Veevid

Remote MCP server for Veevid AI: list models, check credit cost, and generate AI videos, images and audio from Claude Code, Cursor or Claude Desktop.

Xem trên veevid.ai

Máy chủ MCP được lưu trữ

npx add-mcp 'https://veevid.ai/api/mcp'

Cài vào Claude Code, Codex, Cursor và nhiều công cụ khác

Tài liệu

Quick Start

1. Get Your API Key

Create an API key at veevid.ai/settings/api-keys. Save it securely — it's only shown once.

mkdir -p ~/.config/veevid
echo "vv_sk_your_key_here" > ~/.config/veevid/api_key

2. Get a Quote

Before generating, check the exact credit cost. A quote is free and never starts a generation:

curl -X POST https://veevid.ai/api/quote \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "mode": "veo3",
    "generation_type": "text-to-video",
    "video_quality": "standard"
  }'

Response:

{
  "required_credits": 30,
  "current_balance": 451,
  "sufficient": true
}

3. Generate a Video

curl -X POST https://veevid.ai/api/generate-video \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "A golden retriever running through sunflowers, cinematic lighting",
    "mode": "veo3",
    "generation_type": "text-to-video",
    "aspect_ratio": "16:9",
    "video_quality": "standard"
  }'

Response:

{
  "generation_id": "vg_abc123",
  "status": "processing",
  "video_url": null,
  "image_url": null,
  "provider_task_id": "...",
  "video_quality": "standard"
}

The same endpoint generates images and audio too — pick a mode and generation_type from Models.

4. Check Status

curl https://veevid.ai/api/video-generation/GENERATION_ID/status \
  -H "Authorization: Bearer YOUR_API_KEY"

When complete:

{
  "id": "vg_abc123",
  "status": "completed",
  "media_kind": "video",
  "video_url": "https://cdn.veevid.ai/generated-videos/...mp4",
  "credits_used": 30,
  ...
}

Typical generation time: 60–180 seconds for videos, a few seconds to a minute for images and audio.

Authentication

All API requests require a Bearer token:

Authorization: Bearer vv_sk_your_key_here

Keys start with vv_sk_ prefix
Create and manage keys at /settings/api-keys (key management is only available while signed in on the website, not via the API)
Each key shares your account's credit balance
Maximum 5 active keys per account

Security tips

Never expose your API key in client-side code
Use environment variables or secure config files
Rotate keys if you suspect a leak

MCP Server

Veevid is also a remote MCP (Model Context Protocol) server, so AI assistants such as Claude Code, Cursor and Claude Desktop can list models, quote, generate and fetch results for you. The MCP tools run the same endpoints as the REST API on this page — same API key, same models, same credit balance and the same credit cost as on the website (see pricing).

Setting	Value
Endpoint	`https://veevid.ai/api/mcp`
Transport	Streamable HTTP (POST, stateless)
Authentication	Header `Authorization: Bearer vv_sk_your_key_here`. API keys only — there is no OAuth sign-in, so use one of the setups below rather than a sign-in-based connector.

1. Get an API Key

Sign in on the website and create a key at veevid.ai/settings/api-keys. Save it securely — it's only shown once. In the examples below, replace vv_sk_your_key_here with your key.

2. Connect Your Client

Claude Code

claude mcp add --transport http veevid https://veevid.ai/api/mcp \
  --header "Authorization: Bearer vv_sk_your_key_here"

The server is added to the current project. Add --scope user to use it in all your projects.

Cursor

Add this to ~/.cursor/mcp.json (all projects) or .cursor/mcp.json in a project:

{
  "mcpServers": {
    "veevid": {
      "url": "https://veevid.ai/api/mcp",
      "headers": {
        "Authorization": "Bearer vv_sk_your_key_here"
      }
    }
  }
}

To keep the key out of the file, write "Bearer ${env:VEEVID_API_KEY}" and set VEEVID_API_KEY in your environment.

Claude Desktop

Claude Desktop's config file runs local servers, so connect through the mcp-remote bridge (requires Node.js). Open Settings → Developer → Edit Config (claude_desktop_config.json), add the server, then fully quit and restart Claude Desktop:

{
  "mcpServers": {
    "veevid": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://veevid.ai/api/mcp",
        "--header",
        "Authorization:${VEEVID_AUTH}"
      ],
      "env": {
        "VEEVID_AUTH": "Bearer vv_sk_your_key_here"
      }
    }
  }
}

Keep Authorization:${VEEVID_AUTH} without a space and put the key in env — spaces inside args break on Windows.

3. Tools

get_quote and generate take mode, generation_type and params, where params holds every other field of the request body (prompt, duration, aspect_ratio, video_quality, image, images, video, audio…) exactly as listed under Models. Input files are https URLs. Results are JSON. A failed call is marked as an error and carries an error message, plus http_status when the underlying API endpoint rejected the request.

Tool	What it does
`list_models`	Lists models (mode, generation_type, name, category, summary), optionally filtered by category: video, image, audio, video-tool or effect. Pass mode to get that model’s parameters, input files, upload folders, rules and an example request.
`get_credits`	Returns your current credit balance.
`get_quote`	Returns the exact credit cost of a request (required_credits, current_balance, sufficient) without running it. Free.
`generate`	Starts a video, image or audio generation and returns its generation_id. Spends credits; they are refunded automatically if the job fails.
`get_generation`	Returns the status and result of a generation by id — video_url holds the output file for video, image and audio. Set wait_seconds (up to 40) to wait for it to finish.
`upload_file`	Uploads an image, video or audio file from source_url (https, up to 20MB) or data_base64 (up to 10MB) and returns a url to pass to generate. Use the upload folder list_models gives for the model.
`create_upload_url`	Returns a presigned URL to PUT a large file straight to storage. Currently only for veed-clean-audio (folder remove-background-noise/input, up to 512MB).
`list_creations`	Lists your past generations, newest first. Optional page, page_size (up to 50), generation_type and include_audio.

4. Example: Quote, Generate, Get the Result

Ask your assistant something like "Make an 8-second video of a golden retriever running through sunflowers with Veo 3.1". It calls the tools in this order (numbers below are placeholders):

get_quote — arguments:

{
  "mode": "veo3",
  "generation_type": "text-to-video",
  "params": {
    "prompt": "A golden retriever running through sunflowers, cinematic lighting",
    "video_quality": "standard",
    "resolution": "720p",
    "duration": "8",
    "aspect_ratio": "16:9"
  }
}

Result:

{
  "required_credits": <cost of this request>,
  "current_balance": <your balance>,
  "sufficient": true
}

The server tells the assistant to confirm the cost with you before generating.

generate — the same arguments as get_quote. Result:

{
  "generation_id": "vg_abc123",
  "status": "processing",
  "video_url": null,
  ...
}

get_generation — arguments:

{
  "id": "vg_abc123",
  "wait_seconds": 40
}

Result when completed:

{
  "id": "vg_abc123",
  "status": "completed",
  "media_kind": "video",
  "video_url": "https://cdn.veevid.ai/generated-videos/...mp4",
  "error_message": null,
  "credits_used": <credits charged>,
  ...
}

Videos usually take 1–5 minutes. While status is still "processing", the assistant calls get_generation again. If generate times out, check list_creations before retrying — the job may already have started.

API Reference

Base URL: https://veevid.ai. All endpoints accept and return JSON unless noted.

POST /api/quote

Get the exact credit cost before generating. Send the same body you will send to /api/generate-video, including the prompt — GPT Image, Qwen Image, text-to-speech and music models check it. File URLs are only needed to quote speech-to-text and noise removal (the audio URL); for other models you can quote before uploading. The quote endpoint does not read video URLs and, for some models, does not count the images array, so describe your input files with these fields — each model's rules say which ones it needs:

Quote field	Description
`input_image_count`	Number of images (0–16), for the models whose rules ask for it
`has_video_input`	Always true when you send reference videos, or a source video to edit or extend (not for character swap or the insert-shot source video)
`input_video_count`	Number of reference videos (also read by MiniMax H3 Max and the insert-shot model)
`input_video_duration`	Length in seconds of the input / reference video(s)
`reference_video_duration`	Total length of the reference videos for an inserted shot
has_audio_input, input_audio_count, input_audio_duration	The same for audio inputs

Request:

{
  "mode": "kling-3",
  "generation_type": "text-to-video",
  "duration": "10",
  "model_version": "kling-3-standard",
  "generate_audio": true
}

Response:

{
  "required_credits": 250,
  "current_balance": 451,
  "sufficient": true
}

POST /api/generate-video

Start a generation. Despite the name, this single endpoint handles every model: video, image, audio and video tools. Each model accepts its own parameters — see Models for the full list. Credits are deducted when the job is submitted and refunded automatically if it fails, unless the model's rules say otherwise.

Common Parameters:

Parameter	Type	Description
`mode`	string	Model selector (see Models). Required in practice — the server default is a legacy model.
`generation_type`	string	What to do with the model, e.g. "text-to-video", "image-to-video", "reference-to-video", "text-to-image", "image-to-image", "text-to-speech", "video-upscale". Defaults to "text-to-video".
`prompt`	string	Text prompt (max 20,000 characters; some models allow less). Prompts are sent to the model as-is — the automatic prompt translation on the website is not applied to API calls, so write prompts in English for best results.
`aspect_ratio`	string	"16:9", "9:16", "1:1" etc. — allowed values depend on the model
`duration`	string	Output length in seconds, as a string (model-specific). For tools it is often the length of your input video.
`video_quality`	string	Tier or resolution, depending on the model (e.g. Veo 3.1 tier "lite" / "standard" / "pro"; "720p" / "1080p"; image models "1K" / "2K" / "4K")
`resolution`	string	Output resolution for models that take it separately (e.g. Veo 3.1 "720p" / "1080p" / "4k")
`image`	string	Single image URL (first frame, character photo…)
`images`	string[]	Multiple image URLs (start/end frames, reference images, images to edit)
`video`	string	Single video URL (source video for tools)
`videos`	string[]	Reference video URLs
`audio`	string	Audio URL (driving audio, audio to transcribe…). Some models take an audios array instead.
`model_version`	string	Sub-variant (e.g. "kling-3-pro", "seedance-2.0-fast")

Response:

{
  "generation_id": "vg_abc123",
  "status": "processing",
  "video_url": null,
  "image_url": null,
  "provider_task_id": "...",
  "video_quality": "standard"
}

Most jobs return "processing" — poll the status endpoint. A few return "completed" right away with video_url (videos) or image_url (images).

GET /api/video-generation/{generation_id}/status

Poll for completion. Works for every kind of generation — videos, images and audio.

Response (completed):

{
  "id": "vg_abc123",
  "status": "completed",
  "media_kind": "image",
  "video_url": "https://cdn.veevid.ai/images/...jpg",
  "error_message": null,
  "credits_used": 5,
  "created_at": "2026-10-06T08:00:00.000Z",
  "completed_at": "2026-10-06T08:01:37.000Z",
  "video_metadata": { "content_type": null, "file_name": null, "file_size": null },
  "provider_task_id": "...",
  "video_quality": "1K",
  "result_images": ["https://cdn.veevid.ai/images/...jpg"]
}

Field	Description
`status`	"processing" → "completed" or "failed"
`media_kind`	"video", "image" or "audio"
`video_url`	The result file for every media kind — video, image or audio (for images it is the first image)
`result_images`	All output images, when a model returns more than one
`error_message`	Why the generation failed (credits are normally refunded automatically)
lyrics, transcript, timestamps	Extra audio results: song lyrics (Lyria), the transcript (Scribe V2) and word timestamps (Eleven v4, when requested)
`seedance25_draft`	Only for Seedance 2.5 / Veevid 2.5 Pro 480p drafts: { upgradable, upgrade_credits, expires_at, upgrade_generation_id } — see upgrade-draft below

POST /api/video-generation/{generation_id}/upgrade-draft

Seedance 2.5 and Veevid 2.5 Pro videos generated at 480p are usually drafts (the status response includes seedance25_draft). A completed draft can be upgraded once, within 7 days, to a 1080p version with the same prompt, inputs and seed. No request body is needed; the upgrade is a new generation with its own generation_id to poll. Get its price first by sending only the draft's id to the quote endpoint:

curl -X POST https://veevid.ai/api/quote \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "draft_upgrade_id": "vg_abc123" }'

curl -X POST https://veevid.ai/api/video-generation/vg_abc123/upgrade-draft \
  -H "Authorization: Bearer YOUR_API_KEY"

Response:

{
  "generation_id": "vg_def456",
  "status": "processing",
  "credits_used": 790,
  "draft_id": "vg_abc123"
}

Returns 409 if the draft is not completed yet or was already upgraded, 410 once it has expired, and 502 if the upgrade could not be submitted (the credits are refunded).

GET /api/user/creations

List your generations, newest first.

Query	Description
`page`	Page number, starting at 1 (default 1)
`pageSize`	Items per page (default 12, max 50)
`generationType`	Only return one generation_type, e.g. "text-to-image" (default: all)
`includeAudio`	Audio generations are left out unless you pass includeAudio=1

curl "https://veevid.ai/api/user/creations?page=1&pageSize=20&includeAudio=1" \
  -H "Authorization: Bearer YOUR_API_KEY"

Response:

{
  "creations": [
    {
      "id": "vg_abc123",
      "mode": "veo3",
      "generationType": "text-to-video",
      "status": "completed",
      "prompt": "A golden retriever running through sunflowers...",
      "videoUrl": "https://cdn.veevid.ai/generated-videos/...mp4",
      "creditsUsed": 30,
      "createdAt": "2026-10-06T08:00:00.000Z",
      "mediaKind": "video",
      ...
    }
  ],
  "totalPages": 3,
  "currentPage": 1,
  "totalCount": 41,
  "hasMore": true
}

Note that list items use camelCase field names (videoUrl, generationType), unlike the status endpoint.

GET | DELETE /api/user/creations/{id}

GET returns one generation with the same fields as a list item. DELETE deletes it, including its result files, and returns { "success": true }.

GET /api/credits

Check your current credit balance.

{
  "success": true,
  "credits": 451,
  "balance": 451
}

File Uploads

Image, video and audio inputs are passed as URLs. Most models accept any public https URL, but some tools only accept files uploaded to Veevid (noted in each model's rules). Upload a local file first, then pass the returned url.

POST /api/storage/upload

multipart/form-data with a file and a folder. Use the upload folder listed for the model's input — the folder decides which file types and sizes are accepted (images: JPEG/PNG/WebP, 10MB by default; video and audio folders allow larger files). Requests to this endpoint are capped at 100MB by our hosting platform, even where a folder or model allows more; for bigger files pass a public https URL if the model accepts one.

curl -X POST https://veevid.ai/api/storage/upload \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "file=@./photo.jpg" \
  -F "folder=image-generator"

Response:

{
  "url": "https://cdn.veevid.ai/image-generator/...jpg",
  "key": "image-generator/...jpg"
}

POST /api/storage/presign-upload

Direct upload for large files, currently only for the veed-clean-audio model (folder remove-background-noise/input, up to 512MB). Send { "folder", "filename", "size" } (size in bytes), then PUT the raw file to uploadUrl with exactly the returned contentType as the Content-Type header, and pass url to the model. Accepted extensions:.mp3.wav.m4a.aac.ogg.opus.flac.mp4.mov.m4v.webm. Returns 402 if your balance is below the minimum charge for one minute of audio.

{
  "uploadUrl": "https://...r2.cloudflarestorage.com/...&X-Amz-Signature=...",
  "url": "https://cdn.veevid.ai/remove-background-noise/input/...mp3",
  "key": "remove-background-noise/input/...mp3",
  "contentType": "audio/mpeg"
}

Models

Every model below is called through POST /api/generate-video with its mode and generation_type. Open a model to see its parameters, input files, rules and an example request body. Parameters marked affects price change the credit cost — use /api/quote for the exact amount. Defaults are what the API uses when you leave a field out, which is not always what the website preselects, so send the values you care about explicitly.

Video Models

Text-to-video, image-to-video and reference-to-video generation.

Model	mode	generation_type
Google Veo 3.1	`veo3`	text-to-video, image-to-video, reference-to-video
Grok Imagine	`grok-imagine`	text-to-video, image-to-video
Kling 2.6	`kling-2-6`	text-to-video, image-to-video
Kling 3.0	`kling-3`	text-to-video, image-to-video
LTX 2.5	`ltx-2-5`	text-to-video, image-to-video
LTX 2.3	`ltx-2-3`	text-to-video, image-to-video
Wan Animate	`wan-animate`	image-to-video
Veevid 2.0 Pro	`veevid-2.0-pro`	text-to-video, image-to-video
Veevid 2.5 Pro	`veevid-2.5-pro`	text-to-video, image-to-video
Seedance 2.0	`seedance-2.0`	text-to-video, image-to-video, reference-to-video
Seedance 2.5	`seedance-2.5`	text-to-video, image-to-video, reference-to-video
Wan 2.6	`wan-2-6`	text-to-video, image-to-video, reference-to-video
Wan 2.7	`wan-2.7`	text-to-video, image-to-video, reference-to-video
Wan 3.0	`wan-3.0`	text-to-video, image-to-video, reference-to-video
MiniMax H3	`minimax-h3`	text-to-video, image-to-video, reference-to-video
MiniMax H3 Max	`minimax-h3-max`	text-to-video, image-to-video, reference-to-video
MiniMax H3 Max Camera Controls	`minimax-h3-max-camera`	camera-controls
MiniMax H3 Max Lip Sync	`minimax-h3-max-lip-sync`	lip-sync
Gemini Omni	`gemini-omni`	text-to-video, image-to-video, reference-to-video
Boreal	`creatify-boreal`	text-to-video, image-to-video

Google Veo 3.1 mode: veo3

text-to-video

Generates a 4-8 second video with native audio from a text prompt using Google Veo 3.1 (Lite / Fast / Quality tiers).

Parameter	Values & notes
prompt string · required	Scene description (required, non-empty, max 20000 characters).
video_quality string affects price default "standard"	"lite", "standard", "pro" Tier: lite = Lite, standard = Fast, pro = Quality. Always send `video_quality`.
resolution string affects price default "720p"	"720p", "1080p", "4k" Output resolution.
duration string default "8"	"4", "6", "8" Seconds: "4", "6" or "8" (default 8).
aspect_ratio string default "16:9"	"16:9", "9:16" 16:9 (landscape) or 9:16 (portrait).
watermark string	Optional custom watermark text (max 200 characters).

Rules

resolution, when sent, must be 720p, 1080p or 4k (422 otherwise).
duration, when sent, must be '4', '6' or '8' (400 otherwise); omitted duration becomes '8'.
Send lite, standard or pro: any other accepted video_quality value (e.g. "1080p") is treated as the Quality (pro) tier — set the output size with resolution.
Price depends only on the video_quality tier and resolution; duration and aspect_ratio do not change the price.

Example request body

{
  "mode": "veo3",
  "generation_type": "text-to-video",
  "prompt": "A golden retriever running through sunflowers, cinematic lighting",
  "video_quality": "standard",
  "resolution": "720p",
  "duration": "8",
  "aspect_ratio": "16:9"
}

image-to-video

Animates a first frame (optionally to a last frame) into a 4-8 second Veo 3.1 video with native audio.

Parameter	Values & notes
prompt string · required	Motion/scene description (required, non-empty, max 20000 characters).
images string[] · required	[first_frame_url] or [first_frame_url, last_frame_url]. A single `image` string is also accepted.
video_quality string affects price default "standard"	"lite", "standard", "pro" Tier: lite = Lite, standard = Fast, pro = Quality. Always send `video_quality`.
resolution string affects price default "720p"	"720p", "1080p", "4k" Output resolution; same price table as text-to-video.
duration string default "8"	"4", "6", "8" Seconds: "4", "6" or "8" (default 8).
aspect_ratio string default "16:9"	"16:9", "9:16" 16:9 (landscape) or 9:16 (portrait).
watermark string	Optional custom watermark text (max 200 characters).

Input files

images: 1-2 image URLs: first frame, optional last frame (max 10MB per image)

Upload folders for /api/storage/upload

image: video-generator-veo3

Rules

image or a non-empty images array is required (400 'Image is required for image-to-video generation').
Send 1 or 2 images: the first frame and an optional last frame.
resolution must be 720p/1080p/4k (422) and duration 4/6/8 (400) when sent.
Send lite, standard or pro: any other accepted video_quality value (e.g. "1080p") is treated as the Quality (pro) tier — set the output size with resolution.
Price depends only on the video_quality tier and resolution.
To guide generation with reference images (rather than first/last frames), use generation_type reference-to-video.

Example request body

{
  "mode": "veo3",
  "generation_type": "image-to-video",
  "prompt": "The woman slowly turns toward the camera and smiles",
  "images": [
    "https://example.com/first-frame.jpg"
  ],
  "video_quality": "standard",
  "resolution": "720p",
  "duration": "8",
  "aspect_ratio": "16:9"
}

reference-to-video

Generates an 8-second Veo 3.1 video that keeps subjects consistent with 1-3 reference images.

Parameter	Values & notes
prompt string · required	Scene description (required, non-empty).
images string[] · required	1-3 reference image URLs.
video_quality string affects price default "standard"	"lite", "standard" Only the Lite (lite) and Fast (standard) tiers accept reference images.
resolution string affects price default "720p"	"720p", "1080p", "4k"
duration string default "8"	"8" Locked to 8 seconds.
aspect_ratio string default "16:9"	"16:9" Send 16:9; reference mode supports 16:9 only.
watermark string	Optional custom watermark text (max 200 characters).

Input files

images: 1-3 reference image URLs (max 10MB per image)

Upload folders for /api/storage/upload

image: video-generator-veo3

Rules

images must contain 1-3 URLs (400 'Reference-to-video requires 1-3 images').
video_quality, when sent, must be lite or standard; the Quality tier is rejected with 422.
duration, when sent, must be '8' (422).
resolution must be 720p/1080p/4k (422) when sent.

Example request body

{
  "mode": "veo3",
  "generation_type": "reference-to-video",
  "prompt": "The character from the reference walks through a neon-lit street at night",
  "images": [
    "https://example.com/character.png"
  ],
  "video_quality": "standard",
  "resolution": "720p",
  "duration": "8",
  "aspect_ratio": "16:9"
}

Grok Imagine mode: grok-imagine

text-to-video

Generate a 6-30 second video with synchronized audio from a text prompt using xAI Grok Imagine.

Parameter	Values & notes
prompt string · required	Text prompt (required, non-empty).
duration string affects price default "6"	6–30 Integer seconds sent as a string, 6-30.
video_quality string affects price default "720p"	"480p", "720p" Output resolution: 480p or 720p. Always send `video_quality`.
grokMode string default "normal"	"fun", "normal", "spicy" Grok style mode: fun, normal or spicy.
aspect_ratio string default "16:9"	"2:3", "3:2", "1:1", "9:16", "16:9" Output aspect ratio; use one of the listed values. Always send `aspect_ratio` explicitly (16:9 is used when omitted).

Rules

duration must be an integer string between 6 and 30 (400 otherwise); omitted duration defaults to 6.
video_quality, when sent, must be 480p or 720p (400 otherwise).
grokMode must be fun, normal or spicy; aspect_ratio must be one of the listed values.
prompt must be non-empty.

Example request body

{
  "mode": "grok-imagine",
  "generation_type": "text-to-video",
  "prompt": "A cat DJ spinning records in a neon club, crowd cheering",
  "duration": "6",
  "video_quality": "480p",
  "aspect_ratio": "1:1",
  "grokMode": "normal"
}

image-to-video

Animate one or more images into a 6-30 second video with audio using xAI Grok Imagine.

Parameter	Values & notes
prompt string · required	Text prompt (required, non-empty).
duration string affects price default "6"	6–30 Integer seconds sent as a string, 6-30.
video_quality string affects price default "720p"	"480p", "720p" Output resolution: 480p or 720p. Always send `video_quality`.
grokMode string default "normal"	"fun", "normal", "spicy" Grok style mode: fun, normal or spicy.
aspect_ratio string	"2:3", "3:2", "1:1", "9:16", "16:9" Output aspect ratio; use one of the listed values. When omitted, a single-image request follows the image and a multi-image request uses 16:9.
images array	Optional array of all image URLs when more than one image is used (the first one also goes in `image`).
image string	First image URL. Required unless you send it as images[0].

Input files

image: First image URL.
images: Optional array of up to 7 image URLs when more than one image is used (jpg/png/webp, max 10MB each).

Upload folders for /api/storage/upload

image: video-generator-grok-imagine

Rules

image-to-video requires image or images (400).
Send at most 7 images.
aspect_ratio only applies when more than one image is used; with a single image the output follows the image.
duration must be an integer string between 6 and 30 (400 otherwise); omitted duration defaults to 6.
video_quality, when sent, must be 480p or 720p (400 otherwise).
grokMode must be fun, normal or spicy; aspect_ratio must be one of the listed values.
prompt must be non-empty.

Example request body

{
  "mode": "grok-imagine",
  "generation_type": "image-to-video",
  "prompt": "She turns toward the camera and laughs as the wind blows her hair",
  "image": "https://example.com/portrait.jpg",
  "duration": "6",
  "video_quality": "480p",
  "aspect_ratio": "1:1",
  "grokMode": "normal"
}

Kling 2.6 mode: kling-2-6

text-to-video

Generates a cinematic video with optional synced native audio from a text prompt using Kling 2.6.

Parameter	Values & notes
prompt string · required	Text prompt (required, non-empty, max 20000 characters).
negative_prompt string	Optional things to avoid (max 2000 characters).
duration string affects price default "5"	"5", "10" Seconds, sent as a string: "5" or "10".
aspect_ratio string default "16:9"	"16:9", "9:16", "1:1" Output aspect ratio.
generate_audio boolean affects price default true	Generate native audio.

Rules

duration must be '5' or '10'; aspect_ratio must be one of 16:9, 9:16, 1:1.

Example request body

{
  "mode": "kling-2-6",
  "generation_type": "text-to-video",
  "prompt": "A red fox running through a snowy forest at sunrise, cinematic",
  "duration": "5",
  "aspect_ratio": "16:9",
  "generate_audio": true
}

image-to-video

Animates a start image into a video guided by a text prompt, with optional synced native audio, using Kling 2.6.

Parameter	Values & notes
prompt string · required	Text prompt describing the motion (required, non-empty).
image string · required	Public URL of the start frame image.
negative_prompt string	Optional things to avoid (max 2000 characters).
duration string affects price default "5"	"5", "10" Seconds, sent as a string: "5" or "10".
aspect_ratio string default "16:9"	"16:9" Send 16:9; the output follows the input image's aspect ratio.
generate_audio boolean affects price default true	Generate native audio.

Input files

image: 1 image URL used as the start frame (JPEG/PNG/WebP, max 10MB)

Upload folders for /api/storage/upload

image: video-generator-kling26

Rules

image (or images[0]) is required for image-to-video, otherwise 400 'Image is required for image-to-video generation'.
Only one image is used (start frame); there is no end-frame support for Kling 2.6.
duration must be '5' or '10'.

Example request body

{
  "mode": "kling-2-6",
  "generation_type": "image-to-video",
  "prompt": "The woman slowly turns her head and smiles",
  "image": "https://cdn.veevid.ai/sample/motion-control-image-1.webp",
  "duration": "5",
  "aspect_ratio": "16:9",
  "generate_audio": true
}

Kling 3.0 mode: kling-3

text-to-video

Generates a 3-15 second video with optional native audio from a text prompt or a multi-shot prompt list using Kling 3.0 Standard/Pro.

Parameter	Values & notes
prompt string	Single text prompt. Required unless `multi_prompt` is used; send either `prompt` or `multi_prompt`, not both (send `prompt` as an empty string when using `multi_prompt`).
multi_prompt array	Multi-shot prompts: array of {prompt: string (max 5000 characters), duration: string seconds 3-15}. Set `duration` to the sum of item durations.
model_version string affects price default "kling-3-standard"	"kling-3-standard", "kling-3-pro" Standard or Pro quality tier.
duration string affects price default "5"	3–15 Total seconds as a string, integer step 1.
aspect_ratio string default "16:9"	"16:9", "9:16", "1:1" Output aspect ratio.
generate_audio boolean affects price default false	Native audio. When omitted the video is generated without audio (and priced as silent); the website sends true.
shot_type string default "customize"	"customize", "intelligent" 'customize' or 'intelligent'.
cfg_scale number default 0.5	0–1 Prompt adherence, 0-1 (step 0.01).
negative_prompt string	Optional things to avoid (max 2000 characters).

Rules

duration must be an integer 3-15.
aspect_ratio must be one of 16:9, 9:16, 1:1 when provided.
model_version must be kling-3-standard or kling-3-pro, otherwise 400.
prompt and multi_prompt must not both be non-empty; with multi_prompt, set duration to the sum of item durations.

Example request body

{
  "mode": "kling-3",
  "generation_type": "text-to-video",
  "prompt": "A drone shot flying over a neon-lit city at night, rain reflections",
  "model_version": "kling-3-standard",
  "duration": "5",
  "aspect_ratio": "16:9",
  "generate_audio": true,
  "shot_type": "customize",
  "cfg_scale": 0.5
}

image-to-video

Animates a start frame (optionally to an end frame) into a 3-15 second video with optional native audio (Kling 3.0 Standard or Pro).

Parameter	Values & notes
prompt string	Text prompt. Required unless multi_prompt is used; never send both non-empty.
image string	Public URL of the start frame. If images is also sent, images[0] takes precedence.
images array	[start_frame_url] or [start_frame_url, end_frame_url]. images[0] is the start frame (takes precedence over image), images[1] the optional end frame.
multi_prompt array	Array of {prompt, duration} for multi-shot generation. Do not combine with an end frame.
model_version string affects price default "kling-3-standard"	"kling-3-standard", "kling-3-pro"
duration string affects price default "5"	3–15 Total seconds as a string.
aspect_ratio string default "16:9"	"16:9", "9:16", "1:1"
generate_audio boolean affects price default false	Native audio. When omitted the video is generated without audio (and priced as silent); the website sends true.
shot_type string default "customize"	"customize" Send 'customize'.
cfg_scale number default 0.5	0–1
negative_prompt string	Optional negative prompt (max 2000 chars); omit when empty.

Input files

image: 1 image URL (start frame); JPEG/PNG/WebP, max 10MB
images: 1-2 image URLs: [start_frame] or [start_frame, end_frame]; only the first two are used

Upload folders for /api/storage/upload

image: video-generator-kling3
images: video-generator-kling3

Rules

image or images[0] is required, otherwise 400 'Image is required for image-to-video generation'.
duration must be an integer 3-15; aspect_ratio one of 16:9, 9:16, 1:1; model_version kling-3-standard or kling-3-pro.
images[1] is used as the end frame; an end frame cannot be combined with multi_prompt.

Example request body

{
  "mode": "kling-3",
  "generation_type": "image-to-video",
  "prompt": "The character walks toward the camera as the wind blows her hair",
  "image": "https://cdn.veevid.ai/sample/motion-control-image-1.webp",
  "images": [
    "https://cdn.veevid.ai/sample/motion-control-image-1.webp"
  ],
  "model_version": "kling-3-pro",
  "duration": "5",
  "aspect_ratio": "16:9",
  "generate_audio": true,
  "shot_type": "customize",
  "cfg_scale": 0.5
}

LTX 2.5 mode: ltx-2-5

text-to-video

Generates a 720p-4K video with native audio from a text prompt using LTX 2.5 Fast or Pro; with a driving audio file it animates a scene to that audio (audio-to-video).

Parameter	Values & notes
prompt string · required	Text prompt (required, non-empty).
model_version string affects price default "ltx-2.5-fast"	"ltx-2.5-fast", "ltx-2.5-pro" Fast or Pro tier. Fast: 720p-2160p, 6-20s. Pro: 720p/1080p, 6/8/10s.
duration string affects price default "6"	"6", "8", "10", "12", "14", "16", "18", "20" Seconds, sent as a string. Pro supports only 6/8/10; 12-20 require ltx-2.5-fast. Ignored when `audio` is sent (the clip is as long as the audio).
video_quality string affects price default "1080p"	"720p", "1080p", "1440p", "2160p" Output resolution. Pro supports only 720p/1080p. Fast longer than 10s supports only 720p/1080p. Ignored when `audio` is sent (always 1080p).
aspect_ratio string default "16:9"	"16:9", "9:16" 16:9 or 9:16 ('auto' is also accepted when `audio` is sent). 16:9 when omitted.
target_fps number default 25	24, 25, 48, 50 Frames per second. Pro: 24, 25 or 50. Fast: 24, 25, 48 or 50; Fast longer than 10s only 24 or 25. Ignored when `audio` is sent.
generate_audio boolean default true	Generate a native audio track. Same price either way. Ignored when `audio` is sent.
camera_motion string	"dolly_in", "dolly_out", "dolly_left", "dolly_right", "jib_up", "jib_down", "static", "focus_shift" Optional camera move. Omit to let the prompt decide. Not allowed together with `audio` (400).
audio string affects price	Optional driving audio (2-20s; Pro 2-10s). Switches to audio-to-video: the clip follows the audio and is as long as it, always at 1080p. The price then depends on the audio length. Must be a file uploaded with `POST /api/storage/upload` to folder 'ltx25/audio'; any other URL is rejected with 400 before charging.
input_audio_duration number affects price	2–20 Length of the driving audio in seconds. The server measures the uploaded file and uses the measured length (your value is used only when it is within 0.15s of the measurement). Needed only to quote before uploading the audio.
guidance_scale number	1–50 Audio-to-video only (400 without `audio`). Omit to use the model default.

Input files

audio: Optional: 1 driving audio file (2-20s, Pro 2-10s; max 15MB), uploaded to folder 'ltx25/audio'

Upload folders for /api/storage/upload

audio: ltx25/audio

Rules

text-to-video does not accept images (400); use image-to-video.
duration must be one of '6','8','10' for ltx-2.5-pro and '6'-'20' (even) for ltx-2.5-fast (400 otherwise).
video_quality must be 720p or 1080p for ltx-2.5-pro; 720p, 1080p, 1440p or 2160p for ltx-2.5-fast (400 otherwise).
ltx-2.5-fast longer than 10s supports only 720p/1080p and 24/25 fps (400 otherwise).
Invalid combinations are rejected with 400 before charging; nothing is silently downgraded.
audio must be uploaded first with POST /api/storage/upload (folder 'ltx25/audio'); external URLs and files in other folders are rejected with 400 before charging.
If the server cannot read the length of the uploaded audio file, the request is rejected with 400; re-encode it as MP3 or WAV and upload again.
Audio must be 2-20s long (Pro: 2-10s), measured on the server (400 otherwise).
With audio: duration, video_quality, target_fps and generate_audio are ignored; camera_motion and an end image are rejected (400).
To quote audio-to-video, send the uploaded audio URL (the quote measures the file). Before uploading, send has_audio_input: true and input_audio_duration instead.
Generation is async; poll by generation_id.

Example request body

{
  "mode": "ltx-2-5",
  "generation_type": "text-to-video",
  "prompt": "A slow cinematic dolly shot through a neon-lit rainy street at night",
  "model_version": "ltx-2.5-fast",
  "duration": "6",
  "video_quality": "720p",
  "aspect_ratio": "16:9",
  "target_fps": 25,
  "generate_audio": true
}

image-to-video

Animates a start image (optionally toward an end image) into a 720p-4K video with native audio using LTX 2.5 Fast or Pro; with a driving audio file it animates the image to that audio (audio-to-video).

Parameter	Values & notes
prompt string	Text prompt describing the motion. Required unless `audio` is sent.
image string	Public URL of the start (first-frame) image. Required unless you send it as images[0]; when both are sent they are merged (image first, duplicates removed).
images string[]	[start_url] or [start_url, end_url]. The second image is used as the end frame (not allowed with `audio`). More than 2 images is rejected (400).
model_version string affects price default "ltx-2.5-fast"	"ltx-2.5-fast", "ltx-2.5-pro" Fast or Pro tier. Fast: 720p-2160p, 6-20s. Pro: 720p/1080p, 6/8/10s.
duration string affects price default "6"	"6", "8", "10", "12", "14", "16", "18", "20" Seconds, sent as a string. Pro supports only 6/8/10; 12-20 require ltx-2.5-fast. Ignored when `audio` is sent (the clip is as long as the audio).
video_quality string affects price default "1080p"	"720p", "1080p", "1440p", "2160p" Output resolution. Pro supports only 720p/1080p. Fast longer than 10s supports only 720p/1080p. Ignored when `audio` is sent (always 1080p).
aspect_ratio string default "16:9"	"auto", "16:9", "9:16" 'auto' follows the start image; otherwise 16:9 or 9:16. 16:9 when omitted (send 'auto' to keep the image's framing).
target_fps number default 25	24, 25, 48, 50 Frames per second. Pro: 24, 25 or 50. Fast: 24, 25, 48 or 50; Fast longer than 10s only 24 or 25. Ignored when `audio` is sent.
generate_audio boolean default true	Generate a native audio track. Same price either way. Ignored when `audio` is sent.
camera_motion string	"dolly_in", "dolly_out", "dolly_left", "dolly_right", "jib_up", "jib_down", "static", "focus_shift" Optional camera move. Omit to let the prompt decide. Not allowed together with `audio` (400).
audio string affects price	Optional driving audio (2-20s; Pro 2-10s). Switches to audio-to-video: the clip follows the audio and is as long as it, always at 1080p. The price then depends on the audio length. Must be a file uploaded with `POST /api/storage/upload` to folder 'ltx25/audio'; any other URL is rejected with 400 before charging.
input_audio_duration number affects price	2–20 Length of the driving audio in seconds. The server measures the uploaded file and uses the measured length (your value is used only when it is within 0.15s of the measurement). Needed only to quote before uploading the audio.
guidance_scale number	1–50 Audio-to-video only (400 without `audio`). Omit to use the model default.

Input files

image: 1 image URL used as the first frame
images: Optional: [start, end] - 2 image URLs; the second is used as the last frame
audio: Optional: 1 driving audio file (2-20s, Pro 2-10s; max 15MB), uploaded to folder 'ltx25/audio'

Upload folders for /api/storage/upload

image: video-generator-ltx25
end image: video-generator-ltx25-end
audio: ltx25/audio

Rules

image-to-video requires a start image in image or images (400 'LTX 2.5 image-to-video requires a start image').
Start frame = first image; end frame = second image if present. End image does not change the price.
duration must be one of '6','8','10' for ltx-2.5-pro and '6'-'20' (even) for ltx-2.5-fast (400 otherwise).
video_quality must be 720p or 1080p for ltx-2.5-pro; 720p, 1080p, 1440p or 2160p for ltx-2.5-fast (400 otherwise).
ltx-2.5-fast longer than 10s supports only 720p/1080p and 24/25 fps (400 otherwise).
Invalid combinations are rejected with 400 before charging; nothing is silently downgraded.
audio must be uploaded first with POST /api/storage/upload (folder 'ltx25/audio'); external URLs and files in other folders are rejected with 400 before charging.
If the server cannot read the length of the uploaded audio file, the request is rejected with 400; re-encode it as MP3 or WAV and upload again.
Audio must be 2-20s long (Pro: 2-10s), measured on the server (400 otherwise).
With audio: duration, video_quality, target_fps and generate_audio are ignored; camera_motion and an end image are rejected (400).
To quote audio-to-video, send the uploaded audio URL (the quote measures the file). Before uploading, send has_audio_input: true and input_audio_duration instead.
Generation is async; poll by generation_id.

Example request body

{
  "mode": "ltx-2-5",
  "generation_type": "image-to-video",
  "prompt": "The woman turns toward the camera and smiles as the wind moves her hair",
  "image": "https://example.com/start.jpg",
  "images": [
    "https://example.com/start.jpg"
  ],
  "model_version": "ltx-2.5-fast",
  "duration": "6",
  "video_quality": "1080p",
  "aspect_ratio": "auto",
  "target_fps": 25,
  "generate_audio": true
}

LTX 2.3 mode: ltx-2-3

text-to-video

Generates a high-resolution (1080p-2160p) video with optional native audio from a text prompt using LTX 2.3 or LTX 2.3 Fast.

Parameter	Values & notes
prompt string · required	Text prompt (required, non-empty, max 20000 chars).
model_version string affects price default "ltx-2.3"	"ltx-2.3", "ltx-2.3-fast" Standard or Fast variant. Use one of the listed values.
duration string · required affects price	"6", "8", "10", "12", "14", "16", "18", "20" Seconds, sent as a string. Standard ('ltx-2.3') only supports 6/8/10; 12-20 require 'ltx-2.3-fast'. Always send it explicitly.
video_quality string affects price default "1080p"	"1080p", "1440p", "2160p" Output resolution. Forced to 1080p when model_version is 'ltx-2.3-fast' and duration > 10.
aspect_ratio string default "16:9"	"16:9", "9:16" 16:9 or 9:16 for text-to-video.
target_fps number default 24	24, 25, 48, 50 Frames per second: 24, 25, 48 or 50. Send it explicitly (24 is used when omitted). Forced to 25 when model_version is 'ltx-2.3-fast' and duration > 10.
generate_audio boolean default true	Generate native audio track.

Rules

Always send duration; it must be one of '6','8','10','12','14','16','18','20' (400 otherwise).
model_version 'ltx-2.3' supports only duration 6/8/10; 'ltx-2.3-fast' supports 6-20 (even). Other combinations make the generation fail.
video_quality must be one of 1080p, 1440p, 2160p (400 otherwise); defaults to 1080p.
target_fps must be 24, 25, 48 or 50; other values make the generation fail.
With ltx-2.3-fast and duration > 10, output is always 1080p at 25 fps and priced at 1080p.
aspect_ratio: use one of the listed values.
Generation is async; poll by generation_id.

Example request body

{
  "mode": "ltx-2-3",
  "generation_type": "text-to-video",
  "prompt": "A slow cinematic dolly shot through a neon-lit rainy street at night",
  "model_version": "ltx-2.3",
  "duration": "6",
  "video_quality": "1080p",
  "aspect_ratio": "16:9",
  "target_fps": 25,
  "generate_audio": true
}

image-to-video

Animates a start image (optionally toward an end image) into a high-resolution video with optional native audio using LTX 2.3 or LTX 2.3 Fast.

Parameter	Values & notes
prompt string · required	Text prompt describing the motion (required, non-empty, max 20000 chars).
image string	Public URL of the start (first-frame) image. Required unless you send images; when both are sent, images[0] is used.
images string[]	[start_url] or [start_url, end_url]. images[1], if present, is used as the end frame.
model_version string affects price default "ltx-2.3"	"ltx-2.3", "ltx-2.3-fast" Standard or Fast variant. Use one of the listed values.
duration string · required affects price	"6", "8", "10", "12", "14", "16", "18", "20" Seconds, sent as a string. Standard only supports 6/8/10; 12-20 require 'ltx-2.3-fast'. Always send it explicitly.
video_quality string affects price default "1080p"	"1080p", "1440p", "2160p" Output resolution. Forced to 1080p when Fast and duration > 10.
aspect_ratio string default "16:9"	"auto", "16:9", "9:16" 'auto' follows the input image; otherwise 16:9 or 9:16.
target_fps number default 24	24, 25, 48, 50 Frames per second: 24, 25, 48 or 50. Send it explicitly (24 is used when omitted). Forced to 25 when Fast and duration > 10.
generate_audio boolean default true	Generate native audio track.

Input files

image: 1 image URL used as the first frame (JPEG/PNG/GIF, max 10MB)
images: Optional: [start, end] - 2 image URLs; the second is used as the last frame

Upload folders for /api/storage/upload

image: video-generator-ltx23
end image: video-generator-ltx23-end

Rules

image-to-video requires image or a non-empty images array (400 'Image is required for image-to-video generation').
Start frame = image (or images[0] if image is absent); end frame = images[1] if present.
Always send duration; it must be one of '6','8','10','12','14','16','18','20' (400 otherwise).
model_version 'ltx-2.3' supports only duration 6/8/10; 'ltx-2.3-fast' supports 6-20 (even). Other combinations make the generation fail.
video_quality must be one of 1080p, 1440p, 2160p (400 otherwise); defaults to 1080p.
target_fps must be 24, 25, 48 or 50; other values make the generation fail.
With ltx-2.3-fast and duration > 10, output is always 1080p at 25 fps and priced at 1080p.
aspect_ratio: use one of the listed values.
End image does not change the price.

Example request body

{
  "mode": "ltx-2-3",
  "generation_type": "image-to-video",
  "prompt": "The woman turns toward the camera and smiles as the wind moves her hair",
  "image": "https://example.com/start.jpg",
  "images": [
    "https://example.com/start.jpg"
  ],
  "model_version": "ltx-2.3",
  "duration": "6",
  "video_quality": "1080p",
  "aspect_ratio": "16:9",
  "target_fps": 25,
  "generate_audio": true
}

Wan Animate mode: wan-animate

image-to-video

Animates a character image with the motion of a reference video (animate), or replaces the person in the reference video with the character (replace).

Parameter	Values & notes
image string · required	Public URL of the character image.
video string · required	Public URL of the reference (driving) video, up to 120 seconds.
animationMode string default "animate"	"animate", "replace" 'animate' = character performs the reference video's motion; 'replace' = character replaces the person in the reference video.
video_quality string · required affects price	"480p", "720p" Output resolution. Always send it ('480p' or '720p').
duration string affects price default "5"	1–120 Reference video length in whole seconds (rounded up), as a string. Send the real length of the reference video; it determines the price (5 when omitted).
prompt string	Optional text guidance, max 5000 characters; may be empty.
aspect_ratio string	Optional; can be omitted (output follows the inputs).

Input files

image: 1 character image URL (JPEG/PNG/GIF, max 10MB)
video: 1 reference video URL, max 120 seconds (max 200MB)

Upload folders for /api/storage/upload

image: video-generator
video: wan-animate/reference-videos

Rules

image is required (400 'Character image is required for wan-animate generation').
video is required (400 'Reference video is required for wan-animate generation').
When sent, duration must be an integer > 0 and <= 120 (400).
prompt, if sent, must be <= 5000 characters (400).
generation_type must be 'image-to-video'.
Always send video_quality ('480p' or '720p').
animationMode must be 'animate' or 'replace'; defaults to 'animate'.
Send duration equal to the reference video length in whole seconds (rounded up).
Generation is async; poll by generation_id.

Example request body

{
  "mode": "wan-animate",
  "generation_type": "image-to-video",
  "prompt": "",
  "image": "https://example.com/character.png",
  "video": "https://example.com/dance.mp4",
  "duration": "8",
  "video_quality": "480p",
  "aspect_ratio": "",
  "animationMode": "animate"
}

Veevid 2.0 Pro mode: veevid-2.0-pro

text-to-video

Veevid 2.0 Pro text-to-video, 4-15s, Mini/Fast/Standard tiers up to 4K with native audio.

Parameter	Values & notes
prompt string · required	Text prompt (max 20000 chars). Required for text-to-video.
model_version string affects price default "seedance-2.0"	"seedance-2.0-mini", "seedance-2.0-fast", "seedance-2.0" Quality tier: Mini, Fast or Standard ('seedance-2.0'). Always send `model_version`; when omitted, Standard is used. Mini/Fast support only 480p/720p.
duration string affects price default "5"	4–15 Output length in seconds as an integer string; values outside 4-15 are clamped to 4-15.
video_quality string affects price default "480p"	"480p", "720p", "1080p", "4k" Output resolution. 1080p/4k are available only with Standard ('seedance-2.0'); with Mini/Fast they are reduced to 720p and priced as 720p. Defaults to 480p; use one of the listed values.
aspect_ratio string default "16:9"	"auto", "21:9", "16:9", "4:3", "1:1", "3:4", "9:16" Output aspect ratio; 'auto' adapts to the content. Use one of the listed values; when omitted, 16:9 is used.
generate_audio boolean default true	Generate native audio. Does not change price.
seed integer default -1	Random seed; -1 means random.

Rules

duration is clamped to 4-15 seconds (for both price and output); no range error is returned.
Always send model_version; values other than the Mini/Fast tiers are treated as Standard ('seedance-2.0').
generate_audio and seed do not affect price.

Example request body

{
  "mode": "veevid-2.0-pro",
  "generation_type": "text-to-video",
  "prompt": "A chef flips a pancake in a sunlit kitchen, slow motion, warm tones",
  "model_version": "seedance-2.0-fast",
  "duration": "5",
  "aspect_ratio": "16:9",
  "video_quality": "480p",
  "generate_audio": true,
  "seed": -1
}

image-to-video

Veevid 2.0 Pro image-to-video with first/optional last frame, 4-15s, Mini/Fast/Standard tiers up to 4K with native audio.

Parameter	Values & notes
prompt string	Optional text prompt (max 20000 chars).
image string	Public https URL of the first-frame image. Required unless you send it as images[0].
images array<string>	Optional [first_frame_url, last_frame_url]; when 2+ entries are sent, the last one is used as the last frame.
model_version string affects price default "seedance-2.0"	"seedance-2.0-mini", "seedance-2.0-fast", "seedance-2.0" Quality tier: Mini, Fast or Standard ('seedance-2.0'). Always send `model_version`; when omitted, Standard is used. Mini/Fast support only 480p/720p.
duration string affects price default "5"	4–15 Output length in seconds as an integer string; values outside 4-15 are clamped to 4-15.
video_quality string affects price default "480p"	"480p", "720p", "1080p", "4k" Output resolution. 1080p/4k are available only with Standard ('seedance-2.0'); with Mini/Fast they are reduced to 720p and priced as 720p. Defaults to 480p; use one of the listed values.
aspect_ratio string default "16:9"	"auto", "21:9", "16:9", "4:3", "1:1", "3:4", "9:16" Output aspect ratio; 'auto' adapts to the content. Use one of the listed values; when omitted, 16:9 is used.
generate_audio boolean default true	Generate native audio. Does not change price.
seed integer default -1	Random seed; -1 means random.

Input files

image: 1 image URL used as first frame (JPEG/PNG/WebP, max 30MB)
images: optional [first,last] to add a last frame

Upload folders for /api/storage/upload

image: seedance20

Rules

duration is clamped to 4-15 seconds (for both price and output); no range error is returned.
Always send model_version; values other than the Mini/Fast tiers are treated as Standard ('seedance-2.0').
generate_audio and seed do not affect price.

Example request body

{
  "mode": "veevid-2.0-pro",
  "generation_type": "image-to-video",
  "prompt": "A chef flips a pancake in a sunlit kitchen, slow motion, warm tones",
  "model_version": "seedance-2.0-fast",
  "duration": "5",
  "aspect_ratio": "auto",
  "video_quality": "480p",
  "generate_audio": true,
  "seed": -1,
  "image": "https://example.com/first-frame.jpg",
  "images": [
    "https://example.com/first-frame.jpg"
  ]
}

Veevid 2.5 Pro mode: veevid-2.5-pro

text-to-video

Veevid 2.5 Pro text-to-video, one-take videos 4-30s up to 1080p with native audio.

Parameter	Values & notes
prompt string · required	Text prompt (max 20000 chars). A prompt or at least one image is required.
duration string affects price default "5"	4–30 Output length: integer string 4-30, or '-1' to let the model choose the length. Anything else returns 400.
video_quality string affects price default "480p"	"480p", "720p", "1080p" Output resolution (1080p is 10-bit HEVC). Defaults to 480p; use one of the listed values.
aspect_ratio string default "16:9"	"auto", "21:9", "16:9", "4:3", "1:1", "3:4", "9:16" Output aspect ratio; 'auto' adapts to the content. Use one of the listed values; when omitted, 16:9 is used.
output_format string default "mp4"	"mp4", "mov" Container. mov = H.264 yuv444p + PCM audio (poor browser playback). Only mp4/mov are valid for this mode.
generate_audio boolean default true	Generate native audio. Does not change price.
seed integer default -1	Random seed; -1 means random.

Rules

duration must be a pure integer string 4-30 or '-1'; otherwise 400.
output_format must be mp4 or mov for this mode; image formats (PNG/JPEG) return 400.
output_format, generate_audio and seed do not affect price.
480p results are usually drafts: the status response then includes seedance25_draft, and the draft can be upgraded once to a 1080p version within 7 days with POST /api/video-generation/{id}/upgrade-draft.

Example request body

{
  "mode": "veevid-2.5-pro",
  "generation_type": "text-to-video",
  "prompt": "A lighthouse keeper climbs the spiral stairs at dusk, one continuous take",
  "duration": "5",
  "aspect_ratio": "16:9",
  "video_quality": "480p",
  "output_format": "mp4",
  "generate_audio": true,
  "seed": -1
}

image-to-video

Veevid 2.5 Pro image-to-video with first/optional last frame, one-take videos 4-30s up to 1080p with native audio.

Parameter	Values & notes
prompt string	Text prompt (max 20000 chars). A prompt or at least one image is required.
image string	Public https URL of the first-frame image. Required unless you send it as images[0].
images array<string>	Optional [first_frame_url, last_frame_url]; images[0] is used as first frame and the last entry (when 2+) as last frame.
duration string affects price default "5"	4–30 Output length: integer string 4-30, or '-1' to let the model choose the length. Anything else returns 400.
video_quality string affects price default "480p"	"480p", "720p", "1080p" Output resolution (1080p is 10-bit HEVC). Defaults to 480p; use one of the listed values.
aspect_ratio string default "auto"	"auto", "21:9", "16:9", "4:3", "1:1", "3:4", "9:16" Not applied for image-to-video: the output follows the first frame's aspect ratio. Send 'auto'.
output_format string default "mp4"	"mp4", "mov" Container. mov = H.264 yuv444p + PCM audio (poor browser playback). Only mp4/mov are valid for this mode.
generate_audio boolean default true	Generate native audio. Does not change price.
seed integer default -1	Random seed; -1 means random.

Input files

image: 1 image URL used as first frame (JPEG/PNG/WebP, max 30MB)
images: optional [first,last] to add a last frame

Upload folders for /api/storage/upload

image: seedance25

Rules

image (or images) is required for image-to-video, else 400.
For image-to-video the aspect ratio always follows the first frame, regardless of aspect_ratio.
duration must be a pure integer string 4-30 or '-1'; otherwise 400.
output_format must be mp4 or mov for this mode; image formats (PNG/JPEG) return 400.
output_format, generate_audio and seed do not affect price.
480p results are usually drafts: the status response then includes seedance25_draft, and the draft can be upgraded once to a 1080p version within 7 days with POST /api/video-generation/{id}/upgrade-draft.

Example request body

{
  "mode": "veevid-2.5-pro",
  "generation_type": "image-to-video",
  "prompt": "A lighthouse keeper climbs the spiral stairs at dusk, one continuous take",
  "duration": "5",
  "aspect_ratio": "auto",
  "video_quality": "480p",
  "output_format": "mp4",
  "generate_audio": true,
  "seed": -1,
  "image": "https://example.com/first-frame.jpg",
  "images": [
    "https://example.com/first-frame.jpg"
  ]
}

Seedance 2.0 mode: seedance-2.0

text-to-video

Generates a 4-15s video with optional native audio from a text prompt using ByteDance Seedance 2.0 (Mini/Fast/Standard tiers).

Parameter	Values & notes
prompt string · required	Text prompt (max 20000 chars). Required for text-to-video.
model_version string affects price default "seedance-2.0"	"seedance-2.0-mini", "seedance-2.0-fast", "seedance-2.0" Quality tier: 'seedance-2.0-mini', 'seedance-2.0-fast' or 'seedance-2.0' (Standard). Always send `model_version`; when omitted, Standard is used. Mini and Fast support only 480p/720p.
video_quality string affects price default "480p"	"480p", "720p", "1080p", "4k" Output resolution ('4K' is also accepted as 4k). Defaults to 480p; use one of the listed values. 1080p and 4k are only available with 'seedance-2.0'; with Mini/Fast they fall back to 720p.
duration string affects price default "5"	4–15 Output length in seconds as an integer string (e.g. "5"), 4-15. Out-of-range values are clamped to 4-15; non-numeric values fall back to 5.
aspect_ratio string default "16:9"	"auto", "16:9", "9:16", "1:1", "4:3", "3:4", "21:9" Output aspect ratio; 'auto' (or 'adaptive') adapts to the content. Use one of the listed values; when omitted, 16:9 is used.
generate_audio boolean default true	Generate native audio track.
seed integer default -1	Random seed; -1 = random.

Rules

duration must be a JSON string (e.g. "5"); a number returns 400.
duration is clamped to 4-15 seconds (for both price and output) instead of being rejected.
Always send model_version; values other than seedance-2.0 / seedance-2.0-fast / seedance-2.0-mini are treated as seedance-2.0 (Standard).
seedance-2.0-fast and seedance-2.0-mini support only 480p/720p; requested 1080p/4k is reduced to 720p and priced as 720p.
Credits are deducted at submission (402 if insufficient); unused credits are refunded automatically if the generation fails.

Example request body

{
  "mode": "seedance-2.0",
  "generation_type": "text-to-video",
  "prompt": "A red fox running through fresh snow at sunrise, cinematic tracking shot",
  "model_version": "seedance-2.0-fast",
  "video_quality": "480p",
  "duration": "5",
  "aspect_ratio": "16:9",
  "generate_audio": true,
  "seed": -1
}

image-to-video

Animates a first-frame image (optionally with a last-frame image) into a 4-15s video with Seedance 2.0.

Parameter	Values & notes
prompt string	Optional text prompt (max 20000 chars) when image or reference inputs are given.
model_version string affects price default "seedance-2.0"	"seedance-2.0-mini", "seedance-2.0-fast", "seedance-2.0" Quality tier: 'seedance-2.0-mini', 'seedance-2.0-fast' or 'seedance-2.0' (Standard). Always send `model_version`; when omitted, Standard is used. Mini and Fast support only 480p/720p.
video_quality string affects price default "480p"	"480p", "720p", "1080p", "4k" Output resolution ('4K' is also accepted as 4k). Defaults to 480p; use one of the listed values. 1080p and 4k are only available with 'seedance-2.0'; with Mini/Fast they fall back to 720p.
duration string affects price default "5"	4–15 Output length in seconds as an integer string (e.g. "5"), 4-15. Out-of-range values are clamped to 4-15; non-numeric values fall back to 5.
aspect_ratio string default "16:9"	"auto", "16:9", "9:16", "1:1", "4:3", "3:4", "21:9" Output aspect ratio; 'auto' (or 'adaptive') adapts to the content. Use one of the listed values; when omitted, 16:9 is used.
generate_audio boolean default true	Generate native audio track.
seed integer default -1	Random seed; -1 = random.
image string	First-frame image URL. Required unless you send it as images[0].
images array<string>	[first] or [first, last]. images[0] is used as first frame (overrides image), and the last element becomes the last frame when length > 1.

Input files

image: 1 public image URL used as first frame (JPEG/PNG/WebP, max 30MB)
images: optional [first_frame_url, last_frame_url] for first+last-frame mode

Upload folders for /api/storage/upload

image: seedance20

Rules

duration must be a JSON string (e.g. "5"); a number returns 400.
duration is clamped to 4-15 seconds (for both price and output) instead of being rejected.
Always send model_version; values other than seedance-2.0 / seedance-2.0-fast / seedance-2.0-mini are treated as seedance-2.0 (Standard).
seedance-2.0-fast and seedance-2.0-mini support only 480p/720p; requested 1080p/4k is reduced to 720p and priced as 720p.
Credits are deducted at submission (402 if insufficient); unused credits are refunded automatically if the generation fails.
aspect_ratio is applied as sent for image-to-video.
Price is the same as text-to-video (images do not add cost).

Example request body

{
  "mode": "seedance-2.0",
  "generation_type": "image-to-video",
  "prompt": "The woman turns toward the camera and smiles",
  "image": "https://example.com/first.jpg",
  "images": [
    "https://example.com/first.jpg"
  ],
  "model_version": "seedance-2.0-fast",
  "video_quality": "720p",
  "duration": "5",
  "aspect_ratio": "auto",
  "generate_audio": true,
  "seed": -1
}

reference-to-video

Generates a 4-15s video guided by multimodal references (up to 9 images, 3 videos, 3 audio clips) with Seedance 2.0 omni reference.

Parameter	Values & notes
prompt string	Optional text prompt (max 20000 chars) when image or reference inputs are given.
model_version string affects price default "seedance-2.0"	"seedance-2.0-mini", "seedance-2.0-fast", "seedance-2.0" Quality tier: 'seedance-2.0-mini', 'seedance-2.0-fast' or 'seedance-2.0' (Standard). Always send `model_version`; when omitted, Standard is used. Mini and Fast support only 480p/720p.
video_quality string affects price default "480p"	"480p", "720p", "1080p", "4k" Output resolution ('4K' is also accepted as 4k). Defaults to 480p; use one of the listed values. 1080p and 4k are only available with 'seedance-2.0'; with Mini/Fast they fall back to 720p.
duration string affects price default "5"	4–15 Output length in seconds as an integer string (e.g. "5"), 4-15. Out-of-range values are clamped to 4-15; non-numeric values fall back to 5.
aspect_ratio string default "16:9"	"auto", "16:9", "9:16", "1:1", "4:3", "3:4", "21:9" Output aspect ratio; 'auto' (or 'adaptive') adapts to the content. Use one of the listed values; when omitted, 16:9 is used.
generate_audio boolean default true	Generate native audio track.
seed integer default -1	Random seed; -1 = random.
images array<string>	Reference image URLs (role reference_image). Max 9; extras are ignored.
videos array<string> affects price	Reference video URLs (role reference_video), mp4/mov. Max 3; extras are ignored.
input_video_duration number affects price	0–15 Total seconds of all reference videos (rounded up). Required when videos are sent; must equal the total reference video length. >15 is rejected with 400; values below 2 are priced as 2.
audio string	Optional; if sent, set it to the first reference audio URL (audios[0]).
audios array<string>	Reference audio URLs (role reference_audio). Max 3,.wav/.mp3, total <= 15s; extras are ignored. Audio cannot be the only reference: at least one image or video is required.

Input files

images: 0-9 reference image URLs (JPEG/PNG/WebP; max 30MB each)
videos: 0-3 reference video URLs (mp4/mov, max 50MB each, total <= 15s)
audios: 0-3 reference audio URLs (wav/mp3, max 15MB each, total <= 15s); requires at least one image or video

Upload folders for /api/storage/upload

image: seedance20
video: seedance20/reference-videos
audio: seedance20/audio

Rules

duration must be a JSON string (e.g. "5"); a number returns 400.
duration is clamped to 4-15 seconds (for both price and output) instead of being rejected.
Always send model_version; values other than seedance-2.0 / seedance-2.0-fast / seedance-2.0-mini are treated as seedance-2.0 (Standard).
seedance-2.0-fast and seedance-2.0-mini support only 480p/720p; requested 1080p/4k is reduced to 720p and priced as 720p.
Credits are deducted at submission (402 if insufficient); unused credits are refunded automatically if the generation fails.
At least one reference image or video is required (audio-only is not allowed); otherwise the generation fails.
When sending videos, send input_video_duration equal to the total reference video length in seconds (rounded up); > 15 is rejected with 400.
Send at most 9 images, 3 videos and 3 audios; extras are ignored.
Without reference videos (images/audio only), price equals the text-to-video price.
When quoting with /api/quote, add has_video_input: true and input_video_duration if you send reference videos (the quote endpoint does not read video URLs, so the price would come out too low).

Example request body

{
  "mode": "seedance-2.0",
  "generation_type": "reference-to-video",
  "prompt": "The character from image 1 dances in the style of video 1",
  "images": [
    "https://example.com/character.png"
  ],
  "videos": [
    "https://example.com/dance.mp4"
  ],
  "input_video_duration": 6,
  "model_version": "seedance-2.0-fast",
  "video_quality": "480p",
  "duration": "5",
  "aspect_ratio": "auto",
  "generate_audio": true,
  "seed": -1
}

Seedance 2.5 mode: seedance-2.5

text-to-video

Generates a 4-30s one-take video with optional native audio from a text prompt using ByteDance Seedance 2.5.

Parameter	Values & notes
prompt string · required	Text prompt (max 20000 chars). Required for text-to-video.
video_quality string affects price default "480p"	"480p", "720p", "1080p" Output resolution; always send `video_quality` explicitly. 1080p is 10-bit HEVC (may not play in some browsers). '4k'/'4K'/'2160p' are treated as 1080p.
duration string affects price default "5"	4–30 Output seconds as a string containing a plain integer from 4 to 30, or "-1" to let the model choose the length. Anything else (e.g. "3", "31", "5.0", "0x10") is rejected with 400.
aspect_ratio string default "16:9"	"auto", "16:9", "9:16", "1:1", "4:3", "3:4", "21:9" Output aspect ratio; 'auto' lets the model choose. Send `aspect_ratio` explicitly and use one of the listed values.
output_format string default "mp4"	"mp4", "mov" Output container. mov = H.264 yuv444p with PCM audio (often not playable in browsers). Any other value is rejected with 400.
generate_audio boolean default true	Generate a native audio track. Default true.
seed integer default -1	Random seed; -1 = random.

Rules

duration must be a JSON string matching /^-?\d+$/ and equal -1 or 4-30, otherwise 400
output_format must be mp4 or mov (400 otherwise)
Always send video_quality as 480p, 720p or 1080p ('4k' is treated as 1080p)
No model_version parameter (single tier)
Price can be previewed with POST /api/quote; /api/generate-video returns 402 if credits are insufficient. Credits for failed generations are refunded automatically.
480p results are usually drafts: the status response then includes seedance25_draft, and the draft can be upgraded once to a 1080p version within 7 days with POST /api/video-generation/{id}/upgrade-draft.

Example request body

{
  "mode": "seedance-2.5",
  "generation_type": "text-to-video",
  "prompt": "A lighthouse keeper climbs the spiral stairs at dusk, one continuous shot",
  "video_quality": "480p",
  "duration": "5",
  "aspect_ratio": "16:9",
  "output_format": "mp4",
  "generate_audio": true,
  "seed": -1
}

image-to-video

Animates a first-frame image (optionally to a last-frame image) into a 4-30s video with Seedance 2.5; aspect ratio follows the input image.

Parameter	Values & notes
prompt string	Text prompt (max 20000 chars). Optional when a first-frame image is given.
video_quality string affects price default "480p"	"480p", "720p", "1080p" Output resolution; always send `video_quality` explicitly. 1080p is 10-bit HEVC (may not play in some browsers). '4k'/'4K'/'2160p' are treated as 1080p.
duration string affects price default "5"	4–30 Output seconds as a string containing a plain integer from 4 to 30, or "-1" to let the model choose the length. Anything else (e.g. "3", "31", "5.0", "0x10") is rejected with 400.
aspect_ratio string default "auto"	"auto", "16:9", "9:16", "1:1", "4:3", "3:4", "21:9" Ignored for image-to-video: the output aspect ratio follows the first-frame image.
output_format string default "mp4"	"mp4", "mov" Output container. mov = H.264 yuv444p with PCM audio (often not playable in browsers). Any other value is rejected with 400.
generate_audio boolean default true	Generate a native audio track. Default true.
seed integer default -1	Random seed; -1 = random.
image string	First-frame image URL (or provide images[0]).
images array<string>	[first_frame_url] or [first_frame_url, last_frame_url]. images[0] is the first frame (overrides image); when there is more than one entry, the last element is used as the last frame.

Input files

image: 1 public image URL used as the first frame (JPEG/PNG/WEBP; max 30MB)
images: optional [first_frame_url, last_frame_url]

Upload folders for /api/storage/upload

image: seedance25

Rules

duration must be a JSON string matching /^-?\d+$/ and equal -1 or 4-30, otherwise 400
output_format must be mp4 or mov (400 otherwise)
Always send video_quality as 480p, 720p or 1080p ('4k' is treated as 1080p)
No model_version parameter (single tier)
Price can be previewed with POST /api/quote; /api/generate-video returns 402 if credits are insufficient. Credits for failed generations are refunded automatically.
image-to-video requires image or images (400 'Image is required for image-to-video generation')
aspect_ratio has no effect for image-to-video; the output follows the first-frame image
Price is the same as text-to-video (images do not add cost)
480p results are usually drafts: the status response then includes seedance25_draft, and the draft can be upgraded once to a 1080p version within 7 days with POST /api/video-generation/{id}/upgrade-draft.

Example request body

{
  "mode": "seedance-2.5",
  "generation_type": "image-to-video",
  "prompt": "The camera slowly pushes in as leaves drift past",
  "image": "https://example.com/first.jpg",
  "images": [
    "https://example.com/first.jpg",
    "https://example.com/last.jpg"
  ],
  "video_quality": "720p",
  "duration": "8",
  "aspect_ratio": "auto",
  "output_format": "mp4",
  "generate_audio": true,
  "seed": -1
}

reference-to-video

Generates a 4-30s video guided by up to 30 reference images, 10 reference videos and 10 reference audio clips (audio-only allowed) with Seedance 2.5.

Parameter	Values & notes
prompt string	Text prompt (max 20000 chars). Optional when reference assets are given; can refer to them (e.g. 'image 1', 'video 1').
video_quality string affects price default "480p"	"480p", "720p", "1080p" Output resolution; always send `video_quality` explicitly. 1080p is 10-bit HEVC (may not play in some browsers). '4k'/'4K'/'2160p' are treated as 1080p.
duration string affects price default "5"	4–30 Output seconds as a string containing a plain integer from 4 to 30, or "-1" to let the model choose the length. Anything else (e.g. "3", "31", "5.0", "0x10") is rejected with 400.
aspect_ratio string default "16:9"	"auto", "16:9", "9:16", "1:1", "4:3", "3:4", "21:9" Output aspect ratio; 'auto' lets the model choose. Send `aspect_ratio` explicitly and use one of the listed values.
output_format string default "mp4"	"mp4", "mov" Output container. mov = H.264 yuv444p with PCM audio (often not playable in browsers). Any other value is rejected with 400.
generate_audio boolean default true	Generate a native audio track. Default true.
seed integer default -1	Random seed; -1 = random.
images array<string>	Reference image URLs. Max 30.
videos array<string> affects price	Reference video URLs (mp4/mov). Max 10, total length <= 30s. Including reference videos switches to reference-video pricing.
input_video_duration number affects price	0–30 Total length in seconds of all reference videos (round the sum up to a whole second). Max 30; larger values are rejected with 400. Send it equal to the total reference video length whenever `videos` is sent. Minimum billed input length is 2s.
audio string	Optional single reference audio URL; may be the same as audios[0].
audios array<string>	Reference audio URLs (wav/mp3). Max 10, total length <= 30s. Audio-only reference is allowed.

Input files

images: 0-30 reference image URLs (JPEG/PNG/WEBP; max 30MB each)
videos: 0-10 reference video URLs (mp4/mov, max 50MB each, total <= 30s)
audios: 0-10 reference audio URLs (wav/mp3, max 15MB each, total <= 30s)

Upload folders for /api/storage/upload

image: seedance25
video: seedance25/reference-videos
audio: seedance25/audio

Rules

duration must be a JSON string matching /^-?\d+$/ and equal -1 or 4-30, otherwise 400
output_format must be mp4 or mov (400 otherwise)
Always send video_quality as 480p, 720p or 1080p ('4k' is treated as 1080p)
No model_version parameter (single tier)
Price can be previewed with POST /api/quote; /api/generate-video returns 402 if credits are insufficient. Credits for failed generations are refunded automatically.
At least one reference image, video or audio is required; the generation fails otherwise
Send input_video_duration equal to the total reference video length (seconds) whenever videos is sent; values > 30 are rejected with 400
aspect_ratio is honored for reference-to-video
Without reference videos (images/audio only), price equals text-to-video price
480p results are usually drafts: the status response then includes seedance25_draft, and the draft can be upgraded once to a 1080p version within 7 days with POST /api/video-generation/{id}/upgrade-draft.
When quoting with /api/quote, add has_video_input: true and input_video_duration if you send reference videos (the quote endpoint does not read video URLs, so the price would come out too low).

Example request body

{
  "mode": "seedance-2.5",
  "generation_type": "reference-to-video",
  "prompt": "The product from image 1 rotates on a marble table, lit like video 1",
  "images": [
    "https://example.com/product.png"
  ],
  "videos": [
    "https://example.com/lighting.mp4"
  ],
  "input_video_duration": 8,
  "video_quality": "480p",
  "duration": "10",
  "aspect_ratio": "9:16",
  "output_format": "mp4",
  "generate_audio": true,
  "seed": -1
}

Wan 2.6 mode: wan-2-6

text-to-video

Generates a 5-15 second video with synchronized audio from a text prompt using Wan 2.6.

Parameter	Values & notes
prompt string · required	Text prompt (max 20000 chars). Always send a non-empty prompt.
negative_prompt string	Optional negative prompt (max 2000 chars). Omit when empty.
duration string affects price default "5"	"5", "10", "15" Output seconds, sent as a string.
video_quality string affects price default "720p"	"720p", "1080p" Output resolution. Use one of the listed values.
seed integer default -1	Random seed; -1 = random, otherwise 0-2147483647.
enable_prompt_expansion boolean default true	Let the model rewrite/expand the prompt.
shot_type string default "single"	"single", "multi" multi = multi-shot video.
aspect_ratio string default "16:9"	"16:9", "9:16", "1:1", "4:3", "3:4" Output aspect ratio. Use one of the listed values.
audio string	Optional driving/background audio URL (mp3/ogg/wav/m4a/aac, max 15MB) for audio-visual sync.

Input files

audio: optional 1 audio URL (mp3/ogg/wav/m4a/aac)

Upload folders for /api/storage/upload

audio: wan26/audio

Rules

duration must be one of 5, 10, 15 (default 5)
video_quality and aspect_ratio are not validated; always send one of the listed values

Example request body

{
  "mode": "wan-2-6",
  "generation_type": "text-to-video",
  "prompt": "A red fox running through fresh snow at sunrise, cinematic",
  "duration": "5",
  "video_quality": "720p",
  "aspect_ratio": "16:9"
}

image-to-video

Animates a single image into a 5-15 second video with Wan 2.6.

Parameter	Values & notes
prompt string · required	Text prompt (max 20000 chars). Always send a non-empty prompt.
negative_prompt string	Optional negative prompt (max 2000 chars). Omit when empty.
image string · required	Public https URL of the source image (used as the first frame).
duration string affects price default "5"	"5", "10", "15" Output seconds, sent as a string.
video_quality string affects price default "720p"	"720p", "1080p" Output resolution. Use one of the listed values.
seed integer default -1	Random seed; -1 = random, otherwise 0-2147483647.
enable_prompt_expansion boolean default true	Let the model rewrite/expand the prompt.
shot_type string default "single"	"single", "multi" multi = multi-shot video.
audio string	Optional driving/background audio URL (mp3/ogg/wav/m4a/aac, max 15MB) for audio-visual sync.

Input files

image: 1 image URL (first frame); jpg/png/webp, max 10MB
audio: optional 1 audio URL (mp3/ogg/wav/m4a/aac)

Upload folders for /api/storage/upload

image: video-generator
audio: wan26/audio

Rules

image (or images[0]) is required
aspect_ratio is not used for image-to-video; the output follows the image
duration must be one of 5, 10, 15

Example request body

{
  "mode": "wan-2-6",
  "generation_type": "image-to-video",
  "prompt": "The character turns and smiles at the camera",
  "image": "https://example.com/input.jpg",
  "duration": "5",
  "video_quality": "720p"
}

reference-to-video

Generates a new video guided by 1-3 reference videos (subject/motion reference) with Wan 2.6.

Parameter	Values & notes
prompt string · required	Text prompt (max 20000 chars). Always send a non-empty prompt.
negative_prompt string	Optional negative prompt (max 2000 chars). Omit when empty.
videos array<string> · required	1-3 reference video URLs.
video string	Optional single reference video URL; either `video` or `videos` satisfies the requirement.
duration string affects price default "5"	"5", "10" Output seconds: 5 or 10 (15 is rejected for reference-to-video).
video_quality string affects price default "720p"	"720p", "1080p" Output resolution. Use one of the listed values.
seed integer default -1	Random seed; -1 = random, otherwise 0-2147483647.
enable_prompt_expansion boolean default true	Let the model rewrite/expand the prompt.
shot_type string default "single"	"single", "multi" multi = multi-shot video.
aspect_ratio string default "16:9"	"16:9", "9:16", "1:1", "4:3", "3:4" Output aspect ratio. Use one of the listed values.

Input files

videos: 1-3 reference video URLs (mp4/mov/mkv; max 10MB each)

Upload folders for /api/storage/upload

video: wan26/reference-videos

Rules

At least one reference video (video or videos) is required
At most 3 reference videos
duration 15 is rejected (only 5 or 10)
Audio input is not supported for reference-to-video; do not send audio

Example request body

{
  "mode": "wan-2-6",
  "generation_type": "reference-to-video",
  "prompt": "The character from the reference video dances in a neon-lit street",
  "videos": [
    "https://example.com/ref.mp4"
  ],
  "duration": "5",
  "video_quality": "720p",
  "aspect_ratio": "16:9"
}

Wan 2.7 mode: wan-2.7

text-to-video

Generates a 2-15 second video from a text prompt with Wan 2.7 (smoother motion, better scene fidelity).

Parameter	Values & notes
prompt string · required	Text prompt (max 20000 chars). Always send a non-empty prompt.
negative_prompt string	Optional negative prompt (max 2000 chars). Omit when empty.
duration string affects price default "5"	2–15 Output seconds as an integer string.
video_quality string affects price default "720p"	"720p", "1080p" Output resolution.
aspect_ratio string default "16:9"	"16:9", "9:16", "1:1", "4:3", "3:4" Output aspect ratio.
enable_prompt_expansion boolean default true	Let the model expand the prompt.
audio string	Optional audio URL to drive the video (mp3/ogg/wav/m4a/aac, max 15MB).
seed integer default -1	Random seed; -1 = random.

Input files

audio: optional 1 audio URL (mp3/ogg/wav/m4a/aac)

Upload folders for /api/storage/upload

audio: wan27/audio

Rules

duration must be an integer 2-15 (default 5)
video_quality, if sent, must be 720p or 1080p
aspect_ratio, if sent, must be one of 16:9, 9:16, 1:1, 4:3, 3:4

Example request body

{
  "mode": "wan-2.7",
  "generation_type": "text-to-video",
  "prompt": "A paper boat drifting down a rain-soaked city gutter, macro shot",
  "duration": "5",
  "video_quality": "720p",
  "aspect_ratio": "16:9"
}

image-to-video

Animates a start image (optionally to an end frame) into a 2-15 second video with Wan 2.7.

Parameter	Values & notes
prompt string · required	Text prompt (max 20000 chars). Always send a non-empty prompt.
negative_prompt string	Optional negative prompt (max 2000 chars). Omit when empty.
image string	Start-frame image URL. Required unless you send it as images[0].
images array<string>	Only for first/last-frame mode: [start_image_url, end_image_url]; images[1] becomes the end frame.
duration string affects price default "5"	2–15 Output seconds as an integer string.
video_quality string affects price default "720p"	"720p", "1080p" Output resolution.
enable_prompt_expansion boolean default true	Let the model expand the prompt.
audio string	Optional audio URL to drive the video (mp3/ogg/wav/m4a/aac, max 15MB).
seed integer default -1	Random seed; -1 = random.

Input files

image: 1 start-frame image URL (jpg/png/webp, max 20MB)
images: optional [start, end] pair for first/last-frame interpolation
audio: optional 1 audio URL

Upload folders for /api/storage/upload

image: video-generator-wan27
audio: wan27/audio

Rules

image (or images[0]) is required
aspect_ratio is not needed for image-to-video (the output follows the image); if sent it must be one of the listed values
duration must be an integer 2-15

Example request body

{
  "mode": "wan-2.7",
  "generation_type": "image-to-video",
  "prompt": "Gentle wind moves the hair, camera slowly pushes in",
  "image": "https://example.com/start.jpg",
  "duration": "5",
  "video_quality": "720p"
}

reference-to-video

Generates a video guided by up to 3 reference images and/or up to 3 reference videos with Wan 2.7.

Parameter	Values & notes
prompt string · required	Text prompt (max 20000 chars). Always send a non-empty prompt.
negative_prompt string	Optional negative prompt (max 2000 chars). Omit when empty.
images array<string>	0-3 reference image URLs.
videos array<string>	0-3 reference video URLs.
video string	Optional single reference video URL; may be the same as videos[0].
duration string affects price default "5"	2–15 Output seconds as an integer string.
video_quality string affects price default "720p"	"720p", "1080p" Output resolution; 720p when omitted.
aspect_ratio string default "16:9"	"16:9", "9:16", "1:1", "4:3", "3:4" Output aspect ratio.
shot_type string default "single"	"single", "multi" single or multi-shot output.
seed integer default -1	Random seed; -1 = random (values >= 0 fix the seed).

Input files

images: 0-3 reference image URLs (jpg/png/webp, max 20MB)
videos: 0-3 reference video URLs (mp4/mov, max 100MB each)

Upload folders for /api/storage/upload

image: video-generator-wan27
video: wan27/reference-videos

Rules

At least one reference image or video is required
At most 3 reference images and at most 3 reference videos
duration must be an integer 2-15
audio and enable_prompt_expansion are not supported for reference-to-video

Example request body

{
  "mode": "wan-2.7",
  "generation_type": "reference-to-video",
  "prompt": "The woman from the reference image walks through a sunflower field",
  "images": [
    "https://example.com/ref.jpg"
  ],
  "duration": "5",
  "video_quality": "720p",
  "aspect_ratio": "16:9",
  "shot_type": "single"
}

Wan 3.0 mode: wan-3.0

text-to-video

Generates a 2-30 second video with optional native audio from a text prompt using Wan 3.0.

Parameter	Values & notes
prompt string · required	Text prompt. Required for text-to-video.
duration string affects price default "5"	2–30 Output seconds as an integer string (2-30); -1 (smart duration) is rejected.
video_quality string affects price default "720p"	"480p", "720p", "1080p" Resolution. Use one of the listed values ('4k' is treated as 1080p).
aspect_ratio string default "16:9"	"adaptive", "16:9", "9:16", "1:1", "4:3", "3:4" Output aspect ratio; 'adaptive' lets the model choose. Send `aspect_ratio` explicitly and use one of the listed values.
generate_audio boolean default true	Generate a synchronized soundtrack (default true; send false to turn it off). Does not change the price.
seed integer	Accepted but currently not passed to the model.

Rules

duration must be an integer 2-30; -1 (smart duration) is rejected
generate_audio does not change the price

Example request body

{
  "mode": "wan-3.0",
  "generation_type": "text-to-video",
  "prompt": "A drone shot over a misty mountain lake at dawn, birds taking off",
  "duration": "5",
  "video_quality": "720p",
  "aspect_ratio": "adaptive",
  "generate_audio": true
}

image-to-video

Animates a first frame (optionally to a last frame) into a 2-30 second video with Wan 3.0.

Parameter	Values & notes
prompt string	Motion/scene prompt (optional for image-to-video).
image string	First-frame image URL. Required unless you send it as images[0].
images array<string>	[first_frame] or [first_frame, last_frame]; images[0] is the first frame and the last element (when 2 or more) is the last frame.
duration string affects price default "5"	2–30 Output seconds as an integer string (2-30); -1 (smart duration) is rejected.
video_quality string affects price default "720p"	"480p", "720p", "1080p" Resolution. Use one of the listed values ('4k' is treated as 1080p).
aspect_ratio string default "16:9"	"adaptive", "16:9", "9:16", "1:1", "4:3", "3:4" Output aspect ratio; 'adaptive' lets the model choose. Send `aspect_ratio` explicitly and use one of the listed values.
generate_audio boolean default true	Generate a synchronized soundtrack (default true; send false to turn it off). Does not change the price.
seed integer	Accepted but currently not passed to the model.

Input files

image: 1 first-frame image URL (jpg/png/webp/bmp/tiff, max 20MB)
images: optional [first, last] frame pair

Upload folders for /api/storage/upload

image: wan30

Rules

image or images is required
files (document) and links (webpage) cannot be combined with image-to-video
duration must be an integer 2-30; -1 rejected

Example request body

{
  "mode": "wan-3.0",
  "generation_type": "image-to-video",
  "prompt": "The cat slowly opens its eyes and yawns",
  "image": "https://example.com/first.jpg",
  "images": [
    "https://example.com/first.jpg"
  ],
  "duration": "5",
  "video_quality": "720p",
  "aspect_ratio": "adaptive",
  "generate_audio": true
}

reference-to-video

Generates a 2-30 second video from any mix of reference images, videos, audio, one document (document-to-video) or one webpage (webpage-to-video) with Wan 3.0.

Parameter	Values & notes
prompt string	Instruction describing how to use the references (e.g. refer to 'Image 1', 'Video 1').
images array<string>	0-10 reference image URLs.
videos array<string>	0-5 reference video URLs, total length <= 15s. Always pass reference videos in `videos` (plural), together with `input_video_duration`.
input_video_duration number affects price	0–15 Total seconds of all reference videos (round the sum up to a whole second; max 15). Send it equal to the total reference video length whenever `videos` is sent.
audios array<string>	0-5 reference audio URLs (wav/mp3), total <= 15s.
audio string	Single reference audio URL; used only when `audios` is absent.
files array<string>	Document-to-video: at most 1 document URL (.doc.docx.xls.xlsx.ppt.pptx.pdf.txt.md.key.pages.numbers; <=100MB, <=50 pages). Mutually exclusive with links.
links array<string>	Webpage-to-video: at most 1 public webpage URL. Mutually exclusive with files.
duration string affects price default "5"	2–30 Output seconds as an integer string (2-30); -1 (smart duration) is rejected.
video_quality string affects price default "720p"	"480p", "720p", "1080p" Resolution. Use one of the listed values ('4k' is treated as 1080p).
aspect_ratio string default "16:9"	"adaptive", "16:9", "9:16", "1:1", "4:3", "3:4" Output aspect ratio; 'adaptive' lets the model choose. Send `aspect_ratio` explicitly and use one of the listed values.
generate_audio boolean default true	Generate a synchronized soundtrack (default true; send false to turn it off). Does not change the price.
seed integer	Accepted but currently not passed to the model.

Input files

images: 0-10 image URLs (jpg/png/webp/bmp/tiff, max 20MB each)
videos: 0-5 video URLs (mp4/mov, max 100MB each), combined <= 15s
audio: 0-5 audio URLs via audios (wav/mp3, max 15MB each), combined <= 15s
files: 0-1 document URL (100MB, 50 pages max)
links: 0-1 webpage URL

Upload folders for /api/storage/upload

image: wan30
video: wan30/reference-videos
audio: wan30/audio
document: wan30/documents

Rules

At most 10 images, 5 videos, 5 audios, 1 file, 1 link (exceeding returns 400, never truncated)
files and links cannot be sent together
Provide at least one reference image, video, audio, document or webpage
input_video_duration must be <= 15, and input_video_duration + duration must be <= 30 (400 otherwise)
Always pass reference videos in videos (plural) and send input_video_duration equal to their total length (seconds)
duration must be an integer 2-30; -1 rejected
When quoting with /api/quote, add has_video_input: true and input_video_duration if you send reference videos (the quote endpoint does not read video URLs, so the price would come out too low).

Example request body

{
  "mode": "wan-3.0",
  "generation_type": "reference-to-video",
  "prompt": "Turn this product page into a 10 second promo video",
  "links": [
    "https://example.com/product"
  ],
  "duration": "10",
  "video_quality": "720p",
  "aspect_ratio": "16:9",
  "generate_audio": true
}

MiniMax H3 mode: minimax-h3

text-to-video

Generates a 4-15 second video from a text prompt with MiniMax H3 (Hailuo 03) at 768P or 2K.

Parameter	Values & notes
prompt string · required	Text prompt (max 20000 chars). Always send a non-empty prompt.
duration string affects price default "5"	4–15 Output seconds as an integer string.
video_quality string affects price default "768p"	"768p", "2k" Resolution. '2K', '1440p', '2160p' are treated as 2k. Use one of the listed values.
aspect_ratio string default "16:9"	"21:9", "16:9", "4:3", "1:1", "3:4", "9:16" Output aspect ratio. 'adaptive' is not supported for text-to-video; use one of the listed values.

Rules

duration must be an integer between 4 and 15 (400 otherwise); sent as a string
Seed, negative prompt and audio options are not supported for this model

Example request body

{
  "mode": "minimax-h3",
  "generation_type": "text-to-video",
  "prompt": "A red fox running through fresh snow at sunrise, cinematic tracking shot",
  "duration": "5",
  "aspect_ratio": "16:9",
  "video_quality": "768p"
}

image-to-video

Animates a first-frame image (optionally toward a last frame) into a 4-15 second MiniMax H3 video at 768P or 2K.

Parameter	Values & notes
prompt string · required	Text prompt describing the motion. Always send a non-empty prompt.
duration string affects price default "5"	4–15 Output seconds as an integer string.
video_quality string affects price default "768p"	"768p", "2k"
image string	First frame image URL. Either image or images[0] must be present.
images string[] affects price	[first_frame] or [first_frame, last_frame]. images[0] is the first frame; with 2 or more entries the LAST element is used as the last frame. Max 9.
aspect_ratio string default "adaptive"	"adaptive", "21:9", "16:9", "4:3", "1:1", "3:4", "9:16" Ignored for image-to-video; the output follows the first frame.

Input files

image: 1 public image URL used as the first frame (jpg/png/webp, max 20MB)
images: Optional array [first_frame_url, last_frame_url]; only the first and last entries are used

Upload folders for /api/storage/upload

image: minimax-h3

Rules

image-to-video requires image or a non-empty images array (400 'Image is required for image-to-video generation')
duration must be an integer 4-15
At most 9 images
The last frame is taken from the last element of images only when images has 2 or more entries
Only the first and last images are used, but every image after the 5th is charged: send at most 2. When quoting with /api/quote, send input_image_count.

Example request body

{
  "mode": "minimax-h3",
  "generation_type": "image-to-video",
  "prompt": "The woman turns toward the camera and smiles as wind moves her hair",
  "duration": "5",
  "aspect_ratio": "adaptive",
  "video_quality": "768p",
  "image": "https://example.com/first-frame.jpg",
  "images": [
    "https://example.com/first-frame.jpg"
  ]
}

reference-to-video

Generates a MiniMax H3 video guided by up to 9 reference images, 3 reference videos and 3 reference audio clips.

Parameter	Values & notes
prompt string · required	Text prompt; can refer to the reference assets. Always send a non-empty prompt.
duration string affects price default "5"	4–15 Output seconds as an integer string.
video_quality string affects price default "768p"	"768p", "2k"
aspect_ratio string default "16:9"	"adaptive", "21:9", "16:9", "4:3", "1:1", "3:4", "9:16" 'adaptive' lets the model choose from the references. Use one of the listed values.
images string[] affects price	Reference image URLs, max 9.
videos string[] affects price	Reference video URLs, max 3, total duration <= 15s. Singular 'video' is accepted as a fallback when 'videos' is absent.
input_video_duration number affects price	0–15 Total seconds of the reference videos (max 15). Send it equal to the total reference video length whenever videos are sent.
audios string[]	Reference audio URLs, max 3 (free).
audio string	Singular reference audio; used only if 'audios' is absent.

Input files

images: 0-9 reference image URLs (jpg/png/webp/bmp/tiff, 20MB each)
video: 0-3 reference video URLs via 'videos' (mp4/mov, 100MB each, total <= 15s)
audio: 0-3 reference audio URLs via 'audios' (wav/mp3, 15MB each); cannot be the only reference

Upload folders for /api/storage/upload

image: minimax-h3
video: minimax-h3/reference-videos
audio: minimax-h3/audio

Rules

At most 9 reference images, 3 reference videos, 3 reference audio files (400 if exceeded, no silent truncation)
At least one reference image or video is required; reference audio cannot be used alone
input_video_duration must be <= 15 seconds in total
Send input_video_duration equal to the total reference video length (seconds) whenever videos are sent
duration must be an integer 4-15
When quoting with /api/quote, add has_video_input: true and input_video_duration if you send reference videos (the quote endpoint does not read video URLs, so the price would come out too low). Also send input_image_count (number of reference images).

Example request body

{
  "mode": "minimax-h3",
  "generation_type": "reference-to-video",
  "prompt": "The character from image 1 walks into the room from video 1 and waves",
  "duration": "5",
  "aspect_ratio": "adaptive",
  "video_quality": "768p",
  "images": [
    "https://example.com/character.png"
  ],
  "videos": [
    "https://example.com/room.mp4"
  ],
  "input_video_duration": 6
}

MiniMax H3 Max mode: minimax-h3-max

text-to-video

Fast (turbo) MiniMax H3 Max text-to-video, 5-15 seconds at 480P/768P/1080P.

Parameter	Values & notes
prompt string · required	Text prompt. Required; always send a non-empty prompt.
duration string affects price default "5"	5–15 Output seconds as an integer string.
video_quality string affects price default "768p"	"480p", "768p", "1080p" Use one of the listed values (2k is not supported).
aspect_ratio string default "16:9"	"21:9", "16:9", "4:3", "1:1", "3:4", "9:16" 'adaptive' is not supported for text-to-video. Use one of the listed values.
prompt_expansion_mode string default "balanced"	"disabled", "fast", "balanced", "quality" Prompt rewriting strength. Values outside the list are rejected with 400.
seed integer	Optional; a value >= 0 fixes the random seed.

Rules

generation_type must be text-to-video, image-to-video or reference-to-video for this mode (400 otherwise)
duration must be an integer between 5 and 15

Example request body

{
  "mode": "minimax-h3-max",
  "generation_type": "text-to-video",
  "prompt": "A paper boat drifting down a rainy city gutter, macro shot",
  "duration": "5",
  "aspect_ratio": "16:9",
  "video_quality": "768p",
  "prompt_expansion_mode": "balanced"
}

image-to-video

Fast MiniMax H3 Max image-to-video from a first frame (optional end frame), 5-15 seconds.

Parameter	Values & notes
prompt string · required	Motion prompt. Required; always send a non-empty prompt.
duration string affects price default "5"	5–15 Output seconds as an integer string.
video_quality string affects price default "768p"	"480p", "768p", "1080p"
image string	First frame URL (image or images[0] required).
images string[]	[first_frame] or [first_frame, end_frame]. With 2 or more entries, the last one is used as the end frame.
aspect_ratio string default "16:9"	"21:9", "16:9", "4:3", "1:1", "3:4", "9:16" Ignored for image-to-video; the output follows the first frame.
prompt_expansion_mode string default "balanced"	"disabled", "fast", "balanced", "quality" Prompt rewriting strength.
seed integer	Optional; a value >= 0 fixes the random seed.

Input files

image: 1 public image URL used as the first frame
images: Optional [first_frame_url, end_frame_url]

Upload folders for /api/storage/upload

image: minimax-h3-max

Rules

image-to-video requires a first frame (image or images[0]); 400 otherwise
duration must be an integer 5-15

Example request body

{
  "mode": "minimax-h3-max",
  "generation_type": "image-to-video",
  "prompt": "The camera slowly pushes in as the dog tilts its head",
  "duration": "5",
  "aspect_ratio": "16:9",
  "video_quality": "768p",
  "prompt_expansion_mode": "balanced",
  "image": "https://example.com/dog.jpg",
  "images": [
    "https://example.com/dog.jpg"
  ]
}

reference-to-video

Fast MiniMax H3 Max video guided by reference images, videos and audio (up to 12 files total).

Parameter	Values & notes
prompt string · required	Text prompt. Required; always send a non-empty prompt.
duration string affects price default "5"	5–15 Output seconds as an integer string.
video_quality string affects price default "768p"	"480p", "768p", "1080p"
aspect_ratio string default "16:9"	"adaptive", "21:9", "16:9", "4:3", "1:1", "3:4", "9:16" 'adaptive' is available only for reference-to-video. Use one of the listed values.
prompt_expansion_mode string default "balanced"	"disabled", "fast", "balanced", "quality" Prompt rewriting strength.
images string[] affects price	Reference image URLs, max 9. Resize images so the long edge is <= 1024 px.
videos string[] affects price	Reference video URLs, max 3, each 2-15s, total <= 15s. Singular 'video' accepted as fallback.
audios string[] affects price	Reference audio URLs, max 3, each 2-15s, total <= 15s. Singular 'audio' accepted as fallback.
input_video_duration number affects price	0–15 Total reference video seconds (max 15). Send it equal to the total reference video length whenever videos are sent.
input_audio_duration number affects price	0–15 Total reference audio seconds (max 15). Send it equal to the total reference audio length whenever audios are sent.
seed integer	Optional; a value >= 0 fixes the random seed.

Input files

images: 0-9 reference image URLs (jpg/png/webp, 20MB, long edge <= 1024 px)
video: 0-3 reference video URLs via 'videos' (mp4/mov, 100MB each, 2-15s each, total <= 15s)
audio: 0-3 reference audio URLs via 'audios' (wav/mp3, 15MB each, 2-15s each, total <= 15s); cannot be the only reference

Upload folders for /api/storage/upload

image: minimax-h3-max
video: minimax-h3-max/reference-videos
audio: minimax-h3-max/audio

Rules

At least one reference image, video or audio is required
Reference audio cannot be used alone; add at least one image or video
Max 9 images, 3 videos, 3 audio files, and max 12 files in total
Reference videos and reference audio each total <= 15s (two separate limits); each clip must be at least 2s
Send input_video_duration / input_audio_duration equal to the total reference video / audio length (seconds)
duration must be an integer 5-15
Resize reference images so the long edge is <= 1024 px
When quoting with /api/quote, describe the references with input_image_count, has_video_input + input_video_count + input_video_duration, and has_audio_input + input_audio_count + input_audio_duration (the quote endpoint does not read video URLs, so the price would come out too low).

Example request body

{
  "mode": "minimax-h3-max",
  "generation_type": "reference-to-video",
  "prompt": "The product from image 1 rotates on a marble table, soft studio light",
  "duration": "5",
  "aspect_ratio": "adaptive",
  "video_quality": "768p",
  "prompt_expansion_mode": "balanced",
  "images": [
    "https://example.com/product.jpg"
  ]
}

MiniMax H3 Max Camera Controls mode: minimax-h3-max-camera

camera-controls

Turns a single image into a 5-15 second video with a scripted camera move (orbit, dolly, crane, etc.) using MiniMax H3 Max.

Parameter	Values & notes
image string	Source image URL (`image` or `images[0]`). Output aspect ratio follows the image.
images string[]	Optional; if sent, must contain at most 1 entry (the source image). No end frame.
camera_preset string default "orbit-360"	"orbit-360", "birds-eye", "whip-spin", "hero-reveal", "dolly-in", "dolly-out", "arc-left", "arc-right", "worms-eye" Named trajectory. Unknown names are rejected (400). Ignored when camera_trajectory is provided. Defaults to orbit-360 when both are omitted.
camera_trajectory array	Explicit keyframes. Array of 2-12 objects {time: number 0..1 strictly increasing, azimuth: degrees (signed, multi-turn allowed), elevation: number -90..90, distance: number > 0}. Total \|azimuth\| travel <= 11520 degrees (32 turns). Takes precedence over camera_preset.
prompt string	Optional. If empty, the default instruction 'The entire scene is frozen. Only the camera moves.' is used.
duration string affects price default "5"	5–15 Output seconds as an integer string.
video_quality string affects price default "768p"	"480p", "768p", "1080p" Use one of the listed values.
prompt_expansion_mode string default "balanced"	"disabled", "fast", "balanced", "quality" Prompt rewriting strength.
seed integer	Optional non-negative integer for reproducible results; omit for random.

Input files

image: Exactly 1 public image URL (source image / first frame); no end frame, no reference videos or audio

Upload folders for /api/storage/upload

image: minimax-h3-max

Rules

generation_type must be 'camera-controls' (image-to-video is rejected with 400)
A source image is required; more than 1 image, any video, or any audio is rejected with 400
duration must be an integer 5-15
camera_preset must be one of the 9 preset ids unless camera_trajectory is given
camera_trajectory: 2-12 keyframes, time in [0,1] strictly increasing, elevation in [-90,90], distance > 0, total azimuth travel <= 11520 degrees
aspect_ratio has no effect (output follows the source image)

Example request body

{
  "mode": "minimax-h3-max-camera",
  "generation_type": "camera-controls",
  "prompt": "",
  "duration": "5",
  "video_quality": "768p",
  "prompt_expansion_mode": "balanced",
  "image": "https://example.com/statue.jpg",
  "images": [
    "https://example.com/statue.jpg"
  ],
  "camera_preset": "orbit-360"
}

MiniMax H3 Max Lip Sync mode: minimax-h3-max-lip-sync

lip-sync

Makes a portrait photo speak in sync with an audio track (5-14.8s) using MiniMax H3 Max; output length follows the audio.

Parameter	Values & notes
image string	Portrait image URL (`image` or `images[0]`). Image aspect ratio (width/height) must be between 0.4 and 2.5.
images string[]	Optional; if sent, must contain at most 1 entry (the portrait image).
audios string[] · required	Exactly 1 driving audio URL. Singular `audio` is accepted as a fallback if `audios` is absent.
input_audio_duration number affects price	5–3600 Audio length in seconds. Always send it, equal to the audio length. Billed seconds = ceil(min(input_audio_duration, 14.8)), min 5. Values < 5 are rejected.
video_quality string affects price default "768p"	"480p", "768p", "1080p", "2k" Use one of the listed values.
enable_transcription boolean default true	Transcribe the audio to guide mouth shapes. Defaults to true.
seed integer	Optional non-negative integer for reproducible results; omit for random.

Input files

image: 1 public portrait image URL (aspect ratio 0.4-2.5)
audio: 1 public audio URL via 'audios' (wav/mp3/m4a/aac, up to 15MB); at least 5s; audio over 14.8s is clipped to the first 14.8s

Upload folders for /api/storage/upload

image: minimax-h3-max
audio: minimax-h3-max/audio

Rules

generation_type must be 'lip-sync' (400 otherwise)
Requires one portrait image and one audio track; more than 1 image, more than 1 audio, or any video input is rejected (400)
Portrait image aspect ratio (width/height) must be between 0.4 and 2.5
Send input_audio_duration equal to the audio length; it must be at least 5s (400). Audio longer than 14.8s is allowed and clipped
prompt, duration, aspect_ratio and prompt_expansion_mode have no effect for this mode
Cannot re-sync an existing video; image-to-video lip sync only
If the audio turns out longer than the declared input_audio_duration, the difference is charged after generation.

Example request body

{
  "mode": "minimax-h3-max-lip-sync",
  "generation_type": "lip-sync",
  "image": "https://example.com/portrait.jpg",
  "images": [
    "https://example.com/portrait.jpg"
  ],
  "audios": [
    "https://example.com/speech.mp3"
  ],
  "input_audio_duration": 8.4,
  "video_quality": "768p",
  "enable_transcription": true
}

Gemini Omni mode: gemini-omni

text-to-video

Google Gemini Omni text-to-video with native audio, 4-10s, optionally guided by one reference video.

Parameter	Values & notes
prompt string · required	Text prompt (required, non-empty), max 20000 characters.
duration string affects price default "8"	"4", "6", "8", "10" Seconds. Ignored (and not billed) when a video is supplied; output length then follows the input video clip (max 10s).
video_quality string affects price default "1080p"	"720p", "1080p", "4k" Output resolution. The `resolution` field with the same values is also accepted and takes precedence.
aspect_ratio string default "16:9"	"16:9", "9:16" Only landscape or portrait.
videos string[] affects price	Optional, at most 1 video URL used as motion/style reference.
input_video_duration number	0–30 Length of the supplied input video in seconds. Send it whenever a video is supplied; the output covers the first min(input_video_duration, 10) seconds.
seed integer	0–2147483647 Optional, 0-2147483647; omit for random.

Input files

videos: optional, max 1 video URL (mp4/mov, up to 100MB; input video must be <= 30s)

Upload folders for /api/storage/upload

video: gemini-omni/reference-videos

Rules

duration must be one of 4, 6, 8, 10 (omitted defaults to 8)
video_quality (if sent) must be 720p, 1080p or 4k; aspect_ratio (if sent) must be 16:9 or 9:16
at most 1 video; asset quota images + 2*videos <= 7
input_video_duration must be <= 30s
prompt is required, max 20000 characters
When quoting with /api/quote, add has_video_input: true and input_video_duration if you send a video (the quote endpoint does not read video URLs, so the price would come out too low).

Example request body

{
  "mode": "gemini-omni",
  "generation_type": "text-to-video",
  "prompt": "A golden retriever surfing a wave at sunset, cinematic, with ocean sounds",
  "duration": "8",
  "video_quality": "1080p",
  "aspect_ratio": "16:9"
}

image-to-video

Google Gemini Omni image-to-video: animates up to 7 input images into a 4-10s video with native audio.

Parameter	Values & notes
prompt string · required	Text prompt, max 20000 characters.
images string[] · required	1-7 image URLs. A single `image` field is also accepted.
duration string affects price default "8"	"4", "6", "8", "10" Seconds. Ignored and not billed when a video is supplied.
video_quality string affects price default "1080p"	"720p", "1080p", "4k" Output resolution; `resolution` is also accepted and takes precedence.
aspect_ratio string default "16:9"	"16:9", "9:16"
videos string[] affects price	Optional, at most 1 video URL.
input_video_duration number	0–30 Length of the input video in seconds. Send it whenever a video is supplied; the output covers the first min(input_video_duration, 10) seconds.
seed integer	0–2147483647

Input files

images: 1-7 image URLs (up to 20MB each)
videos: optional, max 1 video URL (mp4/mov, <= 30s, up to 100MB)

Upload folders for /api/storage/upload

image: gemini-omni
video: gemini-omni/reference-videos

Rules

image-to-video requires image or a non-empty images
max 7 images and max 1 video; images + 2*videos <= 7 (so with a video, at most 5 images)
video_quality must be 720p, 1080p or 4k; aspect_ratio must be 16:9 or 9:16
input_video_duration must be <= 30s
same price table as text-to-video
duration must be 4, 6, 8 or 10, also when a video is supplied (it is then ignored).
When quoting with /api/quote, add has_video_input: true and input_video_duration if you send a video (the quote endpoint does not read video URLs, so the price would come out too low).

Example request body

{
  "mode": "gemini-omni",
  "generation_type": "image-to-video",
  "prompt": "The woman in the photo turns toward the camera and smiles as wind moves her hair",
  "images": [
    "https://example.com/portrait.jpg"
  ],
  "duration": "6",
  "video_quality": "1080p",
  "aspect_ratio": "9:16"
}

reference-to-video

Google Gemini Omni reference-to-video: generates a video guided by up to 7 reference images and/or 1 reference video.

Parameter	Values & notes
prompt string · required	Text prompt, max 20000 characters.
images string[]	0-7 reference image URLs. At least one image or one video is required.
videos string[] affects price	0-1 reference video URL.
input_video_duration number	0–30 Length of the reference video in seconds; output length = min(this, 10). Send it whenever a video is supplied.
duration string affects price default "8"	"4", "6", "8", "10" Only used/billed when no video is supplied.
video_quality string affects price default "1080p"	"720p", "1080p", "4k"
aspect_ratio string default "16:9"	"16:9", "9:16"
seed integer	0–2147483647

Input files

images: 0-7 reference image URLs (up to 20MB each)
videos: 0-1 reference video URL (mp4/mov, <= 30s; only the first 10s is used)

Upload folders for /api/storage/upload

image: gemini-omni
video: gemini-omni/reference-videos

Rules

at least one reference image or video is required
reference quota: images <= 7, videos <= 1, images + 2*videos <= 7
input_video_duration must be <= 30s
same price table as text-to-video
duration must be 4, 6, 8 or 10, also when a video is supplied (it is then ignored).
When quoting with /api/quote, add has_video_input: true and input_video_duration if you send a reference video (the quote endpoint does not read video URLs, so the price would come out too low).

Example request body

{
  "mode": "gemini-omni",
  "generation_type": "reference-to-video",
  "prompt": "The character from the reference images walks through a neon-lit Tokyo street at night",
  "images": [
    "https://example.com/character-front.png",
    "https://example.com/character-side.png"
  ],
  "duration": "8",
  "video_quality": "1080p",
  "aspect_ratio": "16:9"
}

Boreal mode: creatify-boreal

text-to-video

Creatify Boreal text-to-video for ads, UGC and presenter clips with native synced speech, 1-20s, optionally driven by an uploaded audio track.

Parameter	Values & notes
prompt string · required	Prompt; may optionally be structured with section headers [VISUAL], [SPEECH], [SOUNDS], [TEXT] (dialogue in quotes).
duration string affects price default "10"	1–20 Integer seconds.
video_quality string affects price default "720p"	"720p", "1080p", "2k" Output resolution ('2K' uppercase also accepted).
aspect_ratio string default "auto"	"auto", "16:9", "9:16", "1:1", "4:3", "3:4"
audio string	Optional driving audio URL; if given, the audio track is kept and the model lip-syncs to it instead of generating speech.

Input files

audio: optional, 1 audio URL (.mp3/.ogg/.wav/.m4a/.aac)

Upload folders for /api/storage/upload

audio: creatify-boreal/audio

Rules

generation_type must be text-to-video or image-to-video
duration must be an integer between 1 and 20 (omitted defaults to 10)
video_quality must be 720p, 1080p, 2k or 2K; aspect_ratio must be auto, 16:9, 9:16, 1:1, 4:3 or 3:4
prompt is required

Example request body

{
  "mode": "creatify-boreal",
  "generation_type": "text-to-video",
  "prompt": "[VISUAL]\nA young woman holds a can of sparkling water in a sunny kitchen and talks to the camera.\n\n[SPEECH]\n\"This is the only drink I grab after my workout.\"",
  "duration": "10",
  "video_quality": "720p",
  "aspect_ratio": "9:16"
}

image-to-video

Creatify Boreal image-to-video: turns a product or presenter image into an ad/UGC clip with native synced speech, optionally driven by audio.

Parameter	Values & notes
prompt string · required	Prompt; optional [VISUAL]/[SPEECH]/[SOUNDS]/[TEXT] sections.
image string · required	1 image URL used as the starting image (`images[0]` is also accepted).
duration string affects price default "10"	1–20 Integer seconds.
video_quality string affects price default "720p"	"720p", "1080p", "2k"
aspect_ratio string default "auto"	"auto", "16:9", "9:16", "1:1", "4:3", "3:4" 'auto' follows the input image.
audio string	Optional driving audio URL; the model lip-syncs to it instead of generating speech.

Input files

image: 1 image URL
audio: optional, 1 audio URL (.mp3/.ogg/.wav/.m4a/.aac)

Upload folders for /api/storage/upload

image: creatify-boreal
audio: creatify-boreal/audio

Rules

image-to-video requires image or non-empty images
duration integer 1-20; video_quality 720p/1080p/2k/2K; aspect_ratio auto/16:9/9:16/1:1/4:3/3:4

Example request body

{
  "mode": "creatify-boreal",
  "generation_type": "image-to-video",
  "prompt": "[VISUAL]\nThe presenter picks up the product and shows it to the camera.\n\n[SPEECH]\n\"Meet the lightest headphones we've ever made.\"",
  "image": "https://example.com/presenter.jpg",
  "duration": "8",
  "video_quality": "1080p",
  "aspect_ratio": "auto"
}

Image Models

Text-to-image and image editing (image-to-image).

Model	mode	generation_type
Google Nano Banana	`nano-banana`	text-to-image
Google Nano Banana (Edit)	`nano-banana-edit`	image-to-image
Google Nano Banana 2.1	`nano-banana-2-1`	text-to-image, image-to-image
Google Nano Banana 2	`nano-banana-2`	text-to-image, image-to-image
Flux Kontext	`flux-kontext`	text-to-image, image-to-image
Krea 3	`krea-3`	text-to-image, image-to-image
FLUX 3 Image	`flux-3-image`	text-to-image, image-to-image
Seedream 5.0	`seedream-5.0-lite`	text-to-image, image-to-image
Seedream 5.0 Pro	`seedream-5.0-pro`	text-to-image, image-to-image
Seedream 5.0 Flash	`seedream-5.0-flash`	text-to-image, image-to-image
Ideogram 4.5	`ideogram-4-5`	text-to-image, image-to-image
GPT Image 2	`gpt-image-2`	text-to-image, image-to-image
GPT Image 2.5	`gpt-image-2-5`	text-to-image, image-to-image
Qwen Image 2.1	`qwen-image-2-1`	text-to-image, image-to-image
Virtual Try-On	`virtual-try-on`	image-to-image
Product Holding	`product-holding`	image-to-image

Google Nano Banana mode: nano-banana

text-to-image

Generates an image from text with Google Nano Banana (or Nano Banana Pro via modelVersion).

Parameter	Values & notes
prompt string · required	Text prompt, up to 20000 characters.
modelVersion string affects price default "nano-banana"	"nano-banana", "nano-banana-pro" nano-banana = base model; nano-banana-pro = Nano Banana Pro.
video_quality string affects price default "standard"	"standard", "1K", "2K", "4K" Send 'standard' with nano-banana; with nano-banana-pro, send the output resolution 1K, 2K or 4K.
aspect_ratio string default "16:9"	"auto", "1:1", "9:16", "16:9", "3:4", "4:3", "3:2", "2:3", "5:4", "4:5", "21:9" Use one of the listed values. nano-banana-pro does not support auto (it produces 1:1).
output_format string default "PNG"	"PNG", "JPEG" Output image format.
duration string default "0"	Send "0" (not used for images).

Rules

aspect_ratio: use one of the listed values; with nano-banana-pro, auto produces 1:1.
Text-to-image does not use images; send images: [] or omit it.
Asynchronous: the response is {generation_id, status:'processing', image_url:null, video_url:null, video_quality}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is then in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "nano-banana",
  "generation_type": "text-to-image",
  "prompt": "A beautiful sunset over mountains with golden light",
  "duration": "0",
  "aspect_ratio": "auto",
  "video_quality": "standard",
  "modelVersion": "nano-banana",
  "output_format": "PNG",
  "images": []
}

Google Nano Banana (Edit) mode: nano-banana-edit

image-to-image

Edits one or more uploaded images from a text instruction with Google Nano Banana (or Nano Banana Pro).

Parameter	Values & notes
prompt string · required	Text prompt, up to 20000 characters.
images string[] · required	Input image URLs to edit (1-10).
modelVersion string affects price default "nano-banana"	"nano-banana", "nano-banana-pro" nano-banana = base model; nano-banana-pro = Nano Banana Pro.
video_quality string affects price default "standard"	"standard", "1K", "2K", "4K" Send 'standard' with nano-banana; with nano-banana-pro, send the output resolution 1K, 2K or 4K.
aspect_ratio string default "16:9"	"auto", "1:1", "9:16", "16:9", "3:4", "4:3", "3:2", "2:3", "5:4", "4:5", "21:9" Use one of the listed values. nano-banana-pro does not support auto (it produces 1:1).
output_format string default "PNG"	"PNG", "JPEG" Output image format.
duration string default "0"	Send "0" (not used for images).

Input files

images: 1-10 image URLs; public https URLs or files uploaded with POST /api/storage/upload

Upload folders for /api/storage/upload

image: image-generator

Rules

aspect_ratio: use one of the listed values; with nano-banana-pro, auto produces 1:1.
image-to-image requires a non-empty images array (or image) (400 otherwise).
Send 1-10 input images.
Asynchronous: the response is {generation_id, status:'processing', image_url:null, video_url:null, video_quality}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is then in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "nano-banana-edit",
  "generation_type": "image-to-image",
  "prompt": "Change the background to a blue sky",
  "duration": "0",
  "aspect_ratio": "auto",
  "video_quality": "standard",
  "modelVersion": "nano-banana",
  "output_format": "PNG",
  "images": [
    "https://example.com/input.jpg"
  ]
}

Google Nano Banana 2.1 mode: nano-banana-2-1

text-to-image

Generates a 1K/2K/4K image from text with Google Nano Banana 2.1.

Parameter	Values & notes
prompt string · required	Text prompt, up to 5000 characters.
resolution string affects price default "1K"	"1K", "2K", "4K" Output tier.
aspect_ratio string default "16:9"	"auto", "1:1", "16:9", "9:16", "4:3", "3:4", "3:2", "2:3", "5:4", "4:5", "21:9", "4:1", "1:4", "8:1", "1:8" auto lets the model pick from the prompt. Always send it (the API default is 16:9).
output_format string default "JPEG"	"JPEG", "PNG" Output format: JPEG or PNG (default JPEG).
duration string default "0"	Send "0" (not used for images).

Rules

Price by resolution only: 1K = 3 credits, 2K = 5, 4K = 8. Aspect ratio and the number of reference images do not change the price.
generation_type must be text-to-image or image-to-image (422).
aspect_ratio must be one of the listed values (400).
resolution must be 1K, 2K or 4K (422).
prompt max 5000 characters (400).
This is a different model from nano-banana-2 (separate mode and price).
Asynchronous: the response is {generation_id, status:'processing', image_url:null, video_url:null, video_quality}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is then in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "nano-banana-2-1",
  "generation_type": "text-to-image",
  "prompt": "A minimalist coffee shop poster with the headline \"SLOW MORNINGS\" in bold serif type",
  "duration": "0",
  "aspect_ratio": "1:1",
  "resolution": "1K",
  "output_format": "JPEG",
  "images": []
}

image-to-image

Edits/combines up to 10 reference images from a prompt with Google Nano Banana 2.1 at 1K-4K.

Parameter	Values & notes
prompt string · required	Text prompt, up to 5000 characters.
images string[] · required	1-10 reference image URLs.
resolution string affects price default "1K"	"1K", "2K", "4K" Output tier.
aspect_ratio string default "16:9"	"auto", "1:1", "16:9", "9:16", "4:3", "3:4", "3:2", "2:3", "5:4", "4:5", "21:9", "4:1", "1:4", "8:1", "1:8" auto keeps the aspect ratio of the first reference image. Always send it (the API default is 16:9).
output_format string default "JPEG"	"JPEG", "PNG" Output format: JPEG or PNG (default JPEG).
duration string default "0"	Send "0" (not used for images).

Input files

images: 1-10 reference image URLs

Upload folders for /api/storage/upload

image: image-generator

Rules

Price by resolution only: 1K = 3 credits, 2K = 5, 4K = 8. Aspect ratio and the number of reference images do not change the price.
generation_type must be text-to-image or image-to-image (422).
aspect_ratio must be one of the listed values (400).
resolution must be 1K, 2K or 4K (422).
prompt max 5000 characters (400).
image-to-image requires a non-empty images array (or image) (400 otherwise).
Max 10 reference images (400).
This is a different model from nano-banana-2 (separate mode and price).
Asynchronous: the response is {generation_id, status:'processing', image_url:null, video_url:null, video_quality}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is then in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "nano-banana-2-1",
  "generation_type": "image-to-image",
  "prompt": "A minimalist coffee shop poster with the headline \"SLOW MORNINGS\" in bold serif type",
  "duration": "0",
  "aspect_ratio": "auto",
  "resolution": "1K",
  "output_format": "JPEG",
  "images": [
    "https://example.com/ref.jpg"
  ]
}

Google Nano Banana 2 mode: nano-banana-2

text-to-image

Generates an image from text with Google Nano Banana 2 (1K-4K).

Parameter	Values & notes
prompt string · required	Text prompt, up to 20000 characters.
video_quality string affects price default "1K"	"1K", "2K", "4K"
aspect_ratio string default "16:9"	"1:1", "2:3", "3:2", "3:4", "4:3", "4:5", "5:4", "9:16", "16:9", "21:9", "1:4", "4:1", "1:8", "8:1" Output aspect ratio. Use one of the listed values.
output_format string default "JPG"	"JPG", "PNG" Output format: JPG or PNG (default JPG).
google_search boolean default false	Enable Google Search grounding.
image_search boolean default false	Optional; has no effect for this model and can be omitted.
duration string default "0"	Send "0" (not used for images).

Rules

Use one of the listed values.
aspect_ratio: use one of the listed values.
Asynchronous: the response is {generation_id, status:'processing', image_url:null, video_url:null, video_quality}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is then in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "nano-banana-2",
  "generation_type": "text-to-image",
  "prompt": "A beautiful portrait with stunning details",
  "duration": "0",
  "aspect_ratio": "1:1",
  "video_quality": "1K",
  "output_format": "JPG",
  "google_search": false,
  "image_search": false,
  "images": []
}

image-to-image

Edits/combines reference images from a text prompt with Google Nano Banana 2.

Parameter	Values & notes
prompt string · required	Text prompt, up to 20000 characters.
images string[] · required	Reference image URLs (1-14).
video_quality string affects price default "1K"	"1K", "2K", "4K"
aspect_ratio string default "16:9"	"match_input_image", "1:1", "2:3", "3:2", "3:4", "4:3", "4:5", "5:4", "9:16", "16:9", "21:9", "1:4", "4:1", "1:8", "8:1" Output aspect ratio. 'match_input_image' keeps the input image's ratio. Use one of the listed values.
output_format string default "JPG"	"JPG", "PNG" Output format: JPG or PNG (default JPG).
google_search boolean default false	Enable Google Search grounding.
image_search boolean default false	Optional; has no effect for this model and can be omitted.
duration string default "0"	Send "0" (not used for images).

Input files

images: 1-14 reference image URLs

Upload folders for /api/storage/upload

image: image-generator

Rules

Use one of the listed values.
aspect_ratio: use one of the listed values.
image-to-image requires a non-empty images array (or image) (400 otherwise).
Send 1-14 reference images.
Asynchronous: the response is {generation_id, status:'processing', image_url:null, video_url:null, video_quality}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is then in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "nano-banana-2",
  "generation_type": "image-to-image",
  "prompt": "A beautiful portrait with stunning details",
  "duration": "0",
  "aspect_ratio": "match_input_image",
  "video_quality": "1K",
  "output_format": "JPG",
  "google_search": false,
  "image_search": false,
  "images": [
    "https://example.com/ref.jpg"
  ]
}

Flux Kontext mode: flux-kontext

text-to-image

Generates an image from text with Flux Kontext (Flux 1 or Flux 2, Pro or Max).

Parameter	Values & notes
prompt string · required	Text prompt.
fluxModelVersion string affects price default "flux1"	"flux1", "flux2" Flux Kontext 1 (flux1) or Flux 2 (flux2).
fluxModel string affects price default "flux-kontext-pro"	"flux-kontext-pro", "flux-kontext-max" Pro or Max (with flux2, Max is Flux 2 Flex).
fluxResolution string affects price default "1K"	"1K", "2K" Output resolution; affects price and output only with flux2.
video_quality string default "1K"	"1K", "2K" Send the same value as fluxResolution.
aspect_ratio string default "16:9"	"16:9", "21:9", "4:3", "1:1", "3:4", "9:16", "16:21", "3:2", "2:3", "auto" With flux1: 16:9, 21:9, 4:3, 1:1, 3:4, 9:16, 16:21. With flux2: 1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3, auto.
output_format string	"PNG", "JPEG" Output format, PNG or JPEG (flux1 only). When omitted the model picks the format.
promptUpsampling boolean default false	Prompt upsampling (flux1 only).
enableTranslation boolean default false	Prompt translation (flux1 only).
duration string default "0"	Send "0" (not used for images).

Rules

aspect_ratio: use a value supported by the selected fluxModelVersion.
Asynchronous: the response is {generation_id, status:'processing', image_url:null, video_url:null, video_quality}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is then in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "flux-kontext",
  "generation_type": "text-to-image",
  "prompt": "Serene mountain landscape at sunset with lake reflection",
  "duration": "0",
  "aspect_ratio": "16:9",
  "video_quality": "1K",
  "fluxModel": "flux-kontext-pro",
  "fluxModelVersion": "flux1",
  "fluxResolution": "1K",
  "promptUpsampling": false,
  "enableTranslation": true,
  "output_format": "PNG"
}

image-to-image

Edits an image (or up to 8 references with Flux 2) from a text instruction with Flux Kontext.

Parameter	Values & notes
prompt string · required	Text prompt.
image string	Input image URL (first image); flux1 uses only this. Required unless you send it as images[0].
images string[]	flux2 only: all reference image URLs, up to 8.
fluxModelVersion string affects price default "flux1"	"flux1", "flux2" Flux Kontext 1 (flux1) or Flux 2 (flux2).
fluxModel string affects price default "flux-kontext-pro"	"flux-kontext-pro", "flux-kontext-max" Pro or Max (with flux2, Max is Flux 2 Flex).
fluxResolution string affects price default "1K"	"1K", "2K" Output resolution; affects price and output only with flux2.
video_quality string default "1K"	"1K", "2K" Send the same value as fluxResolution.
aspect_ratio string default "16:9"	"16:9", "21:9", "4:3", "1:1", "3:4", "9:16", "16:21", "3:2", "2:3", "auto" With flux1: 16:9, 21:9, 4:3, 1:1, 3:4, 9:16, 16:21. With flux2: 1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3, auto.
output_format string	"PNG", "JPEG" Output format, PNG or JPEG (flux1 only). When omitted the model picks the format.
promptUpsampling boolean default false	Prompt upsampling (flux1 only).
enableTranslation boolean default false	Prompt translation (flux1 only).
duration string default "0"	Send "0" (not used for images).

Input files

image: 1 image URL (flux1)
images: 1-8 image URLs (flux2 only)

Upload folders for /api/storage/upload

image: flux-kontext

Rules

aspect_ratio: use a value supported by the selected fluxModelVersion.
image-to-image requires a non-empty images array (or image) (400 otherwise).
flux1 uses only the single input image in image; flux2 accepts up to 8 reference images in images.
Asynchronous: the response is {generation_id, status:'processing', image_url:null, video_url:null, video_quality}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is then in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "flux-kontext",
  "generation_type": "image-to-image",
  "prompt": "Serene mountain landscape at sunset with lake reflection",
  "duration": "0",
  "aspect_ratio": "16:9",
  "video_quality": "1K",
  "fluxModel": "flux-kontext-pro",
  "fluxModelVersion": "flux1",
  "fluxResolution": "1K",
  "promptUpsampling": false,
  "enableTranslation": true,
  "output_format": "PNG",
  "image": "https://example.com/input.jpg"
}

Krea 3 mode: krea-3

text-to-image

Generates a 1K or 2K image from text with Krea 3.

Parameter	Values & notes
prompt string · required	Text prompt, up to 5000 characters.
resolution string affects price default "1K"	"1K", "2K" Output tier: 1K (about 1 MP) or 2K (about 4 MP).
aspect_ratio string default "16:9"	"1:1", "16:9", "9:16", "4:3", "3:4", "3:2", "2:3", "5:4", "4:5", "21:9" Output aspect ratio (no auto). Always send it.
output_format string default "JPEG"	"JPEG", "PNG" Output format: JPEG or PNG (default JPEG).
duration string default "0"	Send "0" (not used for images).

Rules

generation_type must be text-to-image or image-to-image (422).
aspect_ratio must be one of the 10 listed values (400). Always send aspect_ratio.
resolution must be 1K or 2K (422).
prompt max 5000 characters (400).
Asynchronous: the response is {generation_id, status:'processing', image_url:null, video_url:null, video_quality}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is then in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "krea-3",
  "generation_type": "text-to-image",
  "prompt": "A cozy bookshop window at dusk with a cat asleep on a stack of books",
  "duration": "0",
  "aspect_ratio": "1:1",
  "resolution": "1K",
  "output_format": "JPEG",
  "images": []
}

image-to-image

Edits or combines 1-8 reference images from a prompt with Krea 3 at 1K or 2K; refer to them as <image-1>...<image-N>.

Parameter	Values & notes
prompt string · required	Text prompt, up to 5000 characters. Refer to reference images as <image-1>, <image-2>,... in upload order.
images string[] · required affects price	1-8 reference image URLs, in the order of <image-1>...<image-N> in the prompt.
resolution string affects price default "1K"	"1K", "2K" Output tier: 1K (about 1 MP) or 2K (about 4 MP).
aspect_ratio string default "16:9"	"1:1", "16:9", "9:16", "4:3", "3:4", "3:2", "2:3", "5:4", "4:5", "21:9" Output aspect ratio (no auto). Always send it.
output_format string default "JPEG"	"JPEG", "PNG" Output format: JPEG or PNG (default JPEG).
duration string default "0"	Send "0" (not used for images).

Input files

images: 1-8 reference image URLs

Upload folders for /api/storage/upload

image: image-generator

Rules

generation_type must be text-to-image or image-to-image (422).
aspect_ratio must be one of the 10 listed values (400). Always send aspect_ratio.
resolution must be 1K or 2K (422).
prompt max 5000 characters (400).
image-to-image requires a non-empty images array (or image) (400 otherwise).
Max 8 reference images (400); each reference image adds one output-size charge.
Asynchronous: the response is {generation_id, status:'processing', image_url:null, video_url:null, video_quality}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is then in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "krea-3",
  "generation_type": "image-to-image",
  "prompt": "Place the ceramic mug from <image-2> on the wooden table in <image-1>. Keep everything else unchanged.",
  "duration": "0",
  "aspect_ratio": "1:1",
  "resolution": "1K",
  "output_format": "JPEG",
  "images": [
    "https://example.com/scene.jpg",
    "https://example.com/mug.jpg"
  ]
}

FLUX 3 Image mode: flux-3-image

text-to-image

Generates a native 1K/2K/4K image from text with FLUX 3.

Parameter	Values & notes
prompt string · required	Text prompt, up to 5000 characters.
resolution string affects price default "1K"	"1K", "2K", "4K" Output tier.
aspect_ratio string affects price default "16:9"	"21:9", "2:1", "16:9", "3:2", "7:5", "4:3", "5:4", "1:1", "4:5", "3:4", "5:7", "2:3", "9:16", "1:2" Output aspect ratio (no auto). Always send it.
output_format string default "JPEG"	"JPEG", "PNG" Output format: JPEG or PNG (default JPEG).
duration string default "0"	Send "0" (not used for images).

Rules

generation_type must be text-to-image or image-to-image (422).
aspect_ratio must be one of the 14 listed values (400). Always send aspect_ratio.
resolution must be 1K, 2K or 4K (422).
prompt max 5000 characters (400).
Asynchronous: the response is {generation_id, status:'processing', image_url:null, video_url:null, video_quality}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is then in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "flux-3-image",
  "generation_type": "text-to-image",
  "prompt": "A bakery storefront at golden hour with a hand-painted sign that reads \"Morning Loaf\"",
  "duration": "0",
  "aspect_ratio": "1:1",
  "resolution": "1K",
  "output_format": "JPEG",
  "images": []
}

image-to-image

Edits/combines up to 10 reference images from a prompt with FLUX 3 at 1K-4K.

Parameter	Values & notes
prompt string · required	Text prompt, up to 5000 characters.
images string[] · required	1-10 reference image URLs.
resolution string affects price default "1K"	"1K", "2K", "4K" Output tier.
aspect_ratio string affects price default "16:9"	"21:9", "2:1", "16:9", "3:2", "7:5", "4:3", "5:4", "1:1", "4:5", "3:4", "5:7", "2:3", "9:16", "1:2" Output aspect ratio (no auto). Always send it.
output_format string default "JPEG"	"JPEG", "PNG" Output format: JPEG or PNG (default JPEG).
duration string default "0"	Send "0" (not used for images).

Input files

images: 1-10 reference image URLs

Upload folders for /api/storage/upload

image: image-generator

Rules

generation_type must be text-to-image or image-to-image (422).
aspect_ratio must be one of the 14 listed values (400). Always send aspect_ratio.
resolution must be 1K, 2K or 4K (422).
prompt max 5000 characters (400).
image-to-image requires a non-empty images array (or image) (400 otherwise).
Max 10 reference images (400); references do not change the price.
Asynchronous: the response is {generation_id, status:'processing', image_url:null, video_url:null, video_quality}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is then in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "flux-3-image",
  "generation_type": "image-to-image",
  "prompt": "A bakery storefront at golden hour with a hand-painted sign that reads \"Morning Loaf\"",
  "duration": "0",
  "aspect_ratio": "1:1",
  "resolution": "1K",
  "output_format": "JPEG",
  "images": [
    "https://example.com/ref.jpg"
  ]
}

Seedream 5.0 mode: seedream-5.0-lite

text-to-image

Generates HD images (optionally a batch) from text with Seedream 5.0 Lite.

Parameter	Values & notes
prompt string · required	Text prompt.
video_quality string default "2K"	"2K", "3K" Output size: 2K or 3K.
aspect_ratio string default "16:9"	"1:1", "2:3", "3:2", "3:4", "4:3", "9:16", "16:9", "21:9" Output aspect ratio. Use one of the listed values and always send it.
output_format string default "PNG"	"PNG", "JPEG" Output format: PNG or JPEG.
sequential_image_generation string default "disabled"	"disabled", "auto" 'auto' enables batch (multi-image) output.
max_images number affects price default 1	1–15 Number of images in a batch. Set it above 1 only together with sequential_image_generation='auto'.
duration string default "0"	Send "0" (not used for images).

Rules

Set max_images > 1 only with sequential_image_generation='auto'.
If a batch returns fewer images than requested, unused credits are refunded automatically.
Always send aspect_ratio, using one of the listed values.
Asynchronous: the response is {generation_id, status:'processing', image_url:null, video_url:null, video_quality}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is then in video_url and all output images are in result_images (array). The status response has no image_url field. Batch results appear in result_images.

Example request body

{
  "mode": "seedream-5.0-lite",
  "generation_type": "text-to-image",
  "prompt": "A cozy reading nook with warm light",
  "duration": "0",
  "aspect_ratio": "1:1",
  "video_quality": "2K",
  "output_format": "PNG",
  "sequential_image_generation": "disabled",
  "max_images": 1,
  "images": []
}

image-to-image

Edits/blends reference images into new HD images (optionally a batch) with Seedream 5.0 Lite.

Parameter	Values & notes
prompt string · required	Text prompt.
images string[] · required	Reference image URLs (1-14).
video_quality string default "2K"	"2K", "3K" Output size: 2K or 3K.
aspect_ratio string default "16:9"	"match_input_image", "1:1", "2:3", "3:2", "3:4", "4:3", "9:16", "16:9", "21:9" Output aspect ratio; 'match_input_image' keeps the input image's ratio. Use one of the listed values and always send it.
output_format string default "PNG"	"PNG", "JPEG" Output format: PNG or JPEG.
sequential_image_generation string default "disabled"	"disabled", "auto" 'auto' enables batch (multi-image) output.
max_images number affects price default 1	1–15 Number of images in a batch. Set it above 1 only together with sequential_image_generation='auto'.
duration string default "0"	Send "0" (not used for images).

Input files

images: 1-14 reference image URLs

Upload folders for /api/storage/upload

image: image-generator

Rules

Set max_images > 1 only with sequential_image_generation='auto'.
If a batch returns fewer images than requested, unused credits are refunded automatically.
Always send aspect_ratio, using one of the listed values.
image-to-image requires a non-empty images array (or image) (400 otherwise).
Send 1-14 reference images.
Asynchronous: the response is {generation_id, status:'processing', image_url:null, video_url:null, video_quality}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is then in video_url and all output images are in result_images (array). The status response has no image_url field. Batch results appear in result_images.

Example request body

{
  "mode": "seedream-5.0-lite",
  "generation_type": "image-to-image",
  "prompt": "A cozy reading nook with warm light",
  "duration": "0",
  "aspect_ratio": "match_input_image",
  "video_quality": "2K",
  "output_format": "PNG",
  "sequential_image_generation": "disabled",
  "max_images": 1,
  "images": [
    "https://example.com/ref.jpg"
  ]
}

Seedream 5.0 Pro mode: seedream-5.0-pro

text-to-image

Generates images with dense text and infographics from a prompt with Seedream 5.0 Pro.

Parameter	Values & notes
prompt string · required	Text prompt, up to 5000 characters.
resolution string affects price default "1K"	"1K", "2K"
video_quality string default "1K"	"1K", "2K" Send the same value as resolution; resolution determines price and output.
aspect_ratio string default "16:9"	"1:1", "4:3", "3:4", "16:9", "9:16", "2:3", "3:2", "21:9" Output aspect ratio; no auto (also for image-to-image). Always send it.
output_format string	"PNG", "JPEG" Output format: PNG or JPEG. When omitted the model picks the format.
duration string default "0"	Send "0" (not used for images).

Rules

aspect_ratio must be one of 1:1, 4:3, 3:4, 16:9, 9:16, 2:3, 3:2, 21:9 (400). Always send aspect_ratio.
resolution must be 1K or 2K (422).
prompt max 5000 characters (400).
Asynchronous: the response is {generation_id, status:'processing', image_url:null, video_url:null, video_quality}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is then in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "seedream-5.0-pro",
  "generation_type": "text-to-image",
  "prompt": "An infographic poster explaining the water cycle with labeled arrows",
  "duration": "0",
  "aspect_ratio": "1:1",
  "video_quality": "1K",
  "resolution": "1K",
  "output_format": "PNG",
  "images": []
}

image-to-image

Precision-edits up to 10 reference images from a prompt with Seedream 5.0 Pro.

Parameter	Values & notes
prompt string · required	Text prompt, up to 5000 characters.
images string[] · required affects price	1-10 reference image URLs.
resolution string affects price default "1K"	"1K", "2K"
video_quality string default "1K"	"1K", "2K" Send the same value as resolution; resolution determines price and output.
aspect_ratio string default "16:9"	"1:1", "4:3", "3:4", "16:9", "9:16", "2:3", "3:2", "21:9" Output aspect ratio; no auto (also for image-to-image). Always send it.
output_format string	"PNG", "JPEG" Output format: PNG or JPEG. When omitted the model picks the format.
duration string default "0"	Send "0" (not used for images).

Input files

images: 1-10 reference image URLs

Upload folders for /api/storage/upload

image: image-generator

Rules

aspect_ratio must be one of 1:1, 4:3, 3:4, 16:9, 9:16, 2:3, 3:2, 21:9 (400). Always send aspect_ratio.
resolution must be 1K or 2K (422).
prompt max 5000 characters (400).
image-to-image requires a non-empty images array (or image) (400 otherwise).
Max 10 reference images (400).
Asynchronous: the response is {generation_id, status:'processing', image_url:null, video_url:null, video_quality}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is then in video_url and all output images are in result_images (array). The status response has no image_url field.
When quoting with /api/quote, send input_image_count (the number of images) — the quote does not count the images array for this model, so it would come out too low.

Example request body

{
  "mode": "seedream-5.0-pro",
  "generation_type": "image-to-image",
  "prompt": "An infographic poster explaining the water cycle with labeled arrows",
  "duration": "0",
  "aspect_ratio": "1:1",
  "video_quality": "1K",
  "resolution": "1K",
  "output_format": "PNG",
  "images": [
    "https://example.com/ref.jpg"
  ]
}

Seedream 5.0 Flash mode: seedream-5.0-flash

text-to-image

Fast, low-cost text-to-image up to 2K with Seedream 5.0 Flash.

Parameter	Values & notes
prompt string · required	Text prompt, max 5000 characters.
resolution string default "1K"	"1K", "1.5K", "2K" Output size (1K, 1.5K or 2K). Set it with `resolution`; `video_quality` is not used.
aspect_ratio string default "16:9"	"1:1", "4:3", "3:4", "16:9", "9:16", "2:3", "3:2", "21:9" Output aspect ratio; no auto option. Always send it explicitly.
output_format string default "PNG"	"PNG", "JPEG" Output format in uppercase: PNG or JPEG.
duration string default "0"	Send "0".

Rules

generation_type must be text-to-image or image-to-image (422).
aspect_ratio must be one of 1:1, 4:3, 3:4, 16:9, 9:16, 2:3, 3:2, 21:9 (400). Always send aspect_ratio explicitly.
resolution must be 1K, 1.5K or 2K (422).
prompt max 5000 characters (400).
Asynchronous: POST returns {generation_id, status:'processing'}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "seedream-5.0-flash",
  "generation_type": "text-to-image",
  "prompt": "A cozy coffee shop interior at golden hour, warm light through large windows",
  "duration": "0",
  "aspect_ratio": "1:1",
  "resolution": "1K",
  "output_format": "PNG",
  "images": []
}

image-to-image

Fast, low-cost edits of up to 10 reference images up to 2K with Seedream 5.0 Flash.

Parameter	Values & notes
prompt string · required	Text prompt, max 5000 characters.
images string[] · required	1-10 reference image URLs.
resolution string default "1K"	"1K", "1.5K", "2K" Output size (1K, 1.5K or 2K). Set it with `resolution`; `video_quality` is not used.
aspect_ratio string default "16:9"	"1:1", "4:3", "3:4", "16:9", "9:16", "2:3", "3:2", "21:9" Output aspect ratio; no auto option. Always send it explicitly.
output_format string default "PNG"	"PNG", "JPEG" Output format in uppercase: PNG or JPEG.
duration string default "0"	Send "0".

Input files

images: 1-10 reference image URLs.

Upload folders for /api/storage/upload

image: image-generator

Rules

generation_type must be text-to-image or image-to-image (422).
aspect_ratio must be one of 1:1, 4:3, 3:4, 16:9, 9:16, 2:3, 3:2, 21:9 (400). Always send aspect_ratio explicitly.
resolution must be 1K, 1.5K or 2K (422).
prompt max 5000 characters (400).
image-to-image requires a non-empty images array (or image) (400).
At most 10 reference images (400).
Asynchronous: POST returns {generation_id, status:'processing'}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "seedream-5.0-flash",
  "generation_type": "image-to-image",
  "prompt": "A cozy coffee shop interior at golden hour, warm light through large windows",
  "duration": "0",
  "aspect_ratio": "1:1",
  "resolution": "1K",
  "output_format": "PNG",
  "images": [
    "https://example.com/ref.jpg"
  ]
}

Ideogram 4.5 mode: ideogram-4-5

text-to-image

Generates 1-8 images with accurate in-image text from a prompt with Ideogram 4.5.

Parameter	Values & notes
prompt string · required	Required, non-empty, max 10000 characters.
quality string affects price default "low"	"low", "medium", "high"
num_images number affects price default 1	1–8 Images per request (1-8); the price is multiplied by num_images.
enable_prompt_expansion boolean default true	Text-to-image only.
aspect_ratio string default "1:1"	"1:1", "16:9", "9:16", "4:3", "3:4", "3:2", "2:3", "5:4", "4:5", "16:10", "10:16", "2:1", "1:2", "12:5", "5:12", "22:9", "9:22", "23:9", "9:23", "8:3", "3:8", "3:1", "1:3" Output aspect ratio.
resolution string	"1K", "2K" Resolution tier for the chosen aspect_ratio; defaults to the lowest available. 12:5, 5:12, 22:9, 9:22, 23:9, 9:23, 8:3, 3:8, 3:1, 1:3 are 2K only.
seed number	Optional integer seed.
duration string default "0"	Send "0".

Rules

Credits for images that are not delivered are refunded automatically.
generation_type must be text-to-image or image-to-image.
prompt required, max 10000 characters (422).
num_images must be an integer 1-8 (422).
resolution must be valid for the chosen aspect_ratio (422).
text-to-image must not include images (422); edit_precision and mask_url are rejected (422).
aspect_ratio auto is not allowed for text-to-image.
Asynchronous: POST returns {generation_id, status:'processing'}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is in video_url and all output images are in result_images (array). The status response has no image_url field. With num_images > 1, all images are in result_images.

Example request body

{
  "mode": "ideogram-4-5",
  "generation_type": "text-to-image",
  "prompt": "A vintage travel poster with the bold title \"VISIT LISBON\" and a yellow tram",
  "duration": "0",
  "quality": "low",
  "num_images": 1,
  "aspect_ratio": "1:1",
  "resolution": "1K",
  "enable_prompt_expansion": true
}

image-to-image

Edits an image (optionally with a mask and reference images) with accurate text rendering via Ideogram 4.5.

Parameter	Values & notes
prompt string · required	Required, non-empty, max 10000 characters.
images string[] · required	images[0] = image to edit; images[1..] = up to 4 reference images (3 with a mask). http(s) URLs only.
quality string affects price default "very_low"	"very_low", "low", "medium", "high"
num_images number affects price default 1	1–8 Images per request (1-8); the price is multiplied by num_images.
edit_precision string default "regular"	"regular", "high" Edit precision.
mask_url string	Black/white mask URL, same size as the source image; black = area to change.
aspect_ratio string default "auto"	"auto", "1:1", "16:9", "9:16", "4:3", "3:4", "3:2", "2:3", "5:4", "4:5", "16:10", "10:16", "2:1", "1:2", "12:5", "5:12", "22:9", "9:22", "23:9", "9:23", "8:3", "3:8", "3:1", "1:3" auto keeps the source image size.
resolution string	"1K", "2K" Resolution tier for the chosen aspect_ratio; defaults to the lowest available. 12:5, 5:12, 22:9, 9:22, 23:9, 9:23, 8:3, 3:8, 3:1, 1:3 are 2K only. Not allowed with aspect_ratio auto.
seed number	Optional integer seed.
duration string default "0"	Send "0".

Input files

images: 1 image to edit + 0-4 reference image URLs (0-3 with a mask).
mask: Optional mask_url (PNG, black = edit area, same size as the source).

Upload folders for /api/storage/upload

image: image-generator
mask: image-generator

Rules

Credits for images that are not delivered are refunded automatically.
generation_type must be text-to-image or image-to-image.
prompt required, max 10000 characters (422).
num_images must be an integer 1-8 (422).
resolution must be valid for the chosen aspect_ratio (422).
very_low quality is available for image-to-image only.
Needs at least 1 image; at most 4 reference images after the first (3 when mask_url is set). Image URLs must be http(s).
enable_prompt_expansion is rejected for image-to-image (422).
A non-auto aspect_ratio/resolution is only allowed with edit_precision=regular and no mask (422).
mask_url is only accepted for qwen-image-2-1 and ideogram-4-5 and must be an http(s) URL.
Asynchronous: POST returns {generation_id, status:'processing'}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is in video_url and all output images are in result_images (array). The status response has no image_url field. With num_images > 1, all images are in result_images.

Example request body

{
  "mode": "ideogram-4-5",
  "generation_type": "image-to-image",
  "prompt": "Replace the sign text with \"OPEN\"",
  "duration": "0",
  "quality": "very_low",
  "num_images": 1,
  "aspect_ratio": "auto",
  "images": [
    "https://example.com/source.jpg"
  ],
  "edit_precision": "regular"
}

GPT Image 2 mode: gpt-image-2

text-to-image

Generates photoreal images with sharp text rendering from a prompt with GPT Image 2.

Parameter	Values & notes
prompt string · required	Required, non-empty, max 20000 characters.
resolution string affects price default "1K"	"1K", "2K", "4K"
aspect_ratio string default "auto"	"auto", "1:1", "3:2", "2:3", "4:3", "3:4", "5:4", "4:5", "16:9", "9:16", "2:1", "1:2", "3:1", "1:3", "21:9", "9:21" Omitted = auto.
background string default "auto"	"auto", "opaque", "transparent" Values other than auto are only allowed at 1K.
duration string default "0"	Send "0".

Rules

generation_type must be text-to-image or image-to-image; prompt required, max 20000 characters (422).
aspect_ratio auto (or omitted) only supports 1K (422).
1:1 does not support 4K (422).
5:4, 4:5, 3:1, 1:3 and 9:21 are not supported at 2K; 3:1, 1:3 and 9:21 are not supported at 4K (422).
background other than auto is only allowed at 1K (422).
Asynchronous: POST returns {generation_id, status:'processing'}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "gpt-image-2",
  "generation_type": "text-to-image",
  "prompt": "A modern storefront poster with the headline \"Summer Sale\" in bold serif type",
  "duration": "0",
  "aspect_ratio": "auto",
  "resolution": "1K",
  "images": []
}

image-to-image

Edits/combines up to 16 reference images from a prompt with GPT Image 2.

Parameter	Values & notes
prompt string · required	Required, non-empty, max 20000 characters.
images string[] · required	1-16 reference image URLs.
resolution string affects price default "1K"	"1K", "2K", "4K"
aspect_ratio string default "auto"	"auto", "1:1", "3:2", "2:3", "4:3", "3:4", "5:4", "4:5", "16:9", "9:16", "2:1", "1:2", "3:1", "1:3", "21:9", "9:21" Omitted = auto.
background string default "auto"	"auto", "opaque", "transparent" Values other than auto are only allowed at 1K.
duration string default "0"	Send "0".

Input files

images: 1-16 reference image URLs.

Upload folders for /api/storage/upload

image: image-generator

Rules

generation_type must be text-to-image or image-to-image; prompt required, max 20000 characters (422).
aspect_ratio auto (or omitted) only supports 1K (422).
1:1 does not support 4K (422).
5:4, 4:5, 3:1, 1:3 and 9:21 are not supported at 2K; 3:1, 1:3 and 9:21 are not supported at 4K (422).
background other than auto is only allowed at 1K (422).
image-to-image requires a non-empty images array (or image) (400).
At most 16 reference images (400).
Asynchronous: POST returns {generation_id, status:'processing'}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "gpt-image-2",
  "generation_type": "image-to-image",
  "prompt": "A modern storefront poster with the headline \"Summer Sale\" in bold serif type",
  "duration": "0",
  "aspect_ratio": "auto",
  "resolution": "1K",
  "images": [
    "https://example.com/ref.jpg"
  ]
}

GPT Image 2.5 mode: gpt-image-2-5

text-to-image

Generates sharper, detailed images from a prompt with GPT Image 2.5 (Flare/Sunburst).

Parameter	Values & notes
prompt string · required	Required, non-empty, max 20000 characters.
gptImage25Version string default "flare"	"flare", "sunburst" Quality tier; flare and sunburst currently cost the same.
resolution string affects price default "1K"	"1K", "2K", "4K"
aspect_ratio string default "auto"	"auto", "1:1", "3:2", "2:3", "4:3", "3:4", "16:9", "9:16", "21:9", "27:16", "16:27", "9:8", "8:9" Omitted = auto. 27:16, 16:27, 9:8, 8:9 are 1K only.
background string default "auto"	"auto", "opaque", "transparent" Available at all resolutions.
duration string default "0"	Send "0".

Rules

generation_type must be text-to-image or image-to-image; prompt required, max 20000 characters (422).
27:16, 16:27, 9:8, 8:9 are only available at 1K (422).
Asynchronous: POST returns {generation_id, status:'processing'}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "gpt-image-2-5",
  "generation_type": "text-to-image",
  "prompt": "A cinematic night city poster with neon reflections on a rainy street",
  "duration": "0",
  "aspect_ratio": "auto",
  "resolution": "1K",
  "gptImage25Version": "flare",
  "background": "auto",
  "images": []
}

image-to-image

Edits up to 16 reference images with stronger reference fidelity using GPT Image 2.5.

Parameter	Values & notes
prompt string · required	Required, non-empty, max 20000 characters.
images string[] · required	1-16 reference image URLs.
gptImage25Version string default "flare"	"flare", "sunburst" Quality tier; flare and sunburst currently cost the same.
resolution string affects price default "1K"	"1K", "2K", "4K"
aspect_ratio string default "auto"	"auto", "1:1", "3:2", "2:3", "4:3", "3:4", "16:9", "9:16", "21:9", "27:16", "16:27", "9:8", "8:9" Omitted = auto. 27:16, 16:27, 9:8, 8:9 are 1K only.
background string default "auto"	"auto", "opaque", "transparent" Available at all resolutions.
duration string default "0"	Send "0".

Input files

images: 1-16 reference image URLs.

Upload folders for /api/storage/upload

image: image-generator

Rules

generation_type must be text-to-image or image-to-image; prompt required, max 20000 characters (422).
27:16, 16:27, 9:8, 8:9 are only available at 1K (422).
image-to-image requires a non-empty images array (or image) (400).
At most 16 reference images (400).
Asynchronous: POST returns {generation_id, status:'processing'}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "gpt-image-2-5",
  "generation_type": "image-to-image",
  "prompt": "A cinematic night city poster with neon reflections on a rainy street",
  "duration": "0",
  "aspect_ratio": "auto",
  "resolution": "1K",
  "gptImage25Version": "flare",
  "background": "auto",
  "images": [
    "https://example.com/ref.jpg"
  ]
}

Qwen Image 2.1 mode: qwen-image-2-1

text-to-image

Generates images (including native transparent PNGs) from a prompt with Qwen Image 2.1.

Parameter	Values & notes
prompt string · required	Required, non-empty, max 5000 characters.
resolution string affects price default "1K"	"1K", "2K"
aspect_ratio string default "1:1"	"1:1", "4:3", "3:4", "3:2", "2:3", "16:9", "9:16" Text-to-image default 1:1; auto is not available.
background string default "opaque"	"opaque", "transparent" transparent returns a native transparent PNG; 'auto' is rejected.
duration string default "0"	Send "0".

Rules

generation_type must be text-to-image or image-to-image; prompt required, max 5000 characters (422).
resolution must be 1K or 2K (422).
background must be opaque or transparent (422).
Text-to-image aspect ratios: 1:1, 4:3, 3:4, 3:2, 2:3, 16:9, 9:16 (auto is rejected, 422).
mask_url is rejected for text-to-image (422).
Asynchronous: POST returns {generation_id, status:'processing'}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "qwen-image-2-1",
  "generation_type": "text-to-image",
  "prompt": "A glossy cartoon sticker of a corgi astronaut, transparent background",
  "duration": "0",
  "aspect_ratio": "1:1",
  "resolution": "1K",
  "background": "transparent",
  "images": []
}

image-to-image

Edits up to 10 reference images, or inpaints one image with a mask, using Qwen Image 2.1.

Parameter	Values & notes
prompt string · required	Required, non-empty, max 5000 characters.
images string[] · required	1-10 reference image URLs (exactly 1 with mask_url).
resolution string affects price default "1K"	"1K", "2K"
aspect_ratio string default "auto"	"auto", "1:1", "4:3", "3:4", "3:2", "2:3", "16:9", "9:16" Image-to-image default auto. When mask_url is set, auto is always used.
background string default "opaque"	"opaque", "transparent" transparent returns a native transparent PNG; 'auto' is rejected.
mask_url string	Black/white mask URL for inpainting; white = area to change.
duration string default "0"	Send "0".

Input files

images: 1-10 reference image URLs (exactly 1 when using a mask).
mask: Optional mask_url (black/white PNG, white = edit area).

Upload folders for /api/storage/upload

image: image-generator
mask: image-generator

Rules

generation_type must be text-to-image or image-to-image; prompt required, max 5000 characters (422).
resolution must be 1K or 2K (422).
background must be opaque or transparent (422).
image-to-image requires a non-empty images array (or image) (400).
At most 10 reference images (422).
mask_url requires exactly 1 image and cannot be combined with background=transparent (422); with a mask, aspect_ratio is always auto.
Image-to-image aspect ratios: auto, 1:1, 4:3, 3:4, 3:2, 2:3, 16:9, 9:16 (422).
Asynchronous: POST returns {generation_id, status:'processing'}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "qwen-image-2-1",
  "generation_type": "image-to-image",
  "prompt": "A glossy cartoon sticker of a corgi astronaut, transparent background",
  "duration": "0",
  "aspect_ratio": "auto",
  "resolution": "1K",
  "background": "opaque",
  "images": [
    "https://example.com/ref.jpg"
  ]
}

Virtual Try-On mode: virtual-try-on

image-to-image

Dresses the person in a photo in 1-3 garment images.

Parameter	Values & notes
images string[] · required	[person photo, 1-3 garment images]; images[0] is always the person.
prompt string	Optional extra instruction appended to the built-in try-on instruction, max 1000 characters.
aspect_ratio string default "auto"	"auto", "1:1", "2:3", "3:2", "3:4", "4:3", "4:5", "5:4", "9:16", "16:9" Omit or send auto to follow the person photo's ratio.
duration string default "0"	Send "0".

Input files

images: images[0] = person photo URL, images[1..3] = 1-3 garment image URLs.

Upload folders for /api/storage/upload

image: virtual-try-on

Rules

generation_type must be image-to-image (422).
images must be [person, 1-3 garment images] (422).
prompt max 1000 characters (422).
aspect_ratio must be auto or one of 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9 (422).
Asynchronous: POST returns {generation_id, status:'processing'}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "virtual-try-on",
  "generation_type": "image-to-image",
  "prompt": "",
  "duration": "0",
  "images": [
    "https://example.com/person.jpg",
    "https://example.com/garment.jpg"
  ]
}

Product Holding mode: product-holding

image-to-image

Places 1-3 product images into the hands of the person in a photo.

Parameter	Values & notes
images string[] · required	[person photo, 1-3 product images]; images[0] is always the person.
prompt string	Optional extra instruction appended to the built-in product-holding instruction, max 1000 characters.
aspect_ratio string default "auto"	"auto", "1:1", "2:3", "3:2", "3:4", "4:3", "4:5", "5:4", "9:16", "16:9" Omit or send auto to follow the person photo's ratio.
duration string default "0"	Send "0".

Input files

images: images[0] = person photo URL, images[1..3] = 1-3 product image URLs.

Upload folders for /api/storage/upload

image: virtual-try-on

Rules

generation_type must be image-to-image (422).
images must be [person, 1-3 product images] (422).
prompt max 1000 characters (422).
aspect_ratio must be auto or one of 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9 (422).
Asynchronous: POST returns {generation_id, status:'processing'}. Poll GET /api/video-generation/{generation_id}/status until status='completed'; the first image URL is in video_url and all output images are in result_images (array). The status response has no image_url field.

Example request body

{
  "mode": "product-holding",
  "generation_type": "image-to-image",
  "prompt": "",
  "duration": "0",
  "images": [
    "https://example.com/person.jpg",
    "https://example.com/product.jpg"
  ]
}

Audio Models

Text-to-speech, music, speech-to-text and audio cleanup. Audio results are returned in video_url with media_kind "audio".

Model	mode	generation_type
Grok TTS	`grok-tts`	text-to-speech
Eleven v4	`eleven-v4`	text-to-speech
Eleven v4 Turbo	`eleven-v4-turbo`	text-to-speech
Lyria 3.5 (Text to Music)	`lyria-music`	text-to-music
Lyria 3.5 (Image to Music)	`lyria-music`	text-to-music
Scribe V2	`scribe-v2`	speech-to-text
VEED Clean Audio	`veed-clean-audio`	remove-background-noise

Grok TTS mode: grok-tts

text-to-speech

Converts text to speech with xAI Grok TTS, with selectable voice, language and output codec.

Parameter	Values & notes
prompt string · required affects price	1–15000 Text to speak (sent as-is, never auto-translated). Length is counted in characters.
duration string default "0"	"0" Send "0" (audio has no duration input).
voice string default "eve"	"carina", "zagan", "helix", "orion", "luna", "iris", "altair", "zenith", "perseus", "helios", "lux", "kepler", "rigel", "cosmo", "celeste", "ursa", "sirius", "lumen", "castor", "naksh", "atlas", "aurora", "liora", "ara", "eve", "leo", "rex", "sal" Voice name.
language string default "auto"	"auto", "en", "zh", "ja", "ko", "fr", "de", "it", "es-ES", "es-MX", "pt-BR", "pt-PT", "ru", "tr", "vi", "id", "hi", "bn", "ar-EG", "ar-SA", "ar-AE" BCP-47 language; 'auto' = auto-detect.
audio_codec string default "mp3"	"mp3", "wav", "pcm", "mulaw", "alaw" Output format. pcm/mulaw/alaw are headerless raw data (download only, not playable in a browser).
sample_rate number default 24000	8000, 16000, 22050, 24000, 44100, 48000 Output sample rate in Hz.
bit_rate number default 128000	32000, 64000, 96000, 128000, 192000 Only applies when audio_codec is mp3. The maximum depends on sample_rate: up to 8000 Hz → 64000, up to 24000 Hz → 160000 (so 128000 is the highest allowed value), above → 192000.
seed number	Accepted for backward compatibility; can be omitted.

Rules

prompt is required and must be non-empty after trim; max 15000 characters.
generation_type must be 'text-to-speech'; text-to-speech only accepts modes eleven-v4-turbo, eleven-v4, grok-tts.
voice, language, audio_codec, sample_rate and bit_rate must be one of the listed values when provided, else 400.
bit_rate above the max for the sample_rate (sample_rate defaults to 24000 when omitted) is rejected with 400.
Eleven-only fields (stability, similarity_boost, audio_output_format, text_normalization, timestamps=true) are rejected with 400.

Result

Async: POST returns {generation_id, status:'processing'}. Poll GET /api/video-generation/{id}/status; when status='completed', video_url is the generated audio file, media_kind is 'audio', and video_metadata gives content_type/file_name/file_size. No lyrics/transcript/timestamps.

Example request body

{
  "mode": "grok-tts",
  "generation_type": "text-to-speech",
  "prompt": "Welcome to Veevid. Let's make something amazing today.",
  "duration": "0",
  "voice": "eve",
  "language": "auto",
  "audio_codec": "mp3",
  "sample_rate": 24000,
  "bit_rate": 128000
}

Eleven v4 mode: eleven-v4

text-to-speech

Converts text to speech with ElevenLabs Eleven v4, the higher-quality tier.

Parameter	Values & notes
prompt string · required affects price	1–5000 Text to speak (not auto-translated). Max 5000 characters.
duration string default "0"	"0" Send "0".
voice string default "Rachel"	"Rachel", "Aria", "Roger", "Sarah", "Laura", "Charlie", "George", "Callum", "River", "Liam", "Charlotte", "Alice", "Matilda", "Will", "Jessica", "Eric", "Chris", "Brian", "Daniel", "Lily", "Bill" Preset voice name (case-sensitive).
language string default "auto"	"auto", "en", "zh", "yue", "es", "pt", "fr", "de", "it", "ja", "ko", "hi", "ar", "ru", "af", "as", "ast", "az", "be", "bg", "bn", "bs", "ca", "ceb", "cs", "cy", "da", "el", "et", "fa", "fi", "fil", "gl", "gu", "ha", "he", "hr", "hu", "hy", "id", "is", "jv", "ka", "kk", "kn", "ky", "lb", "ln", "lt", "lv", "mi", "mk", "ml", "mn", "mr", "ms", "mt", "my", "ne", "nl", "no", "oc", "or", "pa", "pl", "ps", "ro", "sd", "sk", "sl", "so", "sr", "sv", "sw", "ta", "te", "tg", "th", "tr", "uk", "ur", "uz", "vi" ISO 639-1 code (a few 3-letter codes: yue, ast, ceb, fil); 'auto' = let the model detect.
stability number default 0.5	0–1 Lower = more expressive, higher = more stable.
similarity_boost number default 0.75	0–1 Higher = closer to the original voice.
seed number	0–4294967295 Integer seed; omit for a random seed.
audio_output_format string default "mp3_44100_128"	"mp3_22050_32", "mp3_44100_32", "mp3_44100_64", "mp3_44100_96", "mp3_44100_128", "mp3_44100_192", "opus_48000_32", "opus_48000_64", "opus_48000_96", "opus_48000_128", "opus_48000_192", "pcm_8000", "pcm_16000", "pcm_22050", "pcm_24000", "pcm_44100", "pcm_48000", "ulaw_8000", "alaw_8000" codec_samplerate_bitrate. opus ->.ogg; pcm/ulaw/alaw are headerless raw files (.pcm/.ulaw/.alaw).
text_normalization string default "auto"	"auto", "on", "off" Whether numbers/dates/abbreviations are expanded before synthesis.
timestamps boolean default false	Return per-character timestamps in the status response.

Rules

prompt is required and non-empty after trim; max 5000 characters.
generation_type must be 'text-to-speech'.
Grok-only fields audio_codec / sample_rate / bit_rate are rejected with 400 (use audio_output_format).
voice, language, audio_output_format and text_normalization must be one of the listed values when provided, else 400.
stability and similarity_boost must be finite numbers in [0,1]; seed must be an integer in [0, 4294967295].

Result

Async: poll GET /api/video-generation/{id}/status. On completed, video_url is the generated audio file and media_kind is 'audio'. If timestamps:true was sent, the response also has 'timestamps': an array of word groups like {characters:[...], character_start_times_seconds:[...], character_end_times_seconds:[...]} (only present when non-empty).

Example request body

{
  "mode": "eleven-v4",
  "generation_type": "text-to-speech",
  "prompt": "Welcome to Veevid. Let's make something amazing today.",
  "duration": "0",
  "voice": "Rachel",
  "language": "auto",
  "stability": 0.5,
  "similarity_boost": 0.75,
  "audio_output_format": "mp3_44100_128",
  "text_normalization": "auto",
  "timestamps": false
}

Eleven v4 Turbo mode: eleven-v4-turbo

text-to-speech

Converts text to speech with ElevenLabs Eleven v4 Turbo: half the price of Eleven v4, with lower latency.

Parameter	Values & notes
prompt string · required affects price	1–5000 Text to speak (not auto-translated). Max 5000 characters.
duration string default "0"	"0" Send "0".
voice string default "Rachel"	"Rachel", "Aria", "Roger", "Sarah", "Laura", "Charlie", "George", "Callum", "River", "Liam", "Charlotte", "Alice", "Matilda", "Will", "Jessica", "Eric", "Chris", "Brian", "Daniel", "Lily", "Bill" Preset voice name (case-sensitive).
language string default "auto"	"auto", "en", "zh", "yue", "es", "pt", "fr", "de", "it", "ja", "ko", "hi", "ar", "ru", "af", "as", "ast", "az", "be", "bg", "bn", "bs", "ca", "ceb", "cs", "cy", "da", "el", "et", "fa", "fi", "fil", "gl", "gu", "ha", "he", "hr", "hu", "hy", "id", "is", "jv", "ka", "kk", "kn", "ky", "lb", "ln", "lt", "lv", "mi", "mk", "ml", "mn", "mr", "ms", "mt", "my", "ne", "nl", "no", "oc", "or", "pa", "pl", "ps", "ro", "sd", "sk", "sl", "so", "sr", "sv", "sw", "ta", "te", "tg", "th", "tr", "uk", "ur", "uz", "vi" Same language list as eleven-v4.
stability number default 0.5	0–1
similarity_boost number default 0.75	0–1
seed number	0–4294967295 Integer seed; omit for a random seed.
audio_output_format string default "mp3_44100_128"	"mp3_22050_32", "mp3_44100_32", "mp3_44100_64", "mp3_44100_96", "mp3_44100_128", "mp3_44100_192", "opus_48000_32", "opus_48000_64", "opus_48000_96", "opus_48000_128", "opus_48000_192", "pcm_8000", "pcm_16000", "pcm_22050", "pcm_24000", "pcm_44100", "pcm_48000", "ulaw_8000", "alaw_8000"
text_normalization string default "auto"	"auto", "on", "off"
timestamps boolean default false	Return per-character timestamps in the status response.

Rules

Same validation rules as eleven-v4; only the price differs.

Result

Same as eleven-v4: video_url is the generated audio file; optional 'timestamps' array when timestamps:true.

Example request body

{
  "mode": "eleven-v4-turbo",
  "generation_type": "text-to-speech",
  "prompt": "Welcome to Veevid. Let's make something amazing today.",
  "duration": "0",
  "voice": "Rachel",
  "language": "auto",
  "stability": 0.5,
  "similarity_boost": 0.75,
  "audio_output_format": "mp3_44100_128",
  "text_normalization": "auto",
  "timestamps": false
}

Lyria 3.5 (Text to Music) mode: lyria-music

text-to-music

Generates a music track (about 1-2 minutes, vocals or instrumental) from a text prompt with Google Lyria 3.5.

Parameter	Values & notes
prompt string · required	1–5000 Music description (style, mood, instruments, tempo, structure). There is no duration parameter; length can only be described in the prompt (observed minimum ~60s).
duration string default "0"	"0" Send "0".
image string	Optional single reference image https URL.

Input files

image: Optional, 1 image https URL (jpeg/png/webp, max 10MB; upload it with POST /api/storage/upload). Only one image is used.

Upload folders for /api/storage/upload

image: lyria-music

Rules

prompt is required and non-empty after trim; max 5000 characters.
generation_type must be 'text-to-music' (and text-to-music only accepts lyria-music).
image and every entry of images, if provided, must start with https://.
Prompts imitating real singers or copyrighted lyrics are rejected by the model; unused credits are refunded automatically.

Result

Async: poll GET /api/video-generation/{id}/status. On completed, video_url is the generated music audio file and media_kind is 'audio'. 'lyrics' (string) is present only when the track has vocals; it is raw Lyria markup (section tags like [[A0]], line timing like [12.0:]), so clean it before displaying.

Example request body

{
  "mode": "lyria-music",
  "generation_type": "text-to-music",
  "prompt": "Upbeat lo-fi hip hop with warm piano chords, soft vinyl crackle and a relaxed 80 BPM groove. About two minutes long.",
  "duration": "0"
}

Lyria 3.5 (Image to Music) mode: lyria-music

text-to-music

Generates music that matches the mood of a reference image plus a text prompt with Google Lyria 3.5 (image-to-music).

Parameter	Values & notes
prompt string · required	1–5000 Still required. Example: 'Instrumental background music that matches the mood of this image. About one minute long.'
image string	Reference image https URL; the main input for image-to-music (optional at the API level).
duration string default "0"	"0" Send "0".

Input files

image: 1 image https URL (jpeg/png/webp, max 10MB; upload it with POST /api/storage/upload). Only one image is used.

Upload folders for /api/storage/upload

image: lyria-music

Rules

Same mode, generation_type, validation and price as the text-to-music entry; there is no separate 'image-to-music' generation_type.
image must be https; a prompt is still required.

Result

Same as text-to-music: video_url is the music audio file; optional 'lyrics'.

Example request body

{
  "mode": "lyria-music",
  "generation_type": "text-to-music",
  "prompt": "Instrumental background music that matches the mood of this image. About one minute long.",
  "duration": "0",
  "image": "https://cdn.veevid.ai/image-to-music/i2m-sample-coffee.webp"
}

Scribe V2 mode: scribe-v2

speech-to-text

Transcribes an audio file to text with speaker diarization and audio-event tags using ElevenLabs Scribe V2.

Parameter	Values & notes
audio string · required	https URL of the audio file to transcribe.
input_audio_duration number · required affects price	0–3600 Audio length in seconds (must be > 0, max 3600); send the actual length. Billing basis: ceil(seconds/60) minutes, min 1.
duration string default "0"	"0" Send "0".
diarize boolean default true	Identify speakers.
tag_audio_events boolean default true	Tag non-speech events like [laughter].
keyterms array<string> affects price	Proper nouns/brand names to bias recognition; max 100 items, each non-empty and <=50 chars. Send it only when non-empty; keyterms raise the price.

Input files

audio: 1 audio https URL;.mp3,.wav,.m4a,.ogg,.flac,.aac only (video files such as mp4 are not supported). Max 60 minutes; upload with POST /api/storage/upload (max 100MB).

Upload folders for /api/storage/upload

audio: speech-to-text/audio

Rules

generation_type must be 'speech-to-text' and vice versa.
audio is required and must be an https URL; audios (array) is rejected - single file only.
If the URL has a file extension it must be one of.mp3.wav.m4a.ogg.flac.aac; URLs without an extension are accepted.
input_audio_duration is required, > 0 and <= 3600 seconds.
keyterms: max 100, each non-empty and <= 50 characters after trim.
Send input_audio_duration equal to the actual audio length (seconds); if it is under-declared, the difference is charged after transcription.
Failed transcriptions are refunded automatically.
To quote, send the https audio URL and input_audio_duration (the quote checks the URL but does not download it).

Result

Async and fast (poll GET /api/video-generation/{id}/status, e.g. every 4s). On completed: 'transcript' object {v:1, language (ISO 639-3 e.g. 'eng', or null), duration (seconds, end of last word), speakers (count), text (full text), cues: [{s: start_sec, e: end_sec, p: speaker_index_or_null, t: text}]}. Note: video_url here is the original source audio URL (not a result file) and is kept for 14 days. media_kind 'audio'.

Example request body

{
  "mode": "scribe-v2",
  "generation_type": "speech-to-text",
  "duration": "0",
  "audio": "https://example.com/interview.mp3",
  "input_audio_duration": 95.4,
  "diarize": true,
  "tag_audio_events": true
}

VEED Clean Audio mode: veed-clean-audio

remove-background-noise

Removes background noise from an audio or video file and returns a cleaned, loudness-normalized audio file with VEED Clean Audio.

Parameter	Values & notes
audio string · required	https URL of the audio OR video file to clean (video: only the audio track is used).
input_audio_duration number · required affects price	0–1800 Media length in seconds (must be > 0, max 1800); send the actual length.
duration string default "0"	"0" Send "0".
strength number default 0.874	0–1 How much of the original background is kept under speech; suppression floor = 1 - strength (0.874 = -18 dB). Silence between words is always cleaned.
target_lufs number\|null default -19	-40–-8 Output integrated loudness (LUFS). null = disable loudness normalization (keep original volume); omitted = -19. Gain only, peak capped at -1.1 dBTP so high targets may not be reached.
audio_output_format string default "flac"	"flac", "wav" Output is always 48kHz mono 16-bit

Input files

audio: 1 https URL; audio.mp3.wav.m4a.aac.ogg.opus.flac or video.mp4.mov.m4v.webm. Max 30 minutes and 512MB (upload it with POST /api/storage/presign-upload, see File Uploads). Video without an audio track fails and credits are refunded automatically.

Rules

generation_type must be 'remove-background-noise' and vice versa.
audio is required and must be an https URL; audios (array) is rejected - single file only.
If the URL has a file extension it must be one of.mp3.wav.m4a.aac.ogg.opus.flac.mp4.mov.m4v.webm; URLs without an extension are accepted.
input_audio_duration is required, > 0 and <= 1800 seconds (30 minutes).
strength must be in [0,1]; target_lufs must be null or in [-40,-8]; audio_output_format must be flac or wav.
Send input_audio_duration equal to the actual media length (seconds); if it is under-declared, the difference is charged after processing.
Failed generations (e.g. 'File has no audio track') are refunded automatically.
To quote, send the https audio URL and input_audio_duration (the quote checks the URL but does not download it).

Result

Async (processing takes about 0.2-0.35x the media length; poll GET /api/video-generation/{id}/status). On completed, video_url is the cleaned audio file (.flac or.wav) and media_kind is 'audio'. No lyrics/transcript/timestamps.

Example request body

{
  "mode": "veed-clean-audio",
  "generation_type": "remove-background-noise",
  "duration": "0",
  "audio": "https://example.com/noisy-recording.mp3",
  "input_audio_duration": 162.3,
  "strength": 0.874,
  "target_lufs": -19,
  "audio_output_format": "flac"
}

Video Tools & Effects

Upscale, resize, edit, extend, insert shots, swap characters, lip-sync, motion control and video effects.

Model	mode	generation_type
Kling 3.0 Motion Control	`motion-control-standard`	motion-control
Kling 3.0 Motion Control	`motion-control-pro`	motion-control
AI Hug Video Generator	`ai-hug`	ai-hug
AI Kissing Video Generator	`ai-kissing`	ai-kissing
AI Video Upscaler	`upscale`	video-upscale
AI Video Resizer	`reframe`	video-reframe
Wan 2.7 Edit	`wan-2.7-edit`	edit-video
Wan 3.0	`wan-3.0`	edit-video, video-extend
Seedance 2.0	`seedance-2.0`	edit-video, video-extend
Seedance 2.5	`seedance-2.5`	edit-video, video-extend
LTX 2.3	`ltx-2-3`	video-extend
MiniMax H3 Max	`minimax-h3-max-extend`	video-extend
MiniMax H3 Max	`minimax-h3-max-insert`	video-insert
MiniMax H3 Max	`minimax-h3-max-recast`	character-swap
Video Lip Sync	`video-lip-sync`	lip-sync
Google Veo 3.1	`veo3`	video-extend
Grok Imagine	`grok-imagine`	video-extend

Kling 3.0 Motion Control mode: motion-control-standard

motion-control

Transfers the motion from a driving video onto the character in a reference image (Kling 3.0 Motion Control, Standard).

Parameter	Values & notes
image string · required	Public URL of the character image.
video string · required	Public URL of the driving (motion reference) video.
duration string (whole seconds) · required affects price	Driving video length as a whole number of seconds, rounded up (e.g. "5"; decimals are rejected); it determines the price. When the driving video is on Veevid storage the server also measures it and charges by the measured length if that is longer.
character_orientation string · required	"image", "video" 'image' keeps the character orientation from the image (video up to 10s); 'video' follows the video (up to 30s).
model_version string affects price default "motion-control-standard"	"motion-control-standard", "motion-control-pro" Optional; the Pro tier is used when either mode or model_version is motion-control-pro.
prompt string	Optional text guidance, e.g. 'No distortion, the character's movements are consistent with the video.' Omit if blank.
keep_original_sound boolean default true	Keep the driving video's audio.
elements array	At most 1 element: {frontal_image_url?: string, reference_image_urls: string[1-3]}. Only allowed when character_orientation is 'video'. Send [] when unused.
aspect_ratio string	"original" Send 'original'; output follows the inputs.

Input files

image: 1 character image URL (JPEG/PNG/WebP, max 10MB)
video: 1 driving video URL; up to 10s when character_orientation=image, up to 30s when =video (mp4/webm/mov, max 200MB)
images: optional element images via elements[0].frontal_image_url and elements[0].reference_image_urls (1-3)

Upload folders for /api/storage/upload

image: kling-motion-control/reference-images
video: kling-motion-control/reference-videos
images: kling-motion-control/reference-images

Rules

image, video and character_orientation are required (400 otherwise).
elements are only allowed when character_orientation=video; at most 1 element with 1-3 reference_image_urls.
Send duration equal to the driving video length in whole seconds (rounded up).
duration must be a whole number of seconds (decimals are rejected); max 10 when character_orientation=image, max 30 when =video (400 otherwise).
mode motion-control-standard / motion-control-pro must be used with generation_type motion-control, and vice versa (400 otherwise).
If the output video is longer than the charged duration, the difference is charged after generation; if your balance cannot cover it, the result is not delivered and the generation fails without a refund.

Example request body

{
  "mode": "motion-control-standard",
  "generation_type": "motion-control",
  "model_version": "motion-control-standard",
  "prompt": "No distortion, the character's movements are consistent with the video.",
  "image": "https://cdn.veevid.ai/sample/motion-control-image-1.webp",
  "video": "https://cdn.veevid.ai/sample/motion-control-video-1.mp4",
  "duration": "5",
  "aspect_ratio": "original",
  "character_orientation": "image",
  "keep_original_sound": true,
  "elements": []
}

Kling 3.0 Motion Control mode: motion-control-pro

motion-control

Transfers the motion from a driving video onto the character in a reference image at Pro quality (Kling 3.0 Motion Control, Pro).

Parameter	Values & notes
image string · required	Public URL of the character image.
video string · required	Public URL of the driving (motion reference) video.
duration string (whole seconds) · required affects price	Driving video length as a whole number of seconds, rounded up (e.g. "5"; decimals are rejected); it determines the price. When the driving video is on Veevid storage the server also measures it and charges by the measured length if that is longer.
character_orientation string · required	"image", "video" 'image': video up to 10s; 'video': video up to 30s.
model_version string affects price default "motion-control-pro"	"motion-control-standard", "motion-control-pro" Optional; the Pro tier is used when either mode or model_version is motion-control-pro.
prompt string	Optional text guidance (omit if blank).
keep_original_sound boolean default true	Keep the driving video's audio.
elements array	At most 1 element: {frontal_image_url?: string, reference_image_urls: string[1-3]}; only with character_orientation='video'. Send [] when unused.
aspect_ratio string	"original" Send 'original'; output follows the inputs.

Input files

image: 1 character image URL (JPEG/PNG/WebP, max 10MB)
video: 1 driving video URL; up to 10s when character_orientation=image, up to 30s when =video (max 200MB)
images: optional element images via elements[0].frontal_image_url and elements[0].reference_image_urls (1-3)

Upload folders for /api/storage/upload

image: kling-motion-control/reference-images
video: kling-motion-control/reference-videos
images: kling-motion-control/reference-images

Rules

image, video and character_orientation are required (400 otherwise).
elements are only allowed when character_orientation=video; at most 1 element with 1-3 reference_image_urls.
Send duration equal to the driving video length in whole seconds (rounded up).
duration must be a whole number of seconds (decimals are rejected); max 10 when character_orientation=image, max 30 when =video (400 otherwise).
mode motion-control-standard / motion-control-pro must be used with generation_type motion-control, and vice versa (400 otherwise).
If the output video is longer than the charged duration, the difference is charged after generation; if your balance cannot cover it, the result is not delivered and the generation fails without a refund.

Example request body

{
  "mode": "motion-control-pro",
  "generation_type": "motion-control",
  "model_version": "motion-control-pro",
  "prompt": "No distortion, the character's movements are consistent with the video.",
  "image": "https://cdn.veevid.ai/sample/motion-control-image-1.webp",
  "video": "https://cdn.veevid.ai/sample/motion-control-video-1.mp4",
  "duration": "5",
  "aspect_ratio": "original",
  "character_orientation": "video",
  "keep_original_sound": true,
  "elements": []
}

AI Hug Video Generator mode: ai-hug

ai-hug

Generates a video of two people (or a person and a pet) from two separate photos hugging in a described scene.

Parameter	Values & notes
prompt string · required	Only the scene sentence (max 500 chars). It is automatically wrapped in a fixed template that references the two photos as @Element1/@Element2; do not write the full prompt.
images string[] · required	Exactly 2 image URLs: [person A, person B]; images[0] = @Element1.
resolution string affects price default "720p"	"720p", "1080p", "4k" Uses `resolution`, not video_quality.
duration string affects price default "5"	3–15 Integer seconds.
aspect_ratio string default "16:9"	"16:9", "9:16", "1:1"
generate_audio boolean affects price default false	Generate ambient audio.

Input files

images: exactly 2 image URLs, one subject each (up to 10MB per photo)

Upload folders for /api/storage/upload

image: ai-hug

Rules

mode and generation_type must both be 'ai-hug'
images must contain exactly 2 URLs
prompt (scene) is required and at most 500 characters
resolution must be 720p, 1080p or 4k; duration integer 3-15; aspect_ratio 16:9, 9:16 or 1:1
scene_template is rejected for this mode (only ai-kissing accepts it)

Example request body

{
  "mode": "ai-hug",
  "generation_type": "ai-hug",
  "prompt": "They spot each other in a sunlit park, walk closer and share a warm, tight hug.",
  "images": [
    "https://example.com/person-a.jpg",
    "https://example.com/person-b.jpg"
  ],
  "resolution": "720p",
  "duration": "5",
  "aspect_ratio": "16:9",
  "generate_audio": false
}

AI Kissing Video Generator mode: ai-kissing

ai-kissing

Generates a video of two adults kissing, either from two separate photos or from one photo of both people.

Parameter	Values & notes
prompt string · required	Only the scene sentence (max 500 chars); may refer to the two people only as @Element1 and @Element2 (exact spelling). It is automatically wrapped in a fixed template.
images string[] · required	1 URL (one photo of both people, used as first frame) or 2 URLs ([person 1, person 2], images[0] = @Element1). Count selects the input mode.
scene_template string default "default"	"default", "stylized" 'stylized' for anime/3D style scenes (drops realism constraints).
resolution string affects price default "720p"	"720p", "1080p", "4k" Uses `resolution`, not video_quality.
duration string affects price	3–15 Integer seconds. Default 5 with 2 photos, 8 with 1 photo.
aspect_ratio string default "16:9"	"16:9", "9:16", "1:1" No effect in one-photo mode (output follows the photo).
generate_audio boolean affects price default false

Input files

images: 1 image URL (couple photo) or 2 image URLs (one per person); up to 10MB per photo

Upload folders for /api/storage/upload

image: ai-kissing

Rules

mode and generation_type must both be 'ai-kissing' (legacy mode 'normal' is rejected)
images must contain 1 or 2 URLs
prompt (scene) is required, max 500 characters; only @Element1/@Element2 references allowed
scene_template must be default or stylized
resolution 720p/1080p/4k; duration integer 3-15; aspect_ratio 16:9, 9:16 or 1:1
When quoting with /api/quote, send images or input_image_count (1 or 2) — the input mode decides the default duration (422 otherwise).

Example request body

{
  "mode": "ai-kissing",
  "generation_type": "ai-kissing",
  "prompt": "Standing face to face in a sunlit park at golden hour, @Element1 tilts their head up and closes their eyes while @Element2 gently cups @Element1's face with one hand and leans in; they share a soft, tender kiss on the lips.",
  "images": [
    "https://example.com/person-1.jpg",
    "https://example.com/person-2.jpg"
  ],
  "scene_template": "default",
  "resolution": "720p",
  "duration": "5",
  "aspect_ratio": "16:9",
  "generate_audio": false
}

AI Video Upscaler mode: upscale

video-upscale

Upscales an existing video to 720p, 1080p or 4K and optionally raises its frame rate.

Parameter	Values & notes
video string · required	URL of the source video (singular field).
video_quality string · required affects price	"720p", "1080p", "4K" Target resolution. Always send `video_quality`.
target_fps number affects price default 30	15–60 Target frame rate. Above 30 fps costs more at 1080p and 4K.
duration string affects price default "5"	Source video length in whole seconds, as a string. Send the actual length of the source video; it determines the price (5 when omitted).
video_source string	"upload", "url" 'upload' for files uploaded with `POST /api/storage/upload`, 'url' for an external link. With 'url' the link must answer a HEAD request with a supported video type and be <=200MB.
prompt string default "Upscale"	Optional; has no effect and can be omitted.

Input files

video: 1 video URL (mp4, mov, webm, m4v, mkv, avi, wmv, ts, vob, mpg, flv, 3gp or gif), up to 200MB.

Upload folders for /api/storage/upload

video: upscale-input

Rules

video is required (400 'Video is required for video-upscale generation').
When sent, duration must be a positive integer (400).
target_fps must be 15-60.
With video_source 'url': the URL must answer a HEAD request within 5s, have a supported video content-type or file extension, and be <=200MB.
Processing is asynchronous: the response has status 'processing' and a generation_id to poll.

Example request body

{
  "mode": "upscale",
  "generation_type": "video-upscale",
  "prompt": "Upscale",
  "video": "https://cdn.veevid.ai/upscale-input/clip.mp4",
  "video_quality": "1080p",
  "target_fps": 30,
  "duration": "10",
  "video_source": "upload"
}

AI Video Resizer mode: reframe

video-reframe

Changes a video's aspect ratio by using AI to extend the frame (outpainting) instead of cropping.

Parameter	Values & notes
video string · required	URL of the source video (singular field).
aspect_ratio string default "16:9"	"1:1", "16:9", "9:16", "4:3", "3:4", "21:9", "9:21" Target aspect ratio; use one of the listed values (16:9 when omitted).
duration string affects price default "5"	Source video length in whole seconds, as a string. Send the actual length of the source video.
prompt string default "Reframe Video"	Optional description of the scene to guide the extended areas. The literal 'Reframe Video' means no prompt.
video_source string	"upload", "url" Optional: 'upload' or 'url'.
video_quality string default "standard"	Optional; has no effect (send 'standard' or omit).

Input files

video: 1 video URL, up to 100MB and up to 30s.

Upload folders for /api/storage/upload

video: reframe-input

Rules

video is required (400 'Video is required for video-reframe generation').
Always send duration equal to the source video length (whole seconds).
aspect_ratio must be one of the 7 listed values.
The source video must be at most 30s and 100MB.
Processing is asynchronous: poll the returned generation_id.

Example request body

{
  "mode": "reframe",
  "generation_type": "video-reframe",
  "video": "https://cdn.veevid.ai/reframe-input/clip.mp4",
  "aspect_ratio": "9:16",
  "prompt": "Reframe Video",
  "duration": "8",
  "video_source": "upload",
  "video_quality": "standard"
}

Wan 2.7 Edit mode: wan-2.7-edit

edit-video

Edits an existing 2-10s video according to a text instruction, with an optional reference image.

Parameter	Values & notes
prompt string · required	Editing instruction (required).
video string · required	Source video URL (singular field). MP4 or MOV, 2-10s, <=100MB.
duration string · required affects price	2–10 Output length in whole seconds (2-10). To match the input, send the source length rounded up.
video_quality string default "1080p"	"720p", "1080p" Output resolution. Same price for both.
aspect_ratio string	"auto", "16:9", "9:16", "1:1", "4:3", "3:4" Omit (or send 'auto') to keep the source ratio.
audio_setting string default "auto"	"auto", "origin" 'auto' lets the model generate audio; 'origin' keeps the source audio.
video_source string	"upload", "url" 'upload' for files uploaded with `POST /api/storage/upload`; 'url' for an external link, which must pass a HEAD check (supported video type, <=100MB).
image string	Optional reference image URL (http/https) for the edit.

Input files

video: 1 source video URL (MP4/MOV, 2-10s, <=100MB)
image: optional 1 reference image URL (jpg/png/webp)

Upload folders for /api/storage/upload

video: edit-video-input
image: edit-video-ref

Rules

Always send duration.
duration must be an integer string 2-10 (400).
video_quality, if sent, must be 720p or 1080p; aspect_ratio, if sent, must be auto, 16:9, 9:16, 1:1, 4:3 or 3:4 (400).
video (singular) is required (400 'Video is required for edit-video generation').
With video_source 'url' the link must pass a HEAD check (supported video type, <=100MB).
Source videos uploaded to edit-video-input must be.mp4/.mov, up to 100MB.
prompt is required.

Example request body

{
  "mode": "wan-2.7-edit",
  "generation_type": "edit-video",
  "prompt": "Change the jacket to red leather",
  "video": "https://cdn.veevid.ai/edit-video-input/clip.mp4",
  "duration": "6",
  "video_quality": "720p",
  "audio_setting": "auto",
  "video_source": "upload"
}

Wan 3.0 mode: wan-3.0

edit-video

Edits a source video with a text instruction plus optional reference images and audio, and lets you pick the output length.

Parameter	Values & notes
prompt string · required	Editing instruction (required). It is automatically prefixed with 'Edit Video 1: ' when it lacks an edit intent.
videos string[] · required affects price	[source video URL]. Source seconds are billed together with output seconds.
images string[]	Optional reference images, up to 10.
audios string[]	Optional reference audio files, up to 5.
duration string affects price default "5"	2–30 Output length in whole seconds (2-30); -1 (smart duration) is rejected.
video_quality string affects price default "720p"	"480p", "720p", "1080p" Output resolution.
aspect_ratio string default "adaptive"	"adaptive", "16:9", "9:16", "1:1", "4:3", "3:4" Use one of the listed values; 'adaptive' keeps the source ratio.
generate_audio boolean default true	Generate a soundtrack.
input_video_duration number affects price	0–15 Source video length in seconds, rounded up. Always send it with the source video.
has_video_input boolean	Quote-only field: send `true` to `/api/quote` together with the source video length; `/api/generate-video` ignores it.

Input files

video: 1 source video URL in videos[] (1-15s, up to 100MB)
images: 0-10 reference image URLs (up to 20MB each)
audio: 0-5 reference audio URLs in audios[]

Upload folders for /api/storage/upload

video: wan30/reference-videos
image: wan30
audio: wan30/audio

Rules

Send input_video_duration equal to the source video length (seconds); values > 15 are rejected (400).
duration must be an integer 2-30; -1 is rejected (400).
input_video_duration + duration must be <= 30 (400).
At most 10 images, 5 videos and 5 audios (400).
videos[] must contain the source video (400 'Video is required for edit-video generation').
prompt is required.
When quoting with /api/quote, add has_video_input: true and input_video_duration if you send the source video (the quote endpoint does not read video URLs, so the price would come out too low).

Example request body

{
  "mode": "wan-3.0",
  "generation_type": "edit-video",
  "prompt": "Replace the skateboarder with a short-haired woman in a yellow raincoat",
  "videos": [
    "https://cdn.veevid.ai/wan30/reference-videos/clip.mp4"
  ],
  "duration": "5",
  "video_quality": "720p",
  "aspect_ratio": "adaptive",
  "generate_audio": true,
  "input_video_duration": 6
}

video-extend

Continues an uploaded video (up to 15s), keeping its characters, style and sound.

Parameter	Values & notes
prompt string	What happens next. It is automatically prefixed with 'Extend Video 1 by N seconds.' when no extend intent is present.
videos string[] · required affects price	[source video URL]. No taskId needed.
duration string affects price default "5"	2–30 Length of the new footage in seconds. It must be <= 30 - source seconds.
video_quality string affects price default "720p"	"480p", "720p", "1080p" Output resolution.
aspect_ratio string default "16:9"	"adaptive" Send 'adaptive' so the output follows the source; when omitted, 16:9 is used.
generate_audio boolean default true	Generate audio.
input_video_duration number affects price	0–15 Source length in seconds, rounded up. Always send it with the source video.

Input files

video: 1 source video URL in videos[] (1-15s, up to 100MB)

Upload folders for /api/storage/upload

video: wan30-extend

Rules

Send input_video_duration equal to the source video length (seconds); values > 15 are rejected (400).
duration must be an integer 2-30; -1 is rejected; source + duration must be <= 30 (400).
At most 5 videos, 10 images and 5 audios (400); only the source video is needed.
A source video in videos[] is required.
When quoting with /api/quote, add has_video_input: true and input_video_duration if you send the source video (the quote endpoint does not read video URLs, so the price would come out too low).

Example request body

{
  "mode": "wan-3.0",
  "generation_type": "video-extend",
  "prompt": "She turns around and waves at the camera",
  "videos": [
    "https://cdn.veevid.ai/wan30-extend/clip.mp4"
  ],
  "duration": "5",
  "video_quality": "720p",
  "aspect_ratio": "adaptive",
  "generate_audio": true,
  "input_video_duration": 6
}

Seedance 2.0 mode: seedance-2.0

edit-video

Edits a source video (up to 15s) using a prompt plus optional reference images and audio.

Parameter	Values & notes
prompt string · required	Editing instruction (required).
videos string[] · required affects price	[source video URL]. The source length is billed.
images string[]	Optional reference images; up to 9 are used.
audios string[]	Optional reference audio; up to 3 are used.
duration string affects price default "5"	4–15 Output length in seconds, 4-15. Out-of-range values are clamped to 4-15.
video_quality string affects price default "480p"	"480p", "720p", "1080p", "4k" Output resolution. 1080p and 4k require model_version 'seedance-2.0'; with mini or fast use 480p or 720p.
aspect_ratio string default "16:9"	"auto", "16:9", "9:16", "1:1", "4:3", "3:4", "21:9" 'auto' keeps the source ratio. Use one of the listed values.
model_version string affects price default "seedance-2.0"	"seedance-2.0-mini", "seedance-2.0-fast", "seedance-2.0" Quality/price tier. Send `model_version` explicitly.
generate_audio boolean default true	Generate audio.
input_video_duration number affects price	0–15 Source video length in seconds (billed as at least 2s). Always send it with the source video.
has_video_input boolean	Quote-only field: send `true` to `/api/quote` together with the source video length; `/api/generate-video` ignores it.

Input files

video: 1 source video URL in videos[] (up to 15s and 200MB)
images: 0-9 reference image URLs (up to 30MB each)
audio: 0-3 reference audio URLs in audios[] (.wav/.mp3, up to 15MB each, 15s total)

Upload folders for /api/storage/upload

video: seedance20-edit
image: seedance20
audio: seedance20-edit/audio

Rules

Send input_video_duration equal to the source video length (seconds); values > 15 are rejected (400).
Send model_version explicitly; with seedance-2.0-mini or seedance-2.0-fast use 480p or 720p.
videos[] must contain the source video (400 'Video is required for edit-video generation').
Only the first 9 images, 3 videos and 3 audios are used; duration is normalized to 4-15 rather than rejected.
Not supported: output_format (mp4/mov is only valid for Seedance 2.5).
When quoting with /api/quote, add has_video_input: true and input_video_duration if you send the source video (the quote endpoint does not read video URLs, so the price would come out too low).

Example request body

{
  "mode": "seedance-2.0",
  "generation_type": "edit-video",
  "prompt": "Make it snow in the scene",
  "videos": [
    "https://cdn.veevid.ai/seedance20-edit/clip.mp4"
  ],
  "duration": "5",
  "video_quality": "480p",
  "aspect_ratio": "auto",
  "model_version": "seedance-2.0-fast",
  "generate_audio": true,
  "input_video_duration": 8
}

video-extend

Continues an uploaded video by 4-15s with optional reference images and audio.

Parameter	Values & notes
prompt string	What happens next.
videos string[] · required affects price	[source video URL]. No taskId needed.
images string[]	Optional reference images; up to 9 are used.
audios string[]	Optional reference audio; up to 3 are used.
duration string affects price default "5"	4–15 Length of the new footage in seconds; clamped to 4-15.
video_quality string affects price default "480p"	"480p", "720p", "1080p", "4k" 1080p and 4k require model_version 'seedance-2.0'; with mini or fast use 480p or 720p.
aspect_ratio string default "16:9"	"auto", "16:9", "9:16", "1:1", "4:3", "3:4", "21:9" 'auto' keeps the source ratio.
model_version string affects price default "seedance-2.0"	"seedance-2.0-mini", "seedance-2.0-fast", "seedance-2.0" Quality/price tier. Send `model_version` explicitly.
generate_audio boolean default true	Generate audio.
input_video_duration number affects price	0–15 Source length in seconds (billed as at least 2s). Always send it with the source video.
has_video_input boolean	Quote-only field: send `true` to `/api/quote` together with the source video length; `/api/generate-video` ignores it.

Input files

video: 1 source video URL in videos[] (<=15s; mp4/mov/webm up to 200MB)
images: 0-9 reference image URLs
audio: 0-3 reference audio URLs in audios[] (.wav/.mp3, up to 15MB each, 15s total)

Upload folders for /api/storage/upload

video: seedance20-extend
image: seedance20-extend
audio: seedance20-extend/audio

Rules

Send input_video_duration equal to the source video length (seconds); values > 15 are rejected (400).
Send model_version explicitly; with seedance-2.0-mini or seedance-2.0-fast use 480p or 720p.
The source video is required in videos[].
Only the first 9 images, 3 videos and 3 audios are used.
Reference images uploaded to the seedance20-extend folder must be.jpg/.jpeg/.png/.tiff/.gif (not.webp).
When quoting with /api/quote, add has_video_input: true and input_video_duration if you send the source video (the quote endpoint does not read video URLs, so the price would come out too low).

Example request body

{
  "mode": "seedance-2.0",
  "generation_type": "video-extend",
  "prompt": "The camera follows the runner across the bridge",
  "videos": [
    "https://cdn.veevid.ai/seedance20-extend/clip.mp4"
  ],
  "duration": "5",
  "video_quality": "480p",
  "aspect_ratio": "auto",
  "model_version": "seedance-2.0-fast",
  "generate_audio": true,
  "input_video_duration": 8
}

Seedance 2.5 mode: seedance-2.5

edit-video

Edits a source video (up to 30s) with a prompt and optional references; the output keeps the source's length and aspect ratio.

Parameter	Values & notes
prompt string · required	Editing instruction. It is automatically prefixed with 'Edit the video: ' when no editing keyword (edit/add/remove/replace/change/...) is present.
videos string[] · required affects price	[source video URL].
images string[]	Optional reference images; up to 30 are used.
audios string[]	Optional reference audio; up to 10 are used.
duration string · required affects price	"-1" Send "-1". Edit output always matches the source length.
video_quality string affects price default "480p"	"480p", "720p", "1080p" Output resolution. Use one of the listed values; 1080p output is 10-bit HEVC.
aspect_ratio string default "auto"	"auto" Send 'auto' or omit; edits always keep the source aspect ratio.
generate_audio boolean default true	Generate audio.
input_video_duration number affects price	0–30 Source video length in seconds (billed as at least 2s). Always send it with the source video.
output_format string default "mp4"	"mp4", "mov" Optional output container: mp4 or mov.
has_video_input boolean	Quote-only field: send `true` to `/api/quote` together with the source video length; `/api/generate-video` ignores it.

Input files

video: 1 source video URL in videos[] (up to 30s and 200MB)
images: 0-30 reference image URLs (up to 30MB each)
audio: 0-10 reference audio URLs in audios[] (.wav/.mp3, up to 15MB each)

Upload folders for /api/storage/upload

video: seedance25-edit
image: seedance25
audio: seedance25-edit/audio

Rules

duration must be an integer 4-30 or -1 (400); for edits send -1.
Send input_video_duration equal to the source video length (seconds); values > 30 are rejected (400).
output_format accepts only mp4 or mov (400 otherwise).
videos[] must contain the source (400 'Video is required for edit-video generation').
480p results are usually drafts: the status response then includes seedance25_draft, and the draft can be upgraded once to a 1080p version within 7 days with POST /api/video-generation/{id}/upgrade-draft.
When quoting with /api/quote, add has_video_input: true and input_video_duration if you send the source video (the quote endpoint does not read video URLs, so the price would come out too low).

Example request body

{
  "mode": "seedance-2.5",
  "generation_type": "edit-video",
  "prompt": "Replace the car with a red vintage convertible",
  "videos": [
    "https://cdn.veevid.ai/seedance25-edit/clip.mp4"
  ],
  "duration": "-1",
  "video_quality": "480p",
  "aspect_ratio": "auto",
  "generate_audio": true,
  "input_video_duration": 10
}

video-extend

Continues an uploaded video by 4-30s with optional reference images and audio.

Parameter	Values & notes
prompt string	What happens next. It is automatically prefixed with 'Continue the video: ' when there is no extend keyword (extend/continue/continuation).
videos string[] · required affects price	[source video URL]. No taskId needed.
images string[]	Optional reference images; up to 30 are used.
audios string[]	Optional reference audio; up to 10 are used.
duration string affects price default "5"	4–30 Length of the new footage in seconds. -1 lets the model choose and is billed as the source length (min 4).
video_quality string affects price default "480p"	"480p", "720p", "1080p" Output resolution.
output_format string default "mp4"	"mp4", "mov" mov uses H.264 + PCM audio and usually cannot play in browsers.
aspect_ratio string default "auto"	"auto" The output follows the source aspect ratio.
generate_audio boolean default true	Generate audio.
input_video_duration number affects price	0–30 Source length in seconds, rounded up (billed as at least 2s). Always send it with the source video.
has_video_input boolean	Quote-only field: send `true` to `/api/quote` together with the source video length; `/api/generate-video` ignores it.

Input files

video: 1 source video URL in videos[] (<=30s; mp4/mov/webm up to 200MB)
images: 0-30 reference image URLs
audio: 0-10 reference audio URLs in audios[] (.wav/.mp3, up to 15MB each, 30s total)

Upload folders for /api/storage/upload

video: seedance25-extend
image: seedance25-extend
audio: seedance25-extend/audio

Rules

duration must be an integer 4-30 or -1 (400).
Send input_video_duration equal to the source video length (seconds); values > 30 are rejected (400).
output_format accepts only mp4 or mov (400).
A source video in videos[] is required.
Reference images uploaded to the seedance25-extend folder must be.jpg/.jpeg/.png/.tiff/.gif (not.webp).
480p results are usually drafts: the status response then includes seedance25_draft, and the draft can be upgraded once to a 1080p version within 7 days with POST /api/video-generation/{id}/upgrade-draft.
When quoting with /api/quote, add has_video_input: true and input_video_duration if you send the source video (the quote endpoint does not read video URLs, so the price would come out too low).

Example request body

{
  "mode": "seedance-2.5",
  "generation_type": "video-extend",
  "prompt": "Continue the video: the dog runs to the beach",
  "videos": [
    "https://cdn.veevid.ai/seedance25-extend/clip.mp4"
  ],
  "duration": "5",
  "video_quality": "480p",
  "output_format": "mp4",
  "aspect_ratio": "auto",
  "generate_audio": true,
  "input_video_duration": 8
}

LTX 2.3 mode: ltx-2-3

video-extend

Extends an uploaded video by 1-20s at its start or end, with optional context length control.

Parameter	Values & notes
video string · required	Source video URL (singular field).
prompt string	Optional description of the extension.
duration string affects price default "5"	1–20 Seconds to add (1-20). Decimals are allowed (e.g. 0.5s steps).
extend_mode string default "end"	"start", "end" Where to add footage.
context number	1–20 Seconds of the source to use as context. Omit to let the model maximize context.
video_source string	"upload", "url" 'upload' for files uploaded with `POST /api/storage/upload`; 'url' for an external link, which must pass a HEAD check (supported video type, <=200MB).
aspect_ratio string default "original"	"original" Optional; send 'original' or omit. Output keeps the source aspect ratio.

Input files

video: 1 source video URL (mp4/mov/webm/m4v/gif, up to 20s and 200MB)

Upload folders for /api/storage/upload

video: video-generator-ltx23-extend

Rules

duration must be a number between 1 and 20 (400).
context, if sent, must be 1-20.
video (singular) is required (400 'Video is required for LTX 2.3 extend generation').
No taskId is used; model_version and video_quality have no effect for extends.

Example request body

{
  "mode": "ltx-2-3",
  "generation_type": "video-extend",
  "prompt": "The waves keep rolling in at sunset",
  "video": "https://cdn.veevid.ai/video-generator-ltx23-extend/clip.mp4",
  "duration": "5",
  "extend_mode": "end",
  "aspect_ratio": "original",
  "video_source": "upload"
}

MiniMax H3 Max mode: minimax-h3-max-extend

video-extend

Continues an uploaded clip (1.625-60s) by 5-15s at up to 2K resolution.

Parameter	Values & notes
prompt string · required	What happens next (required).
videos string[] · required	[source video URL]; exactly one. The singular 'video' is also accepted.
duration string affects price default "5"	5–15 Seconds of new footage (whole seconds).
video_quality string affects price default "768p"	"480p", "768p", "1080p", "2k", "2K" Output resolution.
aspect_ratio string default "auto"	"auto", "21:9", "16:9", "4:3", "1:1", "3:4", "9:16" auto (or adaptive) follows the source. A fixed ratio crops the output. If omitted, auto is used, not 16:9.
extend_output string default "extended"	"extended", "continuation" 'extended' returns source + new footage; 'continuation' returns only the new footage plus ~1.6s of the source tail.
prompt_expansion_mode string default "balanced"	"balanced", "disabled" 'balanced' enables prompt expansion; 'disabled' turns it off.
input_video_duration number	1.625–60 Source length in seconds (up to 2 decimals). Used for validation only; the source is not billed.
seed number	Optional integer >= 0.

Input files

video: 1 source video URL in videos[] (1.625-60s, <=50MB, aspect ratio 0.4-2.5; mp4/mov/webm/m4v)

Upload folders for /api/storage/upload

video: minimax-h3-max-extend

Rules

The source length and extend_output do not affect price.
generation_type must be video-extend; duration must be an integer 5-15 (400).
Exactly one source video; prompt is required; images and audio are rejected (400).
aspect_ratio must be auto/adaptive/21:9/16:9/4:3/1:1/3:4/9:16; video_quality must be 480p/768p/1080p/2k/2K (400).
A declared input_video_duration must be 1.625-60s (400).
The source video must be <=50MB with an aspect ratio between 0.4 and 2.5; out-of-spec sources fail during generation. Upload the source with POST /api/storage/upload (the upload only checks the 50MB size).

Example request body

{
  "mode": "minimax-h3-max-extend",
  "generation_type": "video-extend",
  "prompt": "The cat jumps onto the windowsill and looks outside",
  "videos": [
    "https://cdn.veevid.ai/minimax-h3-max-extend/clip.mp4"
  ],
  "duration": "5",
  "video_quality": "768p",
  "aspect_ratio": "auto",
  "extend_output": "extended",
  "prompt_expansion_mode": "balanced",
  "input_video_duration": 6.04
}

MiniMax H3 Max mode: minimax-h3-max-insert

video-insert

Inserts a newly generated 5-13s scene between two time points of a source video.

Parameter	Values & notes
video string · required	Source video URL (singular field; 'videos' means reference videos here).
prompt string	Describes the new scene. Required unless images or videos are given.
duration string affects price default "5"	5–13 Length of the new scene in seconds; decimals allowed.
video_quality string affects price default "768p"	"480p", "768p" Output resolution.
start_time number · required	1.625–60 Cut-out point in seconds (24 fps frame grid).
resume_time number · required	1.667–60 Resume point in seconds. It must be at least 1 frame (1/24s) after start_time and leave >= 33/24s (+0.25s safety margin) of the source after it. Source between start and resume is replaced.
images string[] affects price	Reference images for the new scene, up to 9.
videos string[] affects price	Reference videos, up to 3. Each must be >= 2s, <= 15s in total, and uploaded with POST /api/storage/upload.
reference_video_duration number affects price	0–15 Send the total reference-video length in seconds. The server also measures the files and bills the measured value if it is more than 0.5s longer; if it is omitted with reference videos, the 15-second maximum is billed.
input_video_duration number	0–60 Source video length in seconds; used to validate resume_time. The source is not billed.
color_match boolean default true	Match the new scene's color grading to the source.
prompt_expansion_mode string default "balanced"	"balanced", "disabled" Only 'disabled' turns prompt expansion off.
seed number	Optional integer >= 0.

Input files

video: 1 source video URL in 'video' (about 3.3-60s, <=50MB; mp4/mov/webm/m4v)
images: 0-9 reference image URLs in 'images' (jpg/png/webp)
videos: 0-3 reference video URLs in 'videos' (mp4/mov, each >=2s, total <=15s, <=100MB each, uploaded with POST /api/storage/upload)

Upload folders for /api/storage/upload

video: minimax-h3-max-insert
images: minimax-h3-max-insert/reference-images
videos: minimax-h3-max-insert/reference-videos

Rules

When sending reference videos, always send reference_video_duration equal to their total length (seconds).
generation_type must be video-insert; video_quality must be 480p or 768p (400).
duration must be 5-13 (400).
Exactly one source in 'video'. Reference images must use 'images'; a singular 'image' is rejected (400).
A prompt, a reference image or a reference video is required; audio is rejected (400).
Reference videos must be uploaded with POST /api/storage/upload first; external URLs are rejected (400). Each must be >= 2s, the total <= 15s, and at most 3 videos and 9 images.
start_time must be 1.625-60. resume_time must be > start_time by >= 1/24s and <= 60. With a declared source length, resume_time must be <= floor((source - 1.375 - 0.25) x 24)/24 (400).
A declared source length must be ~3.29-60s (400).
When quoting, send input_image_count for reference images, and for reference videos add has_video_input: true, input_video_count and reference_video_duration (the quote endpoint does not read these files, so the price would come out too low).

Example request body

{
  "mode": "minimax-h3-max-insert",
  "generation_type": "video-insert",
  "prompt": "A hot air balloon drifts across the sky",
  "video": "https://cdn.veevid.ai/minimax-h3-max-insert/clip.mp4",
  "duration": "5",
  "video_quality": "768p",
  "start_time": 3,
  "resume_time": 3.042,
  "input_video_duration": 10,
  "color_match": true,
  "prompt_expansion_mode": "balanced"
}

MiniMax H3 Max mode: minimax-h3-max-recast

character-swap

Swaps the people in a 5-30s video with people from 1-4 reference photos, keeping motion, camera and original audio.

Parameter	Values & notes
video string · required	Source video URL (singular field). Must be uploaded with POST /api/storage/upload, MP4/MOV, 5-30s, and no single shot over 15s.
images string[] · required	1-4 reference photos, one per new person, mapped left to right by default.
prompt string	Optional (<=2000 chars), e.g. which person becomes whom.
video_quality string affects price default "1080p"	"768p", "1080p" Output resolution. Always send video_quality.
input_video_duration number affects price	5–30 Source length in seconds (up to 2 decimals). Send the actual length; the server measures the file and bills the measured length.
seed number	Optional integer >= 0.

Input files

video: 1 source video URL in 'video' (MP4/MOV, 5-30s, <=50MB, uploaded with POST /api/storage/upload)
images: 1-4 reference photo URLs in 'images' (jpg/png/webp, <=20MB each)

Upload folders for /api/storage/upload

video: minimax-h3-max-recast
images: minimax-h3-max-recast/reference-images

Rules

Seconds = server-measured source length (minimum 5). Reference photos are free.
generation_type must be character-swap; video_quality must be 768p or 1080p (400). Always send video_quality.
Input files must be uploaded with POST /api/storage/upload first (external URLs are rejected).
If the source video's duration cannot be read from its MP4/MOV header, the request returns 422 and nothing is charged.
A measured length outside 5-30s (+-0.1s) is rejected (400).
If the measured length exceeds the declared input_video_duration by >0.1s, it returns 409 with measured_duration and required_credits and charges nothing; resubmit with input_video_duration = measured_duration.
1-4 images required; a singular 'image', reference videos or audio are rejected (400); prompt <= 2000 chars.
Duration and aspect ratio are not inputs: the output matches the source.

Example request body

{
  "mode": "minimax-h3-max-recast",
  "generation_type": "character-swap",
  "prompt": "Replace the man on the left with the person in photo 1",
  "video": "https://cdn.veevid.ai/minimax-h3-max-recast/clip.mp4",
  "images": [
    "https://cdn.veevid.ai/minimax-h3-max-recast/reference-images/person1.jpg"
  ],
  "video_quality": "768p",
  "input_video_duration": 8.04
}

Video Lip Sync mode: video-lip-sync

lip-sync

Re-syncs the mouth movements in an existing video to a new voice track; output length equals the audio length.

Parameter	Values & notes
videos string[] · required	[source video URL]; only the first is used. The singular 'video' is also accepted. Must be uploaded with POST /api/storage/upload.
audios string[] · required	[voice audio URL]; only the first is used. The singular 'audio' is also accepted. Must be uploaded with POST /api/storage/upload; use clean vocals.
input_audio_duration number affects price	0–61 Send the audio length in seconds.
lip_sync_mode string default "lite"	"lite", "basic" lite: single front-facing speaker, faster. basic: complex scenes, multiple shots.
lip_sync_separate_vocal boolean default false	Isolate vocals / suppress background noise.
lip_sync_open_scenedet boolean default true	Shot and speaker detection (basic only).
lip_sync_align_audio boolean default true	Loop the video when the audio is longer (lite only; always on in basic).
lip_sync_align_audio_reverse boolean default false	Loop by playing in reverse (lite only; requires align_audio).
lip_sync_templ_start_seconds number default 0	0–3600 Start offset in the source video, floored to whole seconds (lite only).

Input files

video: 1 source video URL (mp4/mov, <=100MB, short side >=360p, uploaded with POST /api/storage/upload)
audio: 1 voice audio URL (mp3/wav/m4a/aac/ogg, <=10MB, <=60s, uploaded with POST /api/storage/upload)

Upload folders for /api/storage/upload

video: video-lip-sync
audio: video-lip-sync/audio

Rules

Always send input_audio_duration equal to the audio length (seconds); if it is omitted you are charged for the 60-second maximum. If the output is longer than declared, the difference is charged after processing (an over-declared length is not refunded).
generation_type must be lip-sync (422).
The video and audio must be uploaded with POST /api/storage/upload first; external URLs are rejected (422).
input_audio_duration > 61 is rejected (422; the limit is 60s plus 1s tolerance).
No prompt, duration or aspect ratio: the output length equals the audio and the resolution follows the source.

Example request body

{
  "mode": "video-lip-sync",
  "generation_type": "lip-sync",
  "videos": [
    "https://cdn.veevid.ai/video-lip-sync/clip.mp4"
  ],
  "audios": [
    "https://cdn.veevid.ai/video-lip-sync/audio/voice.mp3"
  ],
  "input_audio_duration": 12.4,
  "lip_sync_mode": "lite",
  "lip_sync_separate_vocal": false,
  "lip_sync_open_scenedet": true,
  "lip_sync_align_audio": true,
  "lip_sync_align_audio_reverse": false,
  "lip_sync_templ_start_seconds": 0
}

Google Veo 3.1 mode: veo3

video-extend

Extends a Veo 3.1 video you generated earlier, identified by its task ID.

Parameter	Values & notes
taskId string · required affects price	The provider_task_id of one of your own earlier Veo 3.1 generations (returned when it was created and by the status endpoint).
prompt string · required	What happens in the extension. Required.
generation_mode string affects price default "fast"	"lite", "fast", "quality" Extension quality tier.
watermark string	Optional custom watermark text, <=200 chars.
duration string default "8"	"4", "6", "8" Must be 4, 6 or 8 if sent; send "8".
aspect_ratio string default "original"	"original" Send "original" (the value is not used).

Rules

taskId is required (400). It must match one of your generations (404 'Task not found') owned by you (403).
The source must be a Veo 3.1 or Grok Imagine video (400). A Grok Imagine taskId is run and priced as a Grok Imagine extend.
1080p and 4k Veo sources can be extended.
When taskId belongs to a Grok Imagine video, the request is run and billed as a Grok Imagine extend: send duration "6" or "10" (or omit it), not "8".

Example request body

{
  "mode": "veo3",
  "generation_type": "video-extend",
  "taskId": "veo_task_abc123",
  "prompt": "The drone keeps flying over the city at night",
  "generation_mode": "fast",
  "duration": "8",
  "aspect_ratio": "original"
}

Grok Imagine mode: grok-imagine

video-extend

Extends a Grok Imagine video you generated earlier by 6 or 10 seconds from a chosen point.

Parameter	Values & notes
taskId string · required affects price	The provider_task_id of one of your own earlier Grok Imagine generations.
prompt string · required	What happens in the extension. Required.
duration string affects price default "6"	"6", "10" Seconds to add.
extend_at number default 0	Start position in seconds (>= 0); 0 extends from the end of the video.
aspect_ratio string default "original"	"original" Send "original" (the value is not used).

Rules

duration must be '6' or '10' (400 'Grok Imagine Extend only supports 6s or 10s durations').
taskId is required, must exist (404) and belong to you (403). The source must be a Grok Imagine or Veo 3.1 video; a Veo 3.1 taskId is run and priced as a Veo 3.1 extend.

Example request body

{
  "mode": "grok-imagine",
  "generation_type": "video-extend",
  "taskId": "grok_task_abc123",
  "prompt": "The knight raises his sword as the dragon lands",
  "duration": "6",
  "extend_at": 0,
  "aspect_ratio": "original"
}

Code Examples

Generate an image (curl)

curl -X POST https://veevid.ai/api/generate-video \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "mode": "nano-banana-2",
    "generation_type": "text-to-image",
    "prompt": "A cozy coffee shop interior, morning light, watercolor style",
    "aspect_ratio": "1:1",
    "video_quality": "1K"
  }'

To edit your own photo, upload it to /api/storage/upload first and send "generation_type": "image-to-image" with "images": ["UPLOADED_URL"].

Text to speech (curl)

curl -X POST https://veevid.ai/api/generate-video \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "mode": "eleven-v4-turbo",
    "generation_type": "text-to-speech",
    "prompt": "Welcome to Veevid. Let us make something amazing today.",
    "voice": "Rachel"
  }'

Poll the status endpoint; the audio file is in video_url.

Python

import requests
import time

API_KEY = "vv_sk_your_key_here"
BASE = "https://veevid.ai/api"
HEADERS = {"Authorization": f"Bearer {API_KEY}"}

# 1. Upload a first frame
with open("first_frame.jpg", "rb") as f:
    upload = requests.post(
        f"{BASE}/storage/upload",
        headers=HEADERS,
        files={"file": f},
        data={"folder": "video-generator-veo3"},
    ).json()

body = {
    "mode": "veo3",
    "generation_type": "image-to-video",
    "prompt": "The camera slowly pushes in, leaves rustle in the wind",
    "images": [upload["url"]],
    "video_quality": "standard",
}

# 2. Quote
quote = requests.post(f"{BASE}/quote", headers=HEADERS, json=body).json()
print(f"Cost: {quote['required_credits']} credits")

# 3. Generate
result = requests.post(f"{BASE}/generate-video", headers=HEADERS, json=body).json()
gen_id = result["generation_id"]

# 4. Poll
while True:
    status = requests.get(
        f"{BASE}/video-generation/{gen_id}/status", headers=HEADERS
    ).json()
    if status["status"] == "completed":
        print(f"Ready: {status['video_url']}")
        break
    if status["status"] == "failed":
        print(f"Failed: {status['error_message']}")
        break
    time.sleep(10)

JavaScript / TypeScript

const API_KEY = "vv_sk_your_key_here";
const BASE = "https://veevid.ai/api";
const headers = {
  Authorization: \`Bearer ${API_KEY}\`,
  "Content-Type": "application/json",
};

async function generate(body: Record<string, unknown>) {
  // 1. Quote
  const quote = await fetch(\`${BASE}/quote\`, {
    method: "POST",
    headers,
    body: JSON.stringify(body),
  }).then((r) => r.json());
  console.log(\`Cost: ${quote.required_credits} credits\`);

  // 2. Generate
  const result = await fetch(\`${BASE}/generate-video\`, {
    method: "POST",
    headers,
    body: JSON.stringify(body),
  }).then((r) => r.json());

  // 3. Poll
  while (true) {
    const status = await fetch(
      \`${BASE}/video-generation/${result.generation_id}/status\`,
      { headers }
    ).then((r) => r.json());

    if (status.status === "completed") return status.video_url;
    if (status.status === "failed") throw new Error(status.error_message);
    await new Promise((r) => setTimeout(r, 10000));
  }
}

await generate({
  mode: "kling-3",
  generation_type: "text-to-video",
  prompt: "A drone shot flying over a neon-lit city at night",
  duration: "5",
  generate_audio: true,
});

OpenClaw Agent

Install the Veevid skill:

npx clawhub@latest install veevid

Then tell your agent: "Generate a 10-second product video with Kling 3.0"

The agent handles quoting, confirmation, generation, and polling automatically.

Error Codes

Code	Meaning	What to Do
400	Invalid parameters	Check allowed values for the model
401	Invalid or missing API key	Verify your key at /settings/api-keys
402	Insufficient credits	Top up at /pricing
403	Account suspended	Contact support
404	Generation not found	Check the generation_id
409	Conflict: the input video is longer than the declared length (character swap), or a draft is not ready / already upgraded	Nothing was charged; read the error — for character swap, resend with the measured_duration from the response
410	The draft has expired	Generate again
422	Parameter combination not supported by the model	Read the error message and the model rules
500	Server error	Wait a few seconds and retry
502	A draft upgrade could not be submitted (credits refunded)	Wait a few seconds and retry

Error response format:

{
  "error": "Insufficient credits. You need 140 credits but only have 12.",
  "required": 140,
  "balance": 12
}

Limits

Limit	Value
API keys per account	5 active keys
Prompt length	Up to 20,000 characters (model-specific)
Upload size	Per folder: 10MB for images by default, larger for video/audio folders, and at most 100MB per upload request; 512MB via presigned upload (clean audio)
Request rate	No fixed per-key rate limit today — your credit balance is the limit. Please poll status every 5–10 seconds rather than continuously.