ai-avatar-video

作者: halt-catch-fire

透過 inference.sh CLI 建立 AI 虛擬角色與說話頭像影片。推薦:P-Video-Avatar(最快、最便宜、內建 TTS)。另可選:OmniHuman、Fabric、PixVerse。音訊:Inworld TTS-2(支援 100 多種語言、角色情感引導)、ElevenLabs、Kokoro。功能:音訊驅動虛擬角色、文字轉虛擬角色、唇形同步影片、說話頭像生成、虛擬主持人、UGC 內容。適用於:AI 主持人、解說影片、虛擬網紅、配音、行銷影片、UGC 廣告、遊戲虛擬角色等。

npx skills add https://github.com/halt-catch-fire/skills --skill ai-avatar-video

Install the belt CLI skill: npx skills add belt-sh/cli

AI Avatar & Talking Head Videos

Create AI avatars and talking head videos via inference.sh CLI.

AI Avatar & Talking Head Videos

Quick Start

Requires inference.sh CLI (belt). Install instructions

belt login

# Recommended: P-Video-Avatar (fastest, cheapest, built-in TTS)
belt app run pruna/p-video-avatar --input '{
  "image": "https://portrait.jpg",
  "voice_script": "Hello, welcome to our product demo!",
  "voice": "Zephyr (Female)"
}'

Available Models

Start with P-Video-Avatar — it's 18x faster and 6x cheaper than alternatives, with built-in TTS, dynamic backgrounds, and 1080p support.

ModelApp IDBest ForBuilt-in TTS
P-Video-Avatarpruna/p-video-avatarBest overall: speed, cost, quality, controlYes (30 voices, 10 languages)
OmniHuman 1.5bytedance/omnihuman-1-5Multi-character, audio-drivenNo
Fabric 1.0falai/fabric-1-0Image talks with lipsyncYes
PixVerse Lipsyncfalai/pixverse-lipsyncLipsync an existing videoYes

Cost & Speed Comparison

ModelSpeed (per sec of video)Cost per second
P-Video-Avatar~1.83s/s$0.025
OmniHuman 1.5~28s/s (15x slower)$0.16 (6.4x more)
Fabric 1.0~34s/s (18x slower)$0.14 (5.6x more)

Examples

P-Video-Avatar (Recommended)

Generate avatar from portrait + text script with built-in TTS:

belt app run pruna/p-video-avatar --input '{
  "image": "https://portrait.jpg",
  "voice_script": "Welcome to our product walkthrough. Today I will show you three key features.",
  "voice": "Puck (Male)",
  "voice_language": "English (US)",
  "resolution": "720p"
}'

With custom style control:

belt app run pruna/p-video-avatar --input '{
  "image": "https://portrait.jpg",
  "voice_script": "This is exciting news!",
  "voice": "Aoede (Female)",
  "voice_prompt": "Enthusiastic and energetic tone",
  "video_prompt": "The person is presenting on stage with dramatic lighting",
  "resolution": "1080p"
}'

With audio file instead of TTS:

belt app run pruna/p-video-avatar --input '{
  "image": "https://portrait.jpg",
  "audio": "https://speech.mp3"
}'

Full Workflow: Generate Portrait + Avatar

Use Pruna P-Image to generate the portrait, then create the avatar:

# 1. Generate a portrait image
belt app run pruna/p-image --input '{
  "prompt": "professional headshot portrait of a young woman, neutral background, looking at camera, studio lighting, photorealistic",
  "aspect_ratio": "9:16"
}'

# 2. Create avatar video with built-in TTS
belt app run pruna/p-video-avatar --input '{
  "image": "<image-url-from-step-1>",
  "voice_script": "Hi there! Let me walk you through our latest features.",
  "voice": "Zephyr (Female)"
}'

OmniHuman 1.5 (Multi-Character)

belt app run bytedance/omnihuman-1-5 --input '{
  "image": "https://portrait.jpg",
  "audio": "https://speech.mp3"
}'

Supports specifying which character to drive in multi-person images.

Fabric 1.0 (Image Talks)

belt app run falai/fabric-1-0 --input '{
  "image": "https://face.jpg",
  "audio": "https://audio.mp3"
}'

PixVerse Lipsync

belt app run falai/pixverse-lipsync --input '{
  "video": "https://talking-head.mp4",
  "audio": "https://speech.mp3"
}'

Takes a video, not a still image. Omit audio and pass text (plus optional voice_id) to use the built-in TTS.

Full Workflow: TTS + Avatar (Non-TTS Models)

For models without built-in TTS (OmniHuman), generate speech first:

# 1. Generate speech — Inworld TTS-2 for expressive character voices
belt app run inworld/text-to-speech-2 --input '{
  "text": "[friendly] Welcome to our product demo! [excited] Let me show you three features that will change how you work.",
  "voice_id": "Sarah",
  "delivery_mode": "CREATIVE"
}' > speech.json

# 2. Create avatar video with the speech
belt app run bytedance/omnihuman-1-5 --input '{
  "image": "https://presenter-photo.jpg",
  "audio": "<audio-url-from-step-1>"
}'

Tip: For most use cases, P-Video-Avatar with built-in TTS is simpler — no separate audio step needed. Use this workflow only when you specifically need OmniHuman (multi-character) or PixVerse (realistic lipsync).

Full Workflow: Dub Video in Another Language

# 1. Transcribe original video
belt app run openai/gpt-transcribe --input '{"audio": "https://video.mp4"}' > transcript.json

# 2. Translate text (manually or with an LLM)

# 3. Generate speech in new language
belt app run falai/kokoro-tts --input '{"prompt": "<translated-text>"}' > new_speech.json

# 4. Lipsync the original video with new audio
belt app run infsh/latentsync-1-6 --input '{
  "video_path": "https://original-video.mp4",
  "audio_path": "<new-audio-url>"
}'

Avatar UGC Generation

Create UGC-style content with P-Video-Avatar — built-in TTS, no separate audio step needed:

# 1. Generate a relatable UGC-style portrait
belt app run pruna/p-image --input '{
  "prompt": "casual selfie-style photo of a young woman in a cozy room, natural lighting, looking at camera, warm smile, authentic feel",
  "aspect_ratio": "9:16"
}'

# 2. Create UGC avatar video with built-in TTS
belt app run pruna/p-video-avatar --input '{
  "image": "<image-url-from-step-1>",
  "voice_script": "Okay so I just tried this product and honestly? It is a game changer. I was not expecting to love it this much but here we are!",
  "voice": "Zephyr (Female)",
  "voice_prompt": "Excited, casual, authentic tone like talking to a friend",
  "video_prompt": "The person is talking casually to camera in their room, natural gestures",
  "resolution": "1080p"
}'

Why P-Video-Avatar for UGC

  • All-in-one — built-in TTS means no separate audio generation step
  • 30 voices, 10 languages — match your target audience
  • Voice + video prompts — control tone, emotion, body language, and background independently
  • 18x faster, 6x cheaper — produce UGC at scale vs. Fabric/OmniHuman/HeyGen
  • 1080p support — platform-ready vertical video from a single portrait image

Batch UGC: Same Product, Multiple Presenters

# Generate 3 different presenters
for voice in "Zephyr (Female)" "Puck (Male)" "Aoede (Female)"; do
  belt app run pruna/p-video-avatar --input "{
    \"image\": \"https://portrait.jpg\",
    \"voice_script\": \"This changed my morning routine completely. Five minutes and I am done.\",
    \"voice\": \"$voice\",
    \"voice_prompt\": \"Casual, authentic, like a real testimonial\",
    \"video_prompt\": \"Person talking to camera in a bright kitchen\",
    \"resolution\": \"1080p\"
  }"
done

Use Cases

  • UGC & Marketing: Product demos, UGC-style ads with AI presenters
  • Education: Course videos, explainers
  • Localization: Dub content across 10 languages from one image
  • Social Media: Consistent virtual influencer content
  • Corporate: Training videos, announcements
  • Gaming: Character avatars, NPC dialogue

Tips

  • Use high-quality portrait photos (front-facing, good lighting)
  • Audio should be clear with minimal background noise
  • P-Video-Avatar supports built-in TTS — no need for a separate speech generation step
  • P-Video-Avatar output aspect ratio matches the input image
  • Generate portraits with pruna/p-image using 9:16 aspect ratio for vertical videos
  • OmniHuman 1.5 supports multiple people in one image
  • LatentSync is best for syncing existing videos to new audio

Related Skills

# Dedicated P-Video-Avatar skill
npx skills add inference-sh/skills@p-video-avatar

# Full platform skill (all apps)
npx skills add inference-sh/skills@infsh-cli

# Text-to-speech (generate audio for non-TTS avatar models)
npx skills add inference-sh/skills@text-to-speech

# Speech-to-text (transcribe for dubbing)
npx skills add inference-sh/skills@speech-to-text

# Video generation
npx skills add inference-sh/skills@ai-video-generation

# Image generation (create avatar images)
npx skills add inference-sh/skills@ai-image-generation

Browse all video apps: belt app list --category video

Documentation

來自 halt-catch-fire 的更多技能

agent-browser
halt-catch-fire
透過 inference.sh 為 AI 代理提供瀏覽器自動化功能。可導覽網頁、使用 @e 參考與元素互動、擷取螢幕截圖、錄製影片。功能包括:網頁抓取、表單填寫、點擊、打字、拖放、檔案上傳、執行 JavaScript。適用於:網頁自動化、資料擷取、測試、代理瀏覽、研究。觸發詞:瀏覽器、網頁自動化、抓取、導覽、點擊、填寫表單、螢幕截圖、瀏覽網頁、playwright、無頭瀏覽器、網頁代理、上網、錄製影片
browser-automationweb-scrapingtesting
agent-tools
halt-catch-fire
透過 inference.sh CLI 執行 250 多個 AI 應用程式,包括圖片生成、影片創作、大型語言模型、搜尋、3D、Twitter 自動化。模型:FLUX、Veo、Gemini、Grok、Claude、Seedance、OmniHuman、Tavily、Exa、OpenRouter 等。適用於執行 AI 應用程式、生成圖片/影片、呼叫大型語言模型、網路搜尋或自動化 Twitter。觸發詞:inference.sh、infsh、ai model、run ai、serverless ai、ai api、flux、veo、claude api、image generation、video generation、openrouter、tavily、exa search、twitter api、grok
developmentapicreative
ai-image-generation
halt-catch-fire
透過 inference.sh CLI 使用 GPT-Image-2、FLUX、Gemini、Grok、Seedream、Reve 及 50 多種模型生成 AI 圖像。模型包括:GPT-Image-2、FLUX Dev LoRA、FLUX.2 Klein LoRA、Gemini 3 Pro Image、Grok Imagine、Seedream 4.5、Reve、ImagineArt。功能:文字轉圖像、圖像轉圖像、修補、LoRA、圖像編輯、放大、文字渲染。適用於:AI 藝術、產品模型、概念藝術、社交媒體圖形、行銷視覺、插圖。觸發詞:flux、圖像生成、AI 圖像、文字轉...
creativemediaimage
ai-video-generation
halt-catch-fire
We need to translate the given text from English to Traditional Chinese. The instruction says to preserve product names, protocol names, URLs, numbers, and technical terms. So names like Google Veo, Seedance 2.0, HappyHorse, Wan, Grok, inference.sh CLI, Veo 3.1, etc. should remain as is. Also preserve terms like text-to-video, image-to-video, etc. but translate the surrounding Chinese text. The text inside <text> is a description of an agent skill. We need to output only the translated text, no extra labels. Let's translate paragraph by paragraph: "Generate AI videos with Google Veo, Seedance 2.0, HappyHorse, Wan, Grok and 40+ models via inference.sh CLI." -> "通過 inference.sh CLI 使用 Google Veo、Seedance 2.0、HappyHorse、Wan、Grok 及 40 多個模型生成 AI 影片。" "Models: Veo 3.1, Veo 3,
creativevideomedia
app-store-screenshots
halt-catch-fire
依據精確的平台規格製作App Store與Google Play截圖,涵蓋iOS/Android尺寸、圖庫排序、裝置模擬機及預覽影片。適用於:App Store優化、ASO、應用程式截圖、應用程式預覽、Play Store上架。觸發詞:App Store截圖、ASO、App Store優化、Play Store截圖、應用程式預覽、應用程式上架、iOS截圖、Android截圖、App Store圖片、應用程式模擬機、裝置模擬機、應用程式圖庫、商店上架
designmarketingcreative
character-design-sheet
halt-catch-fire
透過參考圖與LoRA技術,確保AI生成角色圖像的一致性。涵蓋轉面視圖、表情圖譜、色票與風格一致性技巧。適用於:角色設計、遊戲美術、插畫、動畫、漫畫、視覺小說。觸發詞:角色設計、角色圖譜、角色一致性、角色參考、轉面圖、表情圖譜、角色美術、一致角色、角色概念、參考圖、角色創作、原創角色設計...
creativedesignmedia
competitor-teardown
halt-catch-fire
結構化競爭分析,包含功能矩陣、SWOT、定位圖與UX審查。涵蓋研究框架、定價比較、評論挖掘與視覺化交付成果。適用於:市場研究、競爭情報、投資人簡報、產品策略、銷售賦能。觸發詞:競爭者分析、競爭分析、競爭者拆解、市場研究、競爭情報、SWOT分析、競爭者比較、市場格局、競爭者審查、競爭...
researchdata-analysismarketing
image-to-video
halt-catch-fire
靜態轉影片轉換指南:模型選擇、動態提示與鏡頭移動。涵蓋Wan 2.5 i2v、Seedance、Fabric、Grok Video及其適用時機。用途:為圖片添加動畫、從靜態圖像製作影片、加入動態效果、產品動畫。觸發詞:image to video、i2v、animate image、still to video、add motion to image、image animation、photo to video、animate still、wan i2v、image2video、bring image to life、animate photo、motion from image