WaveLocus
Überprüfe Sprachaufzeichnungspegel mit WaveLocus in ChatGPT. Audio wird lokal in deinem Browser analysiert, und nur freigegebene Messwerte werden geteilt. Derzeit werden ChatGPT-Verbindungen unterstützt.
Gehosteter MCP-Server
npx add-mcp 'https://assistant.wavelocus.com/mcp'Installiert in Claude Code, Codex, Cursor und mehr
Dokumentation
Availability in ChatGPT
OpenAI has approved WaveLocus for ChatGPT.
What WaveLocus does
WaveLocus measures the digital levels of short voice recordings in your browser. Use it to see whether a narration excerpt has room for a volume increase, check how close its highest sample is to full scale, or compare an original excerpt with a revised copy of the same passage. ChatGPT can explain the measurements you choose to share.
How to use WaveLocus in ChatGPT
If WaveLocus is available in your ChatGPT app directory, follow these steps.
- Find WaveLocus in ChatGPT’s app directory, connect it, and select it in a conversation.
- Ask for a recording review. You can start with the example prompt below.
- Open the private review link within 10 minutes. Confirm that you requested the connection, then choose Connect assistant.
- Add your recording and run the analysis. To compare an edit, start with the original, then use Check a file revised in your editor for the matching revised passage.
- Review the results and approve Share report when ready. Return to the same ChatGPT conversation and say, “I shared my recording report.”
If your link expires, ask for a new recording review in ChatGPT.
Use WaveLocus in Claude
Claude on the web and in the desktop app can connect to WaveLocus as a custom connector. You set it up once.
- In Claude, open Customize > Connectors, choose +, then Add custom connector. On Team and Enterprise plans, an organization Owner adds the connector and members then connect to it.
- Name it WaveLocus, enter the server URL
https://assistant.wavelocus.com/mcp, and choose Continue. - For authentication, choose Sign in now. Keep the OAuth client option Use Claude’s published identity, then choose Add.
- Choose Connect. Review the WaveLocus approval page and approve only a connection you started.
- Ask Claude for a recording review, then follow the same review and Share report steps as above. Afterwards, return to your Claude conversation and say, “I shared my recording report.”
If your organization requires its own OAuth client, choose Use your own OAuth client instead, enter the client ID wavelocus-claude-hosted, and leave the client secret empty.
Use WaveLocus in VS Code
VS Code 1.122 or later can connect to WaveLocus as a remote MCP server for its chat. You set it up once.
- Run MCP: Open User Configuration from the Command Palette, or open
.vscode/mcp.jsonin a workspace. - Add the WaveLocus server with its client ID. Do not add a client secret.
{
"servers": {
"wavelocus": {
"type": "http",
"url": "https://assistant.wavelocus.com/mcp",
"oauth": { "clientId": "wavelocus-vscode" }
}
}
}
- Start the server when VS Code prompts you. Your browser opens the WaveLocus approval page; approve only a connection you started.
- Open Chat in Agent mode and ask for a recording review. When VS Code asks to run a WaveLocus tool, choose Allow. If you skip it, no review link is created.
- Follow the same review and Share report steps as above. Afterwards, return to the chat and say, “I shared my recording report.”
If chat answers without a review link, run Chat: Configure Tools from the Command Palette and make sure the WaveLocus server is selected.
Connections from Claude and VS Code renew without approving again for up to thirty days, and stop renewing after seven days without use. You can approve again at any time. The privacy notice explains what renewal stores.
Worked example
Check recording levels and share the report with ChatGPT.
Supported excerpts
Check up to three mono or stereo excerpts, each up to 60 seconds and 16 MiB. WAV is supported. MP3, M4A/AAC and FLAC depend on your browser's decoding support. Review each recording separately. Different takes are not a controlled before-and-after comparison.
Measurements and limits
The report includes sample peak, whole-excerpt RMS, full-scale samples and DC offset. WaveLocus also estimates true peak and measures gated loudness in LUFS for the selected excerpt when available. Short excerpts, unsupported rates or audio below the loudness gate return an explanation rather than an invented value.
You can select a background-only passage to measure its unweighted digital RMS and sample peak. You identify that passage yourself. This is not verified noise detection, equipment noise floor or SNR. A partial narration-level reference helps interpret sample levels, not a universal recording target or complete publishing approval. Measurements do not assess room echo, intelligibility or overall audio quality. Earlier clipping or distortion can still be present when the report shows zero full-scale samples and an estimated true peak below full scale.
Optional volume adjustment
Where eligible, you can approve a bounded linked gain change and download a new WAV copy. Your original file is unchanged. This changes volume, not inherent audio quality. It cannot remove noise or repair clipping. Listen and recheck before using the result. For a separate file workflow, the audio normalizer targets excerpt LUFS with estimated peak protection and verifies the exported WAV.
Publisher and support
WaveLocus is developed and operated by Sahil P Karawade, an independent developer. Visit WaveLocus support for help.