# ClipQuill

> Free, private video and audio to text. Runs in your browser — the file is processed on your own
> device and zero bytes are uploaded. No signup, no account, no daily cap.

ClipQuill transcribes a video or audio file you already have. The recognition model is downloaded
once (about 44 MB compressed / about 66 MB decompressed) and everything after that happens locally.

## Limits and formats (as published 2026-09-17)

- Input: MP4, MOV, WEBM, M4A, MP3, WAV, AAC, OGG. Anything else is untested and not promised.
- Output: TXT, SRT, VTT, plus one-click copy.
- Limits: 60 minutes of audio and 512 MB per file. A file past either is refused up front.
- Languages: the model carries 99 language tokens, but you must pick the spoken language — this
  build cannot detect it. English and one short Chinese clip are the only ones actually run.
- No fetching from a link. Save the file to your device first.

## Machine-readable resources

- Agent docs: https://clipquill.com/ai/
- Public facts (whole dataset, 16 entries): https://clipquill.com/api/agent/facts
- Fact search: https://clipquill.com/api/agent/lookup?q=<question or topic id>
- OpenAPI 3.1: https://clipquill.com/openapi.json
- MCP server (Streamable HTTP, anonymous): https://clipquill.com/mcp
- MCP server card: https://clipquill.com/.well-known/mcp/server-card.json
- API catalog (RFC 9727): https://clipquill.com/.well-known/api-catalog
- Agent skill: https://clipquill.com/ai/skills/site-lookup/SKILL.md
- Skills discovery index: https://clipquill.com/.well-known/agent-skills/index.json
- ARD manifest: https://clipquill.com/.well-known/ai-catalog.json
- Service index (DNS-AID): https://clipquill.com/ai/index.ilang
- Auth status: https://clipquill.com/auth.md — under construction, no authentication exists or is required
- Homepage as Markdown: request https://clipquill.com/ with `Accept: text/markdown`

## Optional

- Full homepage text: https://clipquill.com/llms-full.txt (same content as https://clipquill.com/ with `Accept: text/markdown`)