Video Transcriber — Claude Code MCP server for video transcripts
Turns videos from links or local files into text with whisper.cpp, running on your own machine. Handy when a coding agent needs the spoken words, not the picture.
Open on GitHubWhat Video Transcriber does
Point the server at a URL or a file path and it hands back a transcript. yt-dlp fetches the media, ffmpeg prepares the audio, and whisper.cpp does the recognition; the whole thing is written in Rust. Five model sizes are available, from tiny up to large, and results come out as TXT, JSON or Markdown.
Two transports. Over stdio it runs locally as video-transcriber-mcp, which is the default. Over HTTP it listens on port 8080 unless you say otherwise, and clients connect to http://localhost:8080/mcp. From there it behaves like any other MCP server your agent already talks to.
yt-dlp and ffmpeg have to be present: without them a link download or the audio step fails. The Homebrew formula pulls both in with the binary. If you install through cargo install or unpack a release archive, you own that setup yourself.
What it helps with
- pull a transcript out of a YouTube link inside the chat
- transcribe a local mp4 without uploading it anywhere
- get JSON output that a script can consume
- handle speech in one of many supported languages
- run one shared transcription server for a team over HTTP
Example prompts for the agent
- Transcribe the video at this link and save the text to
notes.md - Give me a JSON transcript of
lecture.mp4 - Start the transcription server over HTTP on port 3000
How to install Video Transcriber
Install
The MCP registry has no launch command for it: see the repository README
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Video Transcriber
How do I install Video Transcriber?
The Homebrew route covers the binary and its dependencies in one go: brew install nhatvu148/tap/video-transcriber-mcp. With cargo install video-transcriber-mcp you get just the binary and must add yt-dlp, ffmpeg and cmake yourself.
What does Video Transcriber do?
It converts video into text, either from a link or from a file on disk. Recognition happens locally through whisper.cpp, so the media stays on your machine.
Can other machines reach the server?
Yes, run it with --transport http. Clients then use http://localhost:8080/mcp, and --host or --port change where it listens.
