Yueying — Claude Code MCP server: video analysis and setup
Yueying lets an AI client inspect local videos and supported URLs, returning timestamped transcripts, keyframes, and contact sheets. It fits lecture and demo analysis.
claude mcp add --env PYTHONUTF8=<PYTHONUTF8> --env YUEYING_OUT_DIR=<YUEYING_OUT_DIR> --env YUEYING_MODEL=<YUEYING_MODEL> --env YUEYING_DEVICE=<YUEYING_DEVICE> --transport stdio yueying-let-ai-watch-videos -- uvx yueyingWhat Yueying – Let AI watch videos does
Yueying exposes video analysis through MCP. Give it a local file or a YouTube, Bilibili, or other supported video URL, and it produces a timestamped transcript, subtitle files, keyframes, and contact sheets. Existing captions are preferred; when they are unavailable, speech recognition runs locally with faster-whisper.
Each result is written to its own folder with report.md, transcript.txt, transcript.srt, image grids, full-size frames, and manifest.json. You need uv before connecting it because the setup commands run uvx. No API key is needed. ffmpeg comes through imageio-ffmpeg, and downloaded URL videos are removed after processing by default.
What it helps with
- create a timestamped transcript from a video
- pick representative frames from a recording
- build contact sheets for scene review
- export subtitles as
transcript.srt
Example prompts for the agent
- Watch
C:\videos\lecture3.mp4and turn the main steps into notes. - Review
https://www.bilibili.com/video/BV…and find what it says about docker compose. - Show the keyframes and connect them to the relevant transcript passages.
How to install Yueying – Let AI watch videos
Platform CLI
claude mcp add --env PYTHONUTF8=<PYTHONUTF8> --env YUEYING_OUT_DIR=<YUEYING_OUT_DIR> --env YUEYING_MODEL=<YUEYING_MODEL> --env YUEYING_DEVICE=<YUEYING_DEVICE> --transport stdio yueying-let-ai-watch-videos -- uvx yueyingReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from yueying <command>
Platform CLI
codex mcp add yueying-let-ai-watch-videos --env PYTHONUTF8=<PYTHONUTF8> --env YUEYING_OUT_DIR=<YUEYING_OUT_DIR> --env YUEYING_MODEL=<YUEYING_MODEL> --env YUEYING_DEVICE=<YUEYING_DEVICE> -- uvx yueyingReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from yueying <command>
Config file
{
"mcpServers": {
"yueying-let-ai-watch-videos": {
"command": "uvx",
"args": [
"yueying"
],
"env": {
"PYTHONUTF8": "<PYTHONUTF8>",
"YUEYING_OUT_DIR": "<YUEYING_OUT_DIR>",
"YUEYING_MODEL": "<YUEYING_MODEL>",
"YUEYING_DEVICE": "<YUEYING_DEVICE>"
}
}
}
}Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from yueying <command>
Config file
{
"servers": {
"yueying-let-ai-watch-videos": {
"type": "stdio",
"command": "uvx",
"args": [
"yueying"
],
"env": {
"PYTHONUTF8": "<PYTHONUTF8>",
"YUEYING_OUT_DIR": "<YUEYING_OUT_DIR>",
"YUEYING_MODEL": "<YUEYING_MODEL>",
"YUEYING_DEVICE": "<YUEYING_DEVICE>"
}
}
}
}Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from yueying <command>
Platform CLI
gemini mcp add -e PYTHONUTF8=<PYTHONUTF8> -e YUEYING_OUT_DIR=<YUEYING_OUT_DIR> -e YUEYING_MODEL=<YUEYING_MODEL> -e YUEYING_DEVICE=<YUEYING_DEVICE> yueying-let-ai-watch-videos uvx yueyingReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from yueying <command>
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 Yueying – Let AI watch videos
How do I install Yueying in Claude Code?
Run claude mcp add yueying -- uvx yueying.
What does Yueying do?
It gives an MCP client a way to inspect local videos and video URLs, returning timestamped speech text, selected frames, and contact sheets for visual context.
Does Yueying require an API key?
No. It uses platform captions when available and otherwise runs local faster-whisper recognition. The speech model is downloaded when needed.
