MeowMusic YouTube MP3

Package and reuse the MeowMusicServer-patched YouTube fallback workflow: Windows Chrome cookie export/sync to server, server-side yt-dlp/yt-dlp-ejs/ffmpeg setup, old-source-first with YouTube fallback, and MV-to-MP3 extraction/caching. Use when Claude needs to wire YouTube audio acquisition into MeowMusicServer or a similar music service, debug YouTube download failures, refresh cookies from a Windows Chrome profile, or implement a local-cache MP3 flow from YouTube videos.

Use this skill to rebuild the currently working MeowMusicServer-patched solution into another repo/server without rediscovering the whole stack.

Workflow decision tree

1. Need cookie sync from Windows Chrome to server?

   • Use scripts/youtube_cookie_sync.py.
   • If the user wants a double-click workflow on Windows, also use scripts/windows/sync_cookie.bat and scripts/windows/open_youtube.bat.
   • If you need API shape or server handler details, read references/cookie-api-and-sync.md.

2. Need to prepare a Linux server for YouTube download and MP3 extraction?

   • Run scripts/install_server_env.sh.
   • If challenge solving or runtime behavior is flaky, read references/server-runtime-notes.md.

3. Need to patch MeowMusicServer so old sources stay first and YouTube acts only as fallback/补源?

   • Read references/meowmusic-integration.md.
   • Keep source priority as: sources.json → local library → cache → legacy upstreams → YouTube fallback.
   • Do not make YouTube the default first source unless the user explicitly asks.

4. Need MV -> MP3 instead of direct streaming?

   • Prefer yt-dlp to fetch bestaudio, then use ffmpeg to normalize/transcode into a stable cached music.mp3.
   • Read references/meowmusic-integration.md for the patch pattern and command shape.

Core operating rules

• Prefer old/legacy music sources first. YouTube is a补源/fallback path, not the primary path.
• Prefer local reusable cache/library over repeated redownloads.
• Prefer returning a stable cached music.mp3 path to devices instead of a fragile live stream.
• If a cookie file exists, pass it to yt-dlp automatically.
• When YouTube requires extra handling, prefer extractor args compatible with the known-good setup:
  • youtube:player_client=tv,web;formats=missing_pot
  • youtube:player_skip=webpage,configs
• Keep all cookie and bearer-token material out of the skill package. Use placeholders only.

Recommended implementation shape

1. Windows cookie sync

On Windows, export cookies from Chrome with:

yt-dlp --cookies-from-browser chrome:Default --cookies youtube-cookies.txt --skip-download https://www.youtube.com/watch?v=dQw4w9WgXcQ

Then POST the cookie file content to a server endpoint such as:

• POST /api/admin/youtube-cookie/update
• optional status check: GET /api/admin/youtube-cookie/status

Use scripts/youtube_cookie_sync.py unless the repo already has its own equivalent.

2. Server download strategy

Use this shape:

1. Search top YouTube MV/result for a song.
2. Download audio with yt-dlp into a per-track cache directory.
3. Convert the downloaded source into music.mp3 with ffmpeg.
4. Return cached local URLs from the service.

3. MeowMusic source order

Use this order unless the user asks otherwise:

1. curated sources.json
2. local uploaded/downloaded library
3. existing cache hits
4. legacy remote music APIs
5. YouTube fallback

This keeps the product aligned with the current direction: local-first, old-source-first, YouTube only for missing tracks.

Files in this skill

• scripts/install_server_env.sh — server bootstrap for Node 22, yt-dlp, yt-dlp-ejs, ffmpeg.
• scripts/youtube_cookie_sync.py — export Chrome/Edge/Firefox cookies and push them to the server.
• scripts/windows/sync_cookie.bat — Windows double-click wrapper.
• scripts/windows/open_youtube.bat — helper to open YouTube in Chrome.
• references/cookie-api-and-sync.md — cookie API contract and usage notes.
• references/server-runtime-notes.md — server environment and challenge-solver notes.
• references/meowmusic-integration.md — patch strategy and concrete Go snippets.

Execution notes

• Read only the reference file relevant to the current task.
• When editing an existing repo, prefer minimal surgical patches over broad rewrites.
• When a download path already works, preserve behavior and only harden the flaky parts.
• If the target repo differs from MeowMusicServer, reuse the flow, not the exact filenames.