mk-youtube-get-audio

name	mk-youtube-get-audio
description	Download YouTube video audio file. Use when user wants to extract audio or download music/podcast from a video.
license	MIT
metadata	{"version":"1.1.2","author":"kouko","tags":["youtube","audio","download"]}
compatibility	{"claude-code":">=1.0.0"}

YouTube Audio Download

Download video audio file (best available format, no conversion).

Quick Start

/mk-youtube-get-audio <URL> [output_dir] [browser] [--force]

Parameters

Parameter	Required	Default	Description
URL	Yes	-	YouTube video URL
output_dir	No	/tmp/monkey_knowledge/youtube/audio	Output directory for audio file
browser	No	auto	Browser for cookies (chrome, firefox, safari, edge, brave)
--force	No	false	Force re-download even if cached file exists

Examples

• /mk-youtube-get-audio https://youtube.com/watch?v=xxx - Download with auto cookie fallback
• /mk-youtube-get-audio https://youtube.com/watch?v=xxx ~/Music - Save to custom directory
• /mk-youtube-get-audio https://youtube.com/watch?v=xxx /tmp chrome - Use Chrome cookies

How it Works

1. Execute: {baseDir}/scripts/audio.sh "<URL>" "<output_dir>" "<browser>"
2. First attempt: download without authentication
3. If failed: retry with browser cookies (auto-detect or specified)
4. Parse JSON output to get file path

┌─────────────────────────────┐
│   First attempt (no auth)   │
└──────────────┬──────────────┘
               │
       ┌───────┴───────┐
       │               │
    Success         Failed
       │               │
       ▼               ▼
   [Return]    ┌─────────────────────┐
               │ Retry with cookies  │
               │ chrome → firefox →  │
               │ safari → edge →     │
               │ brave               │
               └──────────┬──────────┘
                          │
                  ┌───────┴───────┐
                  │               │
               Success         Failed
                  │               │
                  ▼               ▼
              [Return]        [Error]

Output Format

Success:

{
  "status": "success",
  "file_path": "/tmp/monkey_knowledge/youtube/audio/20091025__VIDEO_ID.m4a",
  "file_size": "5.2M",
  "cached": false,
  "video_id": "dQw4w9WgXcQ",
  "title": "Video Title",
  "channel": "Channel Name",
  "url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
  "duration_string": "3:32"
}

Cache hit (returns existing file):

{
  "status": "success",
  "file_path": "/tmp/monkey_knowledge/youtube/audio/20091025__VIDEO_ID.m4a",
  "file_size": "5.2M",
  "cached": true,
  ...
}

Error:

{
  "status": "error",
  "message": "Download failed (tried with and without cookies)"
}

Filename Format

Files use unified naming with date prefix: {YYYYMMDD}__{video_id}.{ext}

Example: 20091025__dQw4w9WgXcQ.m4a

Browser Cookie Fallback

When download fails (e.g., member-only or age-restricted content), the script automatically:

1. Tries each browser: chrome → firefox → safari → edge → brave
2. For Chrome: tries all profiles (Default, Profile 1, Profile 2, ...)
3. Uses first successful browser/profile combination

Supported browsers:

Browser	Parameter	Chrome Profile Support
Chrome	chrome	Yes (auto-detect all profiles)
Firefox	firefox	Default profile only
Safari	safari	Default profile only
Edge	edge	Default profile only
Brave	brave	Default profile only

Use Cases

• Download audio for speech-to-text when video has no subtitles
• Podcast or music download
• Member-only or age-restricted content (with browser cookies)

Notes

• File caching: If audio file already exists for this video, it will be reused (returns cached: true)
• Force refresh: Use --force flag to re-download even if cached file exists
• Uses system yt-dlp/jq if available, otherwise auto-downloads on first run
• No ffmpeg required (uses best available format without conversion)
• Output format depends on source (typically m4a, webm, or opus)
• Cookie fallback only activates when initial download fails
• Using cookies may risk YouTube account suspension - use secondary account if needed

Next Step

After downloading the audio, invoke /mk-youtube-audio-transcribe with the file_path from the output:

/mk-youtube-audio-transcribe <file_path> [model] [language]

Tip: If you know the video's language from /mk-youtube-get-info, pass it as the language parameter for better model auto-selection (e.g., zh → belle-zh, ja → kotoba-ja).