Bulk Media Summarizer

A tool for scanning and summarizing media content across YouTube, podcast RSS feeds, SoundCloud, and other sources supported by yt-dlp. Ideal for podcast research, conference talk digests, tutorial compilations, and topic research.

Prerequisites

Ensure these dependencies are installed:

• bun: curl -fsSL https://bun.sh/install | bash
• yt-dlp: brew install yt-dlp or pip install yt-dlp
• summarize: See https://summarize.sh (GitHub)

Supported Platforms

Apple Podcasts

Apple Podcasts URLs don't work directly with yt-dlp. Extract the RSS feed URL first:

1. Use WebFetch on the Apple Podcasts URL
2. Find the RSS feed URL in the page (usually in metadata)
3. Use the RSS feed URL in the config

Example:

• Apple URL: https://podcasts.apple.com/us/podcast/huberman-lab/id1545953110
• RSS Feed: https://feeds.megaphone.fm/hubermanlab (use this in config)

Core Commands

bun run /path/to/bulk-summarize.ts [command]

Options

Workflow

Quick Start

# 1. Create config
bun run bulk-summarize.ts init my-research.json

# 2. Edit config to add sources

# 3. Scan and process
bun run bulk-summarize.ts -c my-research.json scan
bun run bulk-summarize.ts -c my-research.json summarize
bun run bulk-summarize.ts -c my-research.json combine --output research.md

YouTube Channels

# Search for channels
yt-dlp "ytsearch10:TOPIC podcast" --flat-playlist \
  --print "%(channel_url)s %(channel)s" 2>/dev/null | sort -u

# Verify channel
yt-dlp --flat-playlist --print "%(playlist_uploader)s" \
  --playlist-end 1 "https://www.youtube.com/@ChannelName/videos"

Podcast RSS Feeds

# Test RSS feed
yt-dlp --flat-playlist --print "%(title)s" \
  "https://feeds.megaphone.fm/hubermanlab" --playlist-end 3

To get RSS feed from Apple Podcasts: use WebFetch on the Apple Podcasts URL and look for the feed URL in the response (typically in page metadata or as a direct link).

Config Structure

{
  "name": "Project Name",
  "keywords": ["keyword1", "keyword2"],
  "sources": [
    {
      "id": "source-id",
      "name": "Display Name",
      "url": "https://...",
      "type": "channel",
      "enabled": true
    }
  ],
  "settings": {
    "maxVideosPerSource": 50,
    "summaryLength": "xl",
    "summaryPrompt": "Your prompt.\n\nTitle: {title}",
    "outputDir": "summaries"
  }
}

Summary Lengths

Audio vs Video Processing

YouTube (has transcripts)

• Fast processing using existing captions
• Highest accuracy for transcribed content

Audio Podcasts (RSS, SoundCloud)

• Requires transcription via Whisper
• summarize handles this automatically
• Slightly slower due to transcription step
• Quality depends on audio clarity

Output Structure

summaries/
  source-id/
    .checkpoint.json    # Progress tracking
    episode1.md         # Individual summaries
    episode2.md

Troubleshooting

Source Not Found

# Test URL directly
yt-dlp --flat-playlist --print "%(title)s" "URL" --playlist-end 3

Transcription Fails

• Audio quality too low
• Non-English content (specify language in summarize)
• Try reducing parallel processing

Additional Resources

Reference Files

• references/config-templates.md - Ready-to-use configs for common use cases
• references/prompt-patterns.md - Effective prompts by content type
• references/platform-guide.md - Platform-specific URL formats and tips

Example Files

• examples/podcast-research.json - Multi-source podcast research
• examples/conference-talks.json - Conference playlist digest
• examples/multi-platform.json - Cross-platform research setup