bulk-summarize

bulk-summarize

Bulk YouTube video summarizer for research. Scans channels and playlists for videos matching keywords, then uses AI to create detailed summaries.

Perfect for:

• Podcast research and episode digests
• Conference talk compilations
• Tutorial series notes
• Any YouTube-based research project

Features

• Keyword filtering - Only summarize videos matching your search terms
• Per-source checkpoints - Each source tracks its own progress in its folder
• Organized output - Summaries grouped by source in separate directories
• Configurable prompts - Customize what the AI extracts from each video
• Combined output - Merge all summaries into a single searchable document

Installation

Prerequisites

1. Bun - JavaScript runtime

   curl -fsSL https://bun.sh/install | bash
   

2. yt-dlp - YouTube metadata downloader

   # macOS
   brew install yt-dlp
   
   # or pip
   pip install yt-dlp
   

3. summarize - AI summarization CLI (GitHub)

   # See https://summarize.sh for installation
   # Requires API key (OpenAI, Anthropic, etc.)
   

Install bulk-summarize

Option 1: Download binary (easiest)

Download the latest release for your platform from Releases:

# macOS (Apple Silicon)
curl -L https://github.com/smerchek/bulk-summarize/releases/latest/download/bulk-summarize-darwin-arm64 -o /usr/local/bin/bulk-summarize
chmod +x /usr/local/bin/bulk-summarize

# macOS (Intel)
curl -L https://github.com/smerchek/bulk-summarize/releases/latest/download/bulk-summarize-darwin-x64 -o /usr/local/bin/bulk-summarize
chmod +x /usr/local/bin/bulk-summarize

# Linux
curl -L https://github.com/smerchek/bulk-summarize/releases/latest/download/bulk-summarize-linux-x64 -o /usr/local/bin/bulk-summarize
chmod +x /usr/local/bin/bulk-summarize

Option 2: From source (requires Bun)

git clone https://github.com/smerchek/bulk-summarize.git
cd bulk-summarize
bun install
bun link --global

Verify installation:

bulk-summarize --help

Quick Start

# 1. Initialize project
bulk-summarize init

# 2. Edit bulk-summarize.json (add sources, keywords)

# 3. Scan for matching videos
bulk-summarize scan

# 4. Summarize pending videos
bulk-summarize summarize

# 5. Combine into one document
bulk-summarize combine

Usage

bulk-summarize [options] <command> [args]

Commands:
  init [name]        Create starter config file
  scan               Scan sources for videos matching keywords
  summarize          Summarize pending videos
  combine            Combine all summaries into one document
  status             Show progress for all sources
  list               List configured sources
  reset [source]     Reset checkpoint (all or specific source)
  help               Show help

Options:
  -c, --config <file>      Config file (default: bulk-summarize.json)
  -o, --output-dir <dir>   Output directory (overrides config)
  -s, --source <id>        Target specific source
  -n, --limit <n>          Limit videos to process
  -d, --delay <ms>         Delay between videos (default: 1000ms)
  -p, --parallel <n>       Concurrent summarizations (default: 1)
  --output <file>          Output file for combine

Output Structure

Each source gets its own folder with checkpoint and summaries:

summaries/
  podcast-name/
    .checkpoint.json     # Tracks pending/done/errors
    abc123.md            # Video summaries
    def456.md
  another-channel/
    .checkpoint.json
    ghi789.md

Configuration

{
  "name": "My Research",
  "keywords": ["topic1", "topic2"],
  "sources": [
    {
      "id": "channel-id",
      "name": "Channel Name",
      "url": "https://www.youtube.com/@ChannelHandle",
      "enabled": true,
      "keywords": ["override", "keywords"]
    }
  ],
  "settings": {
    "maxVideosPerSource": 50,
    "summaryLength": "xl",
    "summaryPrompt": "Extract key insights. Ignore ads.\n\nTitle: {title}",
    "outputDir": "summaries",
    "model": "cli/claude/haiku"
  }
}

Settings

Field	Description
keywords	Default keywords. Videos matching ANY keyword included. Empty = all videos
summaryLength	short, medium, long, xl, xxl
summaryPrompt	AI instructions. {title} and {source} are replaced
maxVideosPerSource	Limit per source
model	Override summarize.sh model (e.g. cli/claude/haiku)

Source Options