Skip to content

README.md

Latest commit

 

History

History
98 lines (70 loc) · 2.45 KB

README.md

File metadata and controls

98 lines (70 loc) · 2.45 KB

YouTube Video Summarizer

This tool creates comprehensive summaries of YouTube videos by analyzing their transcripts using the Gemini API.

Features

  • Fetches YouTube video transcripts
  • Uses MapReduce pattern to process videos of any length
  • Rate-limited API calls to stay within Gemini API quotas
  • Smart chunking with sentence-based segmentation
  • Creates well-structured summaries with key points
  • Supports videos of any length, including multi-hour content
  • Saves summaries as Markdown files

Requirements

Installation

  1. Clone this repository
  2. Install dependencies: npm install
  3. Create a .env file with your Gemini API key:
GEMINI_API_KEY=your_api_key_here

Usage

Run the tool with:

npm start

The tool will:

  1. Ask for your Gemini API key (if not in .env)
  2. Ask for the YouTube video URL
  3. Fetch and analyze the transcript
  4. Generate and save a summary

Options for Long Videos

For videos over 1 hour long, you can use different modes to balance quality and processing time:

Standard Mode (Default)

Processes up to 300K characters (~75K tokens) of transcript

npm start

Long Video Mode

Faster processing with intelligent sampling of very long transcripts

npm start -- --long-video

High Quality Mode

More comprehensive summaries by processing more content (up to 500K characters)

npm start -- --high-quality

Sampling Method Options

When using --long-video, you can specify how the transcript is sampled:

  • --sample-intelligent (default): Takes 6 strategically placed samples throughout the video
  • --sample-uniform: Takes equal samples from beginning, middle, and end
  • --sample-bookend: Focuses more on beginning and end where key info often appears

Example:

npm start -- --long-video --sample-bookend

How It Works

  1. Fetch & Validate: Gets the YouTube transcript and verifies it
  2. Chunk: Divides the transcript into manageable pieces
  3. Map: Summarizes each chunk in parallel (with rate limiting)
  4. Reduce: Combines chunk summaries into a cohesive final summary
  5. Format: Applies Markdown formatting for readability

API Rate Limiting

The tool respects Gemini API free tier limits:

  • 15 requests per minute
  • 1,000,000 tokens per minute
  • 1,500 requests per day

For long videos, processing automatically throttles to stay within these limits.

License

MIT