YouTube Summariser

A command-line tool that summarizes YouTube videos using AI. It extracts transcripts from YouTube videos and generates structured summaries using OpenAI, Anthropic, or OpenRouter (300+ models).

Installation

pip install youtube-summariser

Note: youtube-summariser (British), youtube-summarizer (American), and ys (short alias) commands are all available.

Or install from source:

git clone https://github.com/weijianzhg/youtube-summariser
cd youtube-summariser
pip install -e .

Quick Start

Run the interactive setup to configure your API keys:

youtube-summariser init

This guides you through:

• Selecting your default provider (Anthropic, OpenAI, or OpenRouter)
• Entering your API key (securely masked)
• Optionally configuring additional providers

Configuration

Option 1: Interactive Setup (Recommended)

youtube-summariser init

Settings are saved to a platform-appropriate location:

• macOS/Linux: ~/.youtube-summariser/config.yaml
• Windows: %APPDATA%\youtube-summariser\config.yaml

Re-run init anytime to update your settings.

Option 2: Environment Variables

# For Anthropic (default provider)
export ANTHROPIC_API_KEY=your_anthropic_api_key

# For OpenAI
export OPENAI_API_KEY=your_openai_api_key

# For OpenRouter (access 300+ models)
export OPENROUTER_API_KEY=your_openrouter_api_key

Or create a .env file in your working directory.

Configuration Priority

1. Environment variables (highest priority)
2. User config file (~/.youtube-summariser/config.yaml)
3. Bundled defaults

Default Provider

The default provider is Anthropic. You can change this via init or override per command using --provider.

Usage

# Interactive configuration
youtube-summariser init

# Same command with the short alias
ys init

# Summarize a video (saves to auto-generated filename)
youtube-summariser "https://www.youtube.com/watch?v=VIDEO_ID"

# Specify output filename
youtube-summariser "https://youtu.be/VIDEO_ID" -o my_summary.md

# Print to terminal only (no file saved)
youtube-summariser "https://youtube.com/watch?v=VIDEO_ID" --no-save

# Use a specific provider
youtube-summariser "https://youtu.be/VIDEO_ID" --provider openai

# Use OpenRouter with access to 300+ models
youtube-summariser "https://youtu.be/VIDEO_ID" --provider openrouter

Search by Title

Don't have a URL? Search for videos by title:

# Interactive selection (shows top 5 results)
youtube-summariser search "How to make mass"

# Auto-select first result
youtube-summariser search "Python tutorial" --first

# Show more results
youtube-summariser search "cooking recipes" --max-results 10

Summarize a Channel

Summarize every video published on a channel:

youtube-summariser channel "https://www.youtube.com/@CHANNEL"

For large channels, limit the run to the most recent videos:

youtube-summariser channel "https://www.youtube.com/@CHANNEL" --max-videos 10

Each video goes through the same transcript and LLM summarization pipeline as the single-video command. Summaries are saved individually under channel-summaries/ by default. Use --output-dir to choose another directory.

Note: Without --max-videos, the command processes every video on the channel. This can take a long time and make many LLM API calls on a large channel.

Commands

You can also pass a URL directly without the summarise subcommand for convenience.

Options

zsh Tip (URLs Without Quotes)

In zsh, unquoted YouTube URLs containing ? are treated as glob patterns before the CLI runs.

Use one of these approaches:

# Escape ? in the URL
ys https://www.youtube.com/watch\?v=VIDEO_ID

# Or disable globbing for this command
noglob ys https://www.youtube.com/watch?v=VIDEO_ID

To make this convenient permanently, add to ~/.zshrc:

ys() { noglob command ys "$@"; }

Output Format

Summary files are saved as Obsidian-friendly markdown (.md) with YAML frontmatter:

---
title: "Video Title"
url: "https://www.youtube.com/watch?v=VIDEO_ID"
video_id: VIDEO_ID
channel: "Channel Name"
published_at: 2026-07-20
created: 2026-07-28
content_type: tutorial
series: "Example Course"
series_index: 2
concepts:
  - "gradient descent"
  - "backpropagation"
prerequisites:
  - "basic calculus"
tags:
  - youtube
  - summary
  - "machine-learning"
  - "neural-networks"
model: "anthropic/claude-sonnet-4-5-20250929"
---

# Video Title

## TL;DR
...

## Key Takeaways
- Insight with a linked timestamp [12:34](https://www.youtube.com/watch?v=VIDEO_ID&t=754)

## Detailed Summary
...

## Notable Quotes
...

Generated notes include normalized tags and structured properties for Obsidian.

Requirements

• Python 3.10+
• An API key for OpenAI, Anthropic, or OpenRouter

License

MIT License - see LICENSE for details.