NotebookLM Automation

Automates the creation of multiple audio overviews in Google NotebookLM with different prompts and configurations.

Overview

This skill uses Playwright to automate NotebookLM's browser interface for:

1. Creating a new notebook
2. Adding website sources
3. Generating multiple audio overviews with custom prompts
4. Downloading all generated audio files

The automation handles the repetitive UI interactions, waiting for generation to complete, and downloading all outputs.

Workflow

The automation follows these steps:

1. Prepare input - Create JSON file with sources and audio overview configurations
2. Launch automation - Run the Playwright script
3. Notebook creation - Script creates new notebook and adds sources
4. Audio generation - Script creates each audio overview with specified prompt and length
5. Wait for completion - Script waits ~10 minutes for all audio to generate
6. Download outputs - Script downloads all audio files to /home/claude/

Input Format

Create a JSON file with this structure:

{
  "sources": [
    "https://example.com/article1",
    "https://example.com/page2"
  ],
  "audio_overviews": [
    {
      "prompt": "Focus on technical concepts for developers",
      "length": "longer"
    },
    {
      "prompt": "Create executive summary of business impacts",
      "length": "default"
    }
  ]
}

Field Specifications

sources: Array of website URLs to use as source material

• Must be publicly accessible URLs
• NotebookLM will crawl and index these pages

audio_overviews: Array of audio overview configurations

• prompt: Text describing what the AI should focus on in this episode
• length: Either "default" or "longer" (never use "shorter")

See references/example_input.json for a complete example.

Usage

Basic Usage

python scripts/generate_audio_overviews.py input.json

This runs with visible browser so you can monitor progress and handle Google authentication if needed.

Headless Mode

python scripts/generate_audio_overviews.py input.json --headless

Run browser in headless mode (requires existing authentication).

Custom Timeout

python scripts/generate_audio_overviews.py input.json --timeout 900

Set custom timeout for audio generation (in seconds). Default is 600 seconds (10 minutes).

Prerequisites

Required Software

Install Playwright and dependencies:

pip install playwright
playwright install chromium

Google Account Setup

• Must have access to NotebookLM (https://notebooklm.google.com)
• First run will require Google authentication
• Browser state is not persisted between runs

Output

Audio files are downloaded to /home/claude/ with NotebookLM's default naming:

• Format: MP3
• Naming: Typically includes date/timestamp

The script reports download success for each file and provides summary at completion.

Troubleshooting

Common Issues

Authentication Required

• Run without --headless flag
• Sign in to Google when prompted
• Consider using persistent browser context for repeated use

Selectors Not Found

• NotebookLM's UI may have changed
• See references/troubleshooting.md for selector update guidance
• Run without --headless to observe actual UI

Timeout During Generation

• Audio generation takes ~5-10 minutes per overview
• Increase timeout with --timeout flag
• Script will attempt downloads even after timeout

Sources Not Loading

• Verify URLs are accessible and public
• Check for NotebookLM error messages in browser
• Some content may be blocked or require authentication

Getting Help

For selector updates and detailed troubleshooting, read references/troubleshooting.md.

Limitations

• Requires manual Google authentication on first run
• Browser automation may be fragile if NotebookLM's UI changes significantly
• Cannot create more than ~10 audio overviews at once (NotebookLM limitation)
• Sources must be publicly accessible websites
• Generation time increases with number of sources and overview length