julianobarbosa/notebooklm

NotebookLM Automation

Unofficial Python CLI and API for Google NotebookLM (notebooklm-py). Provides full programmatic access including capabilities the web UI doesn't expose.

Prerequisites

• Python 3.10+
• Google account with NotebookLM access
• One-time browser login via Playwright

Installation

# Install with browser login support
pip install "notebooklm-py[browser]"
playwright install chromium

# Linux only: also run
playwright install-deps chromium

Authentication

First-time setup requires browser login:

notebooklm login
# Opens Chromium → sign into Google → press Enter when done
# Session saved to ~/.notebooklm/storage_state.json

Check auth status: notebooklm auth check --test

For headless/CI environments, copy storage_state.json from a local machine or set NOTEBOOKLM_AUTH_JSON env var.

Environment Variables

| Variable | Description | Default | |----------|-------------|---------| | NOTEBOOKLM_HOME | Config directory | ~/.notebooklm | | NOTEBOOKLM_AUTH_JSON | Inline auth JSON (CI/CD) | — | | NOTEBOOKLM_LOG_LEVEL | DEBUG/INFO/WARNING/ERROR | WARNING | | NOTEBOOKLM_DEBUG_RPC | Enable RPC debug (1) | false |

Core Workflow

The typical workflow is: create notebook → add sources → generate content → download.

1. Notebook Management

notebooklm create "My Research"       # Create notebook
notebooklm list                        # List all notebooks
notebooklm use <id>                    # Set active notebook (supports partial ID)
notebooklm summary                     # AI summary of current notebook
notebooklm rename "New Title"          # Rename
notebooklm delete <id>                 # Delete

2. Adding Sources

Sources are auto-detected by type:

notebooklm source add "https://example.com/article"          # URL
notebooklm source add "https://youtube.com/watch?v=..."       # YouTube
notebooklm source add ./document.pdf                           # File (PDF, MD, DOCX, TXT, audio, video, images)
notebooklm source add-drive <drive-file-id> "Title"           # Google Drive
notebooklm source add-research "climate policy" --mode deep --import-all  # Research agent

Other source commands:

notebooklm source list                # List sources
notebooklm source fulltext <id>       # Get source full text
notebooklm source guide <id>          # AI-generated source guide
notebooklm source rename <id> "New"   # Rename
notebooklm source refresh <id>        # Re-fetch URL source
notebooklm source delete <id>         # Delete

3. Chat / Q&A

notebooklm ask "What are the key findings?" -s <source_id>
notebooklm ask "Compare sources" --json --save-as-note --note-title "Comparison"
notebooklm history                     # View chat history
notebooklm history --save              # Save history as note

4. Content Generation

All generate commands support: -s/--source (repeatable, limit to specific sources), --json, --language, --retry N.

Most are async — use --wait to block until complete.

Audio Overviews (Podcasts)

notebooklm generate audio "Focus on practical applications" \
  --format deep-dive \    # deep-dive | brief | critique | debate
  --length long \         # short | default | long
  --wait

Video Overviews

notebooklm generate video "Explain the architecture" \
  --format explainer \    # explainer | brief
  --style whiteboard \    # auto | classic | whiteboard | kawaii | anime | watercolor | retro-print | heritage | paper-craft
  --wait

Slide Decks

notebooklm generate slide-deck "Executive summary" \
  --format detailed \     # detailed | presenter
  --length default \      # default | short
  --wait

# Revise a specific slide
notebooklm generate revise-slide "Add more data points" \
  -a <artifact_id> --slide 2 --wait    # slide is zero-based

Study Materials

# Quizzes
notebooklm generate quiz --difficulty hard --quantity more --wait

# Flashcards
notebooklm generate flashcards --difficulty medium --wait

Visual & Data

# Infographic
notebooklm generate infographic \
  --orientation landscape \   # landscape | portrait | square
  --detail detailed \         # concise | standard | detailed
  --wait

# Mind map (synchronous, no --wait needed)
notebooklm generate mind-map

# Data table
notebooklm generate data-table "Compare metrics across studies" --wait

Reports

notebooklm generate report "Security analysis" \
  --format briefing-doc \     # briefing-doc | study-guide | blog-post | custom
  --append "Include threat modeling" \
  --wait

5. Downloading Content

All download commands support: -a/--artifact, --all, --latest, --earliest, --name, --force, --no-clobber, --dry-run, --json.

notebooklm download audio ./podcast.mp3
notebooklm download video ./overview.mp4
notebooklm download slide-deck ./slides.pptx --format pptx   # or pdf (default)
notebooklm download infographic ./info.png
notebooklm download report ./report.md
notebooklm download mind-map ./map.json
notebooklm download data-table ./data.csv
notebooklm download quiz --format json ./quiz.json       # json | markdown | html
notebooklm download flashcards --format markdown ./cards.md

6. Sharing

notebooklm share status
notebooklm share public --enable           # Create public link
notebooklm share view-level full           # full | chat
notebooklm share add [email protected] --permission editor -m "Check this out"
notebooklm share remove [email protected]

7. Language

notebooklm language list                   # 80+ languages
notebooklm language get
notebooklm language set ja                 # Set to Japanese

Python API

Fully async API for programmatic workflows:

import asyncio
from notebooklm import NotebookLMClient

async def main():
    async with await NotebookLMClient.from_storage() as client:
        # Create notebook and add sources
        nb = await client.notebooks.create("Research")
        await client.sources.add_url(nb.id, "https://example.com")

        # Generate audio overview
        artifact = await client.artifacts.generate_audio(
            nb.id, description="Deep dive on findings",
            format=AudioFormat.DEEP_DIVE, length=AudioLength.LONG
        )

        # Wait and download
        await client.artifacts.wait(nb.id, artifact.id)
        await client.artifacts.download_audio(nb.id, artifact.id, "output.mp3")

        # Chat with sources
        result = await client.chat.ask(nb.id, "Summarize key points")
        print(result.answer)

asyncio.run(main())

API modules: client.notebooks, client.sources, client.artifacts, client.chat, client.research, client.notes, client.settings, client.sharing

Common Recipes

Research-to-Podcast Pipeline

notebooklm create "Climate Research"
notebooklm use <id>
notebooklm source add "https://en.wikipedia.org/wiki/Climate_change"
notebooklm source add-research "climate change solutions 2025" --mode deep --import-all
notebooklm generate audio "Focus on actionable solutions" --format debate --length long --wait
notebooklm download audio ./climate-debate.mp3

Document Analysis to Study Materials

notebooklm create "Exam Prep"
notebooklm use <id>
notebooklm source add ./textbook.pdf
notebooklm generate quiz --difficulty hard --quantity more --wait
notebooklm generate flashcards --wait
notebooklm download quiz --format markdown ./quiz.md
notebooklm download flashcards --format json ./cards.json

Batch Import + Full Report

notebooklm create "Literature Review"
notebooklm use <id>
for f in ./papers/*.pdf; do notebooklm source add "$f"; done
notebooklm generate report "Systematic review" --format briefing-doc --wait
notebooklm download report ./review.md

Troubleshooting

| Issue | Fix | |-------|-----| | Auth expired | Run notebooklm login again | | playwright not found | pip install "notebooklm-py[browser]" then playwright install chromium | | Generation stuck | Use notebooklm source wait <id> for pending sources, check --retry flag | | Partial ID not matching | Use more characters of the notebook ID | | Debug API calls | Set NOTEBOOKLM_LOG_LEVEL=DEBUG or NOTEBOOKLM_DEBUG_RPC=1 |

---

Gotchas

• Source upload >200MB silently fails after the request returns 200 — downstream shows "processing" forever. Pre-check file size before upload.
• Concurrent audio/video generation per project is rate-limited — second concurrent gen fails with a generic "try again later". Serialize generation jobs.
• gcloud auth vs notebooklm-account differ — a user logged into gcloud may not have notebooklm access; pass --account explicitly when they diverge.
• Source ordering at upload affects citation precedence in generated content — re-uploading to fix order changes the output style.
• Studio types have different timeout windows — audio ~5min, video ~15min, slides ~3min. CLI default of 60s drops mid-generation for the longer types.