Do you want to create a video lesson?

NotebookLM Learning Video/Audio Workflow

Generate a high-quality NotebookLM Audio or Video Overview that teaches the user about topics discussed in the current session.

I. When to Use

When the user wants NotebookLM to generate a podcast-style (audio) or video lesson covering concepts from a completed analysis.

II. How NotebookLM Works (Background for the AI)

NotebookLM takes TWO separate inputs:

1. Source documents — uploaded content that the AI hosts will discuss. This must be pure informational content, NOT prompts or instructions.
2. Customize prompt — a short steering directive that tells the hosts what to focus on, what expertise level to use, and what tone to take.

III. Step 1: Identify the Mode

Ask the user if they are analyzing a Legal Dispute or a General Topic.

IF GENERAL TOPIC:

1. Review the session's work and trace the reasoning chain: what was applied, the conclusion, and why.
2. Build a topic list in logical teaching order.
3. Write notebooklm_source.md with:
   • Executive Summary
   • Context/Background
   • Topic Sections (What it is, how it works, key authorities, application to case, traps/nuances)
   • Summary/Bottom Line

IF LEGAL DISPUTE:

1. Ask the user which disputes to explain.
2. Ask if they want "Teach the Law" mode (general legal education using the best examples) or "Teach the Case" mode (through the lens of this specific dispute's facts).
3. Extract and organize the laws.
4. Write notebooklm_source_[disputename].md using the Law Professor Method:
   • Teach the Law: State the Rule, Break Down the Mechanical Test, Teach with Fact Spectrums (easy yes, easy no, hard middle), Connect the Dots.
   • Teach the Case: The Accusation, The Actual Law/Defense, Fact Spectrum, Where This Case Lands (Predicted Outcome), Connect the Dots.

IV. Step 2: Formatting Rules for Source Document

• CRITICAL ANTI-GARBAGE RULE: NEVER upload raw Python code, massive configuration files, or dense technical markdown rulebooks as the source. NotebookLM is generating a conversational podcast. You MUST synthesize the raw files into a clean, podcast-friendly essay written in plain English for an audience of professionals before uploading.
• Use clear H2 headings for each topic.
• Write in complete, flowing paragraphs — not bullet lists.
• Avoid markdown formatting that won't survive clipboard paste.
• Feed the answers directly. Do NOT leave conclusions for NotebookLM to derive.

V. Step 3: Write the Customize Prompt

Create notebooklm_customize.md (or notebooklm_customize_[disputename].md).

• 2-4 sentences max.
• Set expertise level (e.g., "no legal background").
• Set teaching goal (e.g., "Walk through each issue step by step").
• Tell them to spend extra time on complex topics.

VI. Step 4: Automating NotebookLM (visible_browser MCP)

If you are automating the generation using the visible_browser MCP tools, you MUST first read notebooklm_google_com_browsing skill for the full step-by-step UI guide, known gotchas, and a log of past mistakes. Then follow these strict rules: 0. PRE-FLIGHT STATE CHECK (MANDATORY): If a NotebookLM tab is already open, you MUST take a screenshot FIRST and check: Is a video/audio already generating? Is one already finished and ready to download? If generation is already in progress or complete, DO NOT start a new one. Report what you see and switch to download/extraction mode.

1. Data Entry: Read the source text from the local file and input it into the browser using your MCP tools.
2. Read the screen, don't guess: Video Overview and Audio Overview have completely different UIs with different options. Do NOT assume what buttons exist. Open the correct modal (Video or Audio), take a screenshot, read the actual options on screen, and pick whichever format/style/length makes sense for the context. Never click a button name you haven't visually confirmed exists in that modal.
3. Rename Notebook: Before generating, click the notebook title at the top-left and rename it to a descriptive title (e.g., "Analysis: [Topic Name]").
4. Video vs. Audio — they are NOT interchangeable: If the user wants a video, stay in Video Overview. If they want audio, stay in Audio Overview. Never wander into the other modal looking for missing options — if a button doesn't exist in the current modal, it's because it belongs to the other one.
5. MANDATORY VERIFICATION: Take a screenshot and visually verify your selections BEFORE clicking Generate. Do not click Generate without confirming the right options are checked.
6. Don't click things you aren't asked to click: Only interact with the specific modal you need (Video or Audio). Do not explore other Studio features (flashcards, quizzes, etc.).
7. BATCH GENERATION TRICK: To save time when generating multiple videos, open a new browser tab for each Notebook. Paste the source and prompt, start the generation, and then immediately open a new tab to start the next one. Let them all generate concurrently in the background.
8. BLOB URL CACHE EXTRACTION (CRITICAL): When you click "Download" on a NotebookLM video, the browser may NOT save it to the Downloads folder because NotebookLM generates a Blob URL. Instead, the video gets cached as a raw file without an extension in the profile cache directory.
   • Because these files lack names, you CANNOT download multiple videos at once, or you won't know which is which.
   • You must download them one by one.
   • For each video: First clear the Cache_Data folder (check the notebooklm skill for the exact path). Then click Download. Wait 20 seconds. Then retrieve the new >10MB file from the cache, rename it immediately with the proper .mp4 name, and move it to the destination. Only then can you move to the next tab to download the next video.

VII. Step 5: Manual Guidance (Fallback)

If the user is doing it manually, give these instructions:

• From the NotebookLM dashboard, click "Create new notebook".
• Rename the notebook: By default, it will be called "Untitled notebook" or named after the source. Click the notebook title at the top-left of the screen and type a clear, readable name (e.g., "Fact Check Workflow").
• Click "Copied text" or another source type.
• Insert the source text and wait for it to load.
• Click "Video Overview" or "Audio Overview" in the Studio panel.
• Paste the customize prompt into the customization text box.
• Choose the format and options that fit the goal.
• MANDATORY: You MUST physically type or paste the customization prompt into the text box. Use your type_text or fill_element tools. Do NOT just click the box and skip typing.
• Take a screenshot to visually verify your settings AND that the prompt text is visible in the text box.
• Click Generate.

VIII. Reference: Last-Known UI Options (may be stale)

As of 2026-04-27, the modals looked like this. Do not rely on this — always read the screen:

• Video Overview: Format = Cinematic / Explainer / Brief. Visual style carousel (only for Explainer/Brief). Customization textbox.
• Audio Overview: Format = Deep Dive / Brief / Critique / Debate. Length = Short / Default / Long. Customization textbox.

• 01_notebooklm_script.md — The complete formatted script ready for video production.