NemoClaw Knowledge Wiki

NotebookLM advanced prompts - Tool Drop Channel

5 min read

https://www.youtube.com/watch?v=Lx7Bji_sNWM Here is a comprehensive Markdown guide based on the workflow detailed in the video.

🎥 The “Admin Privileges” AI Video Workflow

Goal: Create high-retention, broadcast-quality documentary videos (motion, high-res, voice-matched) for free, avoiding the low-quality “slideshow” look.

🛠️ The Tech Stack (Free/Freemium)

🚀 Step-by-Step Workflow

Phase 1: Research & Strategy (NotebookLM)

The goal is to ground the video in facts, not hallucinations.

  1. Curate Sources:
    • Go to YouTube.
    • Search your niche (e.g., “Quantum Computers and AI”).
    • Filter: Sort by upload date → “This Month” (Critical for fresh data).
    • Open 10+ high-performing videos and copy their URLs.
  2. Feed NotebookLM:
    • Create a new notebook.
    • Paste the YouTube URLs as sources.
  3. Generate Strategy (Master Prompt #1):
    • Ask NotebookLM to analyze the transcripts.
    • Objective: Find hooks, retention gaps, and structure a unique angle.
    • Action: Click “Save to Note” on the output, then “Convert to Source.” Now the AI “knows” your strategy.

Phase 2: The Script (NotebookLM)

Do not let the AI write a generic blog post. Force it to write for audio.

  1. Generate Script (Master Prompt #2):
    • Prompt the AI to write a script optimized for voiceover (teleprompter format).
    • Instructions: No scene headers, use specific hooks found in Phase 1.
  2. Export:
    • Copy the text to a Google Doc.
    • Warning: Do not use the “Generate Video” button inside NotebookLM (it creates a static slideshow).

Phase 3: Voiceover (Google AI Studio)

Get unlimited, human-sounding speech.

  1. Go to Google AI Studio → Playground.
  2. Select “Text to Speech”.
  3. Model: Choose Gemini 2.5 Pro (or latest robust model).
  4. Settings:
    • Mode: Single Speaker.
    • Voice: Select a deep/authoritative voice (e.g., “Deep American Female”).
    • Style Instructions: “Deep, authoritative, fast-paced.”
  5. Batch Processing:
    • Important: Do not paste the whole script. Copy/paste in 2-minute chunks to prevent the AI from degrading in quality/laziness.
    • Download the audio files.

Phase 4: Visual Roadmap (LM Arena)

Create a precise map so visuals match the audio perfectly.

  1. Go to LM Arena → Direct Chat.
  2. Model: Select Gemini 3 Pro (needs to understand complex logic).
  3. Generate Shot List (Master Prompt #3):
    • Input: Total audio length and clip duration (set to 5 seconds).
    • Paste your script segment.
  4. The Output: The AI should generate a Table with three columns:
    • Column 1: Script Line.
    • Column 2: Image Prompt (Detailed, photographic style).
    • Column 3: Video Motion Prompt (Camera movement instructions).

Phase 5: Image Assets (Google Flow)

  1. Go to Google Flow (ImageFX).
  2. Model: Nano Banana (or Nano Banana Pro).
    • Tip: If you hit the limit on Pro, switch to the standard model (it is unlimited and free).
  3. Action: Copy the Image Prompt from your LM Arena table.
  4. Download: Save the high-res images.

Phase 6: Motion Generation (Meta AI)

Turn static images into video clips.

  1. Go to Meta AI → Create → Video Mode.
  2. The Magic Step:
    • Upload the image you generated in Phase 5.
    • Paste the Video Motion Prompt from your LM Arena table.
  3. Result: The AI animates the image (pans, zooms, lighting shifts) while maintaining the exact composition.
  4. Download the video clip.

Phase 7: Assembly (The Sync Trick)

  1. Open your video editor.
  2. Audio First: Drag your voiceover onto the timeline. Lock the track.
  3. Visuals Second:
    • Do not guess placement.
    • Because you generated clips in 5-second chunks based on the Shot List Table, simply drop the clips in order. They will naturally sync with the script lines they were generated for.
  4. Polish: Add background music and simple dissolve transitions.

📝 The “Master Prompts” Structure

(Note: Adjust the bracketed info for your specific topic) Prompt 1 (Strategy):

“Act as an expert YouTube Producer. Analyze the provided transcripts from top-performing videos. Identify hooks, retention gaps, and structural patterns. Synthesize this data to create a ‘Category King’ video strategy—one that outperforms individual sources. Create a ‘Retention Architecture’ outlining the Opening Loops, The Setup, and The Body.”

Prompt 2 (Script):

“You are a professional YouTube Scriptwriter. Write a script based on the [Strategy Source] we just created. Rules:

  • VOICE OVER ONLY: Do not write scene headers or visual directions.
  • Format: Teleprompter style (continuous text).
  • Tone: High-energy, Motivational, Fast-paced.
  • No Fluff: Do not use conversational filler. Start instantly with the first word of the video.”

Prompt 3 (Visual Shot List):

“Act as an award-winning Documentary Film Director and AI Prompt Engineer. Task: Generate a visual Shot List table. Configuration:

  • Total Audio Length: [Insert Time, e.g., 2 mins]

  • Clip Duration: 5 Seconds (Strict)

  • Visual Style: High-End Documentary, Photorealistic, Cinematic, 16:9. Output: A table with 3 columns:

  1. Narrative Context (Script line)
  2. Image Prompt (Midjourney/Dall-E 3 style description)
  3. Video Motion Prompt (Camera movement, e.g., ‘Slow push in’, ‘Vertical tilt’)”

Prompt 1 https://drive.google.com/file/d/1gF7Y6cYZ5DYVK4NcLquIZU-pry3QYHWO/view

Prompt 2 https://drive.google.com/file/d/1nhD1ZdZ-8KujUog4ZAAiRPTBAQghGo6l/view

Prompt 3 https://drive.google.com/file/d/1x1Yf7AuP-Ex5HdpQaUvxTSR5ex_wc-62/view

Graph View