NotebookLM Now Turns Your Notes Into Cinematic Videos
Google's NotebookLM can now generate documentary-style cinematic videos from uploaded documents using Gemini 3 as creative director and Veo 3 for visuals - a major step beyond its viral audio podcasts.
4 min read
Google's NotebookLM launched Cinematic Video Overviews on March 4 - a feature that turns uploaded documents into short, narrative-led videos with fluid animations and AI-created visuals. It's the same tool that went viral in 2024 with its AI podcast feature, now extended from audio to full video production.
TL;DR
- NotebookLM now generates documentary-style cinematic videos from uploaded sources
- Uses Gemini 3 as "creative director", Veo 3 for video generation, and Nano Banana Pro for narration
- Replaces previous Video Overviews that were essentially narrated slide decks
- Available to Google AI Ultra subscribers ($249/month) - limit of 20 per day
- Accepts PDFs, docs, web articles, transcripts, and YouTube links
- Rolling out in English on web and mobile
How It Works
The pipeline chains three AI models together. When you hit "Generate Cinematic Video Overview," NotebookLM first sends your uploaded sources to Gemini 3, which acts as what Google calls a "creative director" - it reads your materials, selects the key narrative arc, and makes hundreds of structural and stylistic decisions about how to tell the story. Gemini determines shot composition, pacing, transitions, and which parts of your content deserve visual emphasis.
The visual assets are then generated by Nano Banana Pro, Google DeepMind's image generation model built on Gemini 3 Pro. It creates the still and animated visuals - diagrams, infographics, scene imagery - supporting up to 4K resolution with legible text in multiple languages. It's the same model powering NotebookLM's Infographics and Slide Deck features since November 2025.
Finally, Veo 3 takes those assets and produces the actual video - fluid motion, animation, and cinematic transitions stitched into a cohesive sequence.
The result is a short video that looks more like a mini-documentary than a PowerPoint presentation. Google's previous Video Overviews were basically slide decks with a voiceover - static images accompanied by narration. The cinematic version produces actual moving imagery with dynamic visuals.
What You Can Feed It
NotebookLM accepts the same source types as before:
| Source Type | Supported |
|---|---|
| PDFs | Yes |
| Google Docs | Yes |
| Web articles | Yes |
| Transcripts | Yes |
| YouTube links | Yes |
| Pasted text | Yes |
You upload your sources to a notebook, open the Studio panel, and select "Cinematic Video Overview." You can optionally add a text prompt to steer the output - "Create a three-minute explainer for a non-technical audience" or "Compare the two approaches and show trade-offs." Generation can take 30+ minutes for complex content. All generated visuals stay grounded to your uploaded sources rather than drawing from the broader internet.
Availability and Limits
Cinematic Video Overviews are available to Google AI Ultra subscribers (the $249/month tier) on web and mobile. The feature is rolling out in English first, with a cap of 20 video generations per day per user. Users must be 18 or older.
This is notably gated behind Google's most expensive consumer AI subscription. The AI podcast feature that made NotebookLM famous was available on the free tier. Cinematic video is positioned as a premium capability - the compute cost of chaining Gemini 3, Veo 3, and Nano Banana Pro for each generation is substantially higher than text-to-speech alone.
Why It Matters
NotebookLM's audio podcasts went viral in late 2024 because they made a truly useful AI feature that felt magical - upload a paper, get a natural-sounding podcast discussion about it. Cinematic video is the same playbook, raised to a higher level.
The use cases are immediate: researchers turning papers into presentation videos, students creating study materials, educators building lecture content, marketers converting reports into shareable video. The quality gap between "AI-generated video from a PDF" and "professional video production" is closing faster than most people expected.
The broader significance is the multi-model pipeline. Google is chaining its own models - Gemini for reasoning, Nano Banana Pro for visuals, Veo for video - into a single product workflow. This is the kind of vertical integration that's hard to replicate without owning all the pieces.
The limitations are real: $250/month is steep, English-only at launch, 30+ minute generation times, and no clarity on export formats. But the direction is unmistakable - the gap between "upload a document" and "get a professional-looking video" now involves zero human production work.
Sources:
