Short Video Overviews (NotebookLM) launchVerified Organic

The product

What is Short Video Overviews?

The important constraint is that the output is grounded in the user's own uploaded material rather than the open web. That is NotebookLM's whole design premise: it reads what you give it and refuses to wander, which is what makes it usable for study, legal reading, research and internal documentation where a general chatbot's confident invention is disqualifying.

This is a feature launch with a long lineage behind it, and reading the launch without that lineage misses why it landed. Google introduced Audio Overviews, the podcast-style two-host summary of a notebook, at Google I/O in 2024, and it became the product's breakout feature. Video Overviews followed in July 2025 as a horizontal, slide-style visual walkthrough that pulled images, diagrams, quotes and numbers out of the source material, and expanded to 80 languages in January 2026. Short Video Overviews is the third step: the same grounded-explainer engine, cut to the vertical, sub-minute shape people already scroll. Press reporting on the launch attributed the generation to Google's Nano Banana 2 Lite model.

The market it competes in is not really other note apps. It is the short-form educational video lane that TikTok, Reels and Shorts already own, where explainer creators earn attention a minute at a time. What Google is offering is the same format generated from a reader's own sources, which sidesteps the accuracy problem that dogs the human version of that lane.

What it does

• Turns uploaded sources into a 60-second vertical video explaining one concept from that material.
• Drafts the script, selects a visual style and records an AI voiceover automatically.
• Stays grounded in the user's uploaded sources rather than pulling from the open web.
• Launched to Google AI Ultra and Google AI Pro subscribers on mobile and web, with free access stated as coming later.
• Started in English, with more languages stated as following.

The launch was a single post carrying a 30-second vertical video, published at 9:01 AM Pacific on Tuesday 30 June 2026 from the @NotebookLM account.

The post opens on a joke, not a product name: Doom scrolling but make it educational, with a nerd-face emoji. Only the second line introduces the feature, and it does the whole pitch in one sentence, that Short Video Overviews turns your most complex sources into 60-second vertical videos that go deep on any concept. The third line is availability: rolling out now to Google AI Ultra and Pro subscribers on mobile and web, with free users noted as coming soon.

The video is the same shape as the product

The demo clip is vertical and it is 30 seconds long, which is half the length of the videos the feature actually produces. That is a deliberate and effective choice: the format of the launch asset is the format of the product, so a viewer understands what they will get without anyone explaining it. A horizontal demo of a vertical feature would have made the same claim and communicated none of it.

No thread, no giveaway, no dated discount, no creator wave. One post, one vertical clip, three lines of copy. Each element is doing a specific job and none of it is filler.

The tonal decision is the interesting one. A Google feature launch could easily have opened with the product name and the subscription tier. Opening on a self-aware joke about doom scrolling signals that the team knows exactly what format it is borrowing and is not pretending otherwise, and that costs nothing while making the post far more shareable.

The launch post reached 2,560,942 views on 11,923 likes, with 1,043 reposts, 422 quote posts, 377 replies and 7,670 bookmarks. These are matured public metrics, re-pulled directly from the source post on 26 August 2026.

At about 215 views per like this sits at less than half RADAR's roughly 500 organic ceiling, which is one of the tighter reach-to-engagement couplings in the tracked set. Underneath it the public layer is thick: 13,765 total public actions, about 0.537 percent of views. Note the ordering inside that layer, because it is unusual. There were more quote posts than replies, 422 against 377. A reply is a comment on someone else's post; a quote is a person carrying the launch to their own audience with their own framing attached. When quotes outrun replies, the post is being distributed by its readers rather than merely discussed.

The bookmark number is the whole story

The standout figure is the save layer: 7,670 bookmarks against 11,923 likes, a rate of about 0.64 bookmarks per like. For comparison, RADAR routinely sees infrastructure and news announcements save at a fraction of that. A like is a reaction that costs nothing and ends there. A bookmark is a person filing the post because they intend to come back and do something with it, which is exactly what a subscriber does when they see a feature they can go and use inside a product they already pay for. On this launch roughly two people saved it for every three who liked it, and that is the fingerprint of demonstrated intent rather than passive approval.

The verdict is a five-signal read rather than a ratio check. RADAR weighs how the view-velocity curve grew, whether the written engagement climbed in step with the reach or lagged behind a flat floor, whether the replying and quoting accounts read as real and spread out rather than as a coordinated cluster, and whether the quote and repost pattern matches word of mouth or a known amplifier ring. All of them agreed here, which is why the confidence is verified rather than reconstructed.

The numbers moved only slightly between detection and this settled pull. RADAR's detection-time snapshot recorded 2,527,487 views on 11,984 likes, a ratio of 211. The matured figures are 2,560,942 views on 11,923 likes, a ratio of 215: views rose by 33,455 while likes fell by 61 as the count settled. Both readings sit deep inside the same organic band. RADAR re-pulls launches because late engagement and count settling can move a ratio far enough to change a band, and this page shows the settled read.

The first lesson is the cheapest one to copy: make the launch asset the same shape as the product. A vertical 30-second clip announcing a vertical 60-second feature does the explaining without a sentence of explanation, and it removes the gap between what a viewer sees and what they will get. Founders launching a mobile-first product with a horizontal screen recording are throwing that away for free.

The second is the opening line. Three lines of copy and the first one is a joke that names the behaviour the product is borrowing. That is a decision about tone made by a team confident enough not to lead with the product name, and it is a large part of why this post was quoted 422 times. Self-awareness travels; a release note does not.

The third is a measurement lesson rather than a launch one. If you only watch likes you will miss what happened here. The save rate of 0.64 bookmarks per like is what told RADAR this feature had genuine pull, and it is the number a founder should be watching after their own launch, because it separates the people who approved of the post from the people who plan to use the thing.

FORKOFF builds launch videos designed for the format the product actually lives in, and plans the launch around them. See how we approach it on the viral launch video service, or read the best product launch videos we track for more real examples with the numbers attached.