Thirty seconds that ship a company: notes on great launch videos

I have a friend named Theo who has shipped, in his life, four companies. One of them was big. Two were medium. One was small enough that the launch video has been watched, last I checked, four hundred and twelve times — half of those by Theo's mother, who is loyal. He is now working on his fifth. The launch video for the fifth shipped at 4:11 a.m. on a Tuesday in March. I was on the call when it went up.

The reason I was on the call is that Theo had spent six weeks making the launch video and three days, in the middle of those six weeks, making the product. Both numbers are correct. He had agonized over the video — frame rates, music cues, the typeface of the closing card — while the product, a small consumer-products app for sharing photos with people you used to know, sat at 80% feature-complete. He shipped the launch video first. The product followed it by eleven minutes.

I do not endorse this order of operations. Theo, frankly, doesn't either. But he had a reason. He had decided, after watching every great tech launch video he could find, that the thirty seconds before someone hits download is more important than the next thirty days of the product. He was right and he was wrong, and the story of which is which is what this essay is about.

The form of the launch video

Modern tech launch videos have, give or take, four shapes. The shapes are stable across the last decade. If you watch the canonical examples — and I think you should, and the launches wall is a fast way — you will see the four shapes recur with depressing regularity. The good launches use one of them well. The bad ones use three of them at once.

1. The cold demo

Black screen. A cursor. A prompt. The product does the thing. There is no narration. There is, at most, a small ambient sound — a typewriter, a vibration, a hush of wind. The video is between fifteen and forty-five seconds. The OpenAI debut of ChatGPT was a cold demo. The Vercel deployment loop is a cold demo. Most of Anthropic's recent launch videos are cold demos.

Use this when the product is the message. Use it when you trust the visual to do the persuading. Do not use it when the product is conceptual, or when the demo only makes sense with context. A cold demo of a B2B SaaS analytics dashboard is, almost always, a mistake.

2. The talking founder

The founder, alone in a frame, talking to the camera. Sometimes with product cutaways. Sometimes without. The famous one, of course, was the Loom founder, sitting on a couch, recording himself explaining Loom with Loom. The recursion was the marketing.

Use this when the founder is the protagonist of the story. Use it when the company is new and small and the founder's conviction is the thing that's selling. Do not use it when you are a forty-person team trying to look like an indie hacker. The form does not survive scale.

3. The cinematic

A short narrative film. There are characters. There is a setting. There is, often, a single product moment near the end. The Apple product launches are cinematics, sometimes lasting four minutes. The Arc browser launch — the one where the dad is trying to plan a family vacation and the browser helps — was a cinematic.

Use this when you have the budget. Use it when the product is a consumer product and the emotion is the marketing. Do not use it when you are a small team without a film background. A cinematic that doesn't land is the worst kind of launch video — it costs the most and it lands the smallest.

4. The composite

Six product moments cut together with a single piece of music underneath. No narration. Sometimes captions. The video is forty-five to ninety seconds. The Linear product launches are composites. The Notion AI debut was a composite. The Granola launch video was a composite.

Use this when the product has multiple satisfying moments and you want to show range. Use it when you have an editor on the team. Do not use it when each moment, on its own, is not strong enough to earn its three seconds. The composite is unforgiving. A weak moment in a composite reads as a weak product.

Theo's video, in detail

Theo had built a composite. Six moments, ninety seconds, one piece of music that he had paid eight hundred euros to license. The first moment was the app generating a photo collage of a wedding. The second was a chat with an old friend. The third was a notification — Hema reacted — that he'd staged on his own phone with my own name because we were on a call. The fourth was a small celebration animation. The fifth was a friend list scrolling. The sixth was the app opening to a photo that the user had not seen in five years.

I will tell you what I told Theo at 3:55 a.m. on Tuesday, with sixteen minutes until the launch tweet went out. The first, second, fourth, and sixth moments were good. The third and fifth moments were filler.

Theo wanted to fight me. He had spent four hours on the fifth moment, which was a long scroll through a friend list, the kind of moment a designer makes when they want to show off interaction polish. It was, in fact, polished. It was also forgettable. The same fifteen seconds, given to the sixth moment — the photo nobody had seen in five years — would have made the video unforgettable.

We cut moments three and five. The video came in at sixty-one seconds. We pushed the launch by twenty minutes. Theo posted the launch video on X at 4:11 a.m. He went to bed at 4:14. He woke up at 11. The video had eighty-three thousand views. The app had twelve hundred sign-ups.

I am not going to tell you that the two cuts did it. The product is good. The timing was right. But I will tell you that Theo's first instinct was to ship a ninety-second launch video, and the sixty-one-second version performed measurably better, and the difference between those two videos was the willingness to cut his own work.

What I learned watching too many launch videos

I watched, between January and April of this year, two hundred and forty launch videos. I was building a small reference deck for a software agency that asked me to do an audit. Two hundred and forty is, depending on how you count, somewhere between four and six full work weeks. I am only slightly proud of this.

Here is what I learned.

The good launch videos do one thing

Almost every great launch video can be described in a single sentence. A cold demo of ChatGPT writing a poem about whales. A talking-head of Loom's founder explaining Loom with Loom. A composite of Notion AI replacing six dialog boxes in the product. A cinematic of a dad using Arc to plan a family vacation.

The bad launch videos can't be described in a sentence. They are about "the future of work" or "a new way to collaborate" or "reimagining how teams operate." They try to be a brand campaign and a product demo and a recruiting pitch all at once. They fail at all three.

The good launch videos earn the first three seconds

First three seconds: a satisfying moment, or a question the viewer has to keep watching to answer. Not a logo. Not a title card. Not a montage. A thing that earns the next twenty-seven seconds.

If you have to use a title card, put it at the end. The viewer who has watched ninety seconds of your product is the viewer who wants to know your company's name. The viewer in the first three seconds wants to know whether to keep watching.

The good launch videos understand their distribution surface

A launch video on X is a square crop with autoplay-on-mute. The first three seconds have to make sense without sound. A launch video on YouTube has audio and aspect ratio in your favor. A launch video on a product page can be longer, because the viewer chose to be there.

Most launch videos are shot for YouTube and posted to X. They fail on X because the team didn't watch them with the sound off. This is the most common, most fixable, most embarrassing mistake in launch-video production. Watch your video with the sound off. Watch it on a phone. Watch it in a square crop. Then ship it.

The reference wall, if you want one

The launches collection is a wall of the tech launch videos that shaped the feed — frontier-model debuts, devtool launches, AR-glasses reveals, AI-agent demos. Each card opens the brand's launch videos on X. If you are studying the form, this is the wall.

If you got here from a search for "launch videos" specifically, the /for/launch-videos page is the same wall framed for that intent. Both work. The walls are the same. Different doorway.

If you got here because you are studying launch videos because you are about to ship one of your own, the move is this: pick the shape (cold demo, talking founder, cinematic, composite), study the four canonical examples of that shape, write the one sentence your launch video is about, and then build the video to that sentence.

Then cut twenty percent of the seconds.

Then cut another ten.

Theo, six months later

Theo's app has, last I checked, about forty thousand monthly active users. It is not a unicorn. It is a small consumer product that brings people a tiny amount of joy, intermittently. He is working on the next launch video now. He has not asked for my notes, but I am going to give them anyway, because that is what friends are for.

The notes are short. There are three of them. They are in purple, on a printout that is sitting next to his keyboard, because Theo started using a purple pen the week after the launch and has not stopped.

I did not give him the pen. He bought it himself. I think he read an essay somewhere.