How to Make a Launch Video with Veo or Seedance

A launch video does not need a studio anymore. A 20 second clip made of two or three generated shots and a clean screen recording can carry a launch on X, sit at the top of your listing and play on your landing page. Veo and Seedance make the generated part cheap enough to try on a weeknight.

What they do not do is make your product for you. The fastest way to a launch video that looks fake is to ask a video model for your app's interface. This post is the practical version: a shot list for 15 to 30 seconds, which shots to generate and which to screen-record, how to structure a prompt for Veo and for Seedance, four prompts to start from, and how to stitch and caption it all.

A disclosure: I built Prompt Generator, a free tool that writes prompts for specific video and image models, and it launched on Nick Launches this week. It comes up once below. The rest of the post works with any tool.

What the launch video is for

A launch video has one job: show someone who has never heard of the product what it does, before they scroll away. Everything in the shot list serves that.

It also has three homes, and they pull in slightly different directions:

• Your launch listing. On Nick Launches, a YouTube or Vimeo link (or a direct .mp4 file) in the Demo Video field plays inside the gallery as the first slide, ahead of your screenshots. Other links, Loom included, show as a plain link to the video. The player is 16:9, so a landscape cut fills it.
• Social. X, LinkedIn and the rest autoplay in the feed, often without sound. Vertical 9:16 takes more of a phone screen, but 16:9 works everywhere.
• Your landing page. A short loop near the top, usually muted, often without captions because the page around it does the explaining.

Cut one 16:9 master first, then reframe a 9:16 version from it if you post to vertical feeds.

A shot list for 15 to 30 seconds

This is the structure I would start from. Five beats, roughly 20 seconds, stretchable to 30 by giving the screen recording more room.

Two rules keep it tight. The hook shows the problem rather than saying it, because the first two seconds decide whether anyone watches the third. And the screen recording shows one task, not a tour: the single thing someone should remember.

What to generate and what to screen-record

Screen-record anything with your interface in it. Video models redraw interfaces from scratch, and the text on a generated screen tends to come out as near-words. Even when it looks plausible, it is not your product, and anyone who signs up will notice. Record the real thing at a high resolution, zoom in during the edit on the part that matters, and slow the cursor down.

Generate everything around the interface. People reacting, a desk at night, a phone buzzing on a table, the mood shot behind your end card. This is where Veo and Seedance save you a shoot, a location and an actor.

Use a start frame when you can. Both models can animate a still image instead of starting from text alone. If you already made a mockup scene or hero image for your launch visuals, upload it and write the prompt about the motion. The look stays consistent between your images and your video.

How to structure a Veo or Seedance prompt

A video prompt that works covers six things, in roughly this order:

1. Shot. The framing: close-up, medium shot, wide.
2. Subject. Who or what, specific enough to stay the same for the whole clip: "a woman in her 30s in a navy sweater", not "a person".
3. Action. One thing that happens, with a direction: "slides the phone across the desk toward the camera".
4. Camera move. One named move: static, slow push-in, pull-back, orbit, handheld follow. Two moves at once usually confuse the model.
5. Lighting. Time of day, the light source, the color of the grade.
6. Audio and duration. What it sounds like, and how many seconds it should fill.

Write one shot per prompt. A clip of a few seconds cannot hold a story with three scenes; you will get a blur of half-finished actions. If the idea needs a cut, make two clips and cut between them in the edit.

Veo reads one descriptive paragraph well. Veo 3 generates sound together with the picture, so I keep audio on its own line at the end, which stops sound words from leaking into the visuals, and add a duration line. Clips are short, typically around 8 seconds, which is plenty for one beat. If the app you use gives you a fixed length, write the duration anyway so the action is paced for it, and trim in the edit.

Seedance reads short, separate lines well: subject, motion, camera, setting and light, duration. It can also produce several shots in one generation if you name them (Shot 1, Shot 2), which is handy for a hook that needs a quick cut. Whether you get sound depends on the version and the app you use it through, so check before you write audio lines; if there is none, add sound in the edit.

The Veo and Seedance video prompt guide↗ (opens in a new tab) covers both formats in more depth, including dialogue and multi-shot prompts.

Four prompts to start from

Replace the bracketed parts. Each one is a single beat from the shot list.

1. The hook (Veo)

Medium close-up of a man in his late 20s in a gray hoodie at a cluttered desk late at night, rubbing his eyes and leaning back from a laptop whose screen we only see from behind. A stack of sticky notes and a cold cup of coffee beside the keyboard. Static camera at desk height, 35mm lens, shallow depth of field. Light from the laptop screen on his face, a dim warm lamp in the background, muted blue grade.
Audio: quiet room tone, a long tired exhale, a single notification ping near the end.
Duration: 4 seconds.

Swap the setting for the place your user actually has the problem: a café, a warehouse, a classroom.

2. The reveal, ready for your screen recording (Veo)

A laptop on a clean white desk, seen straight on so the screen faces the camera as a flat rectangle. The screen shows an even, bright pure green with nothing on it. The camera pushes in slowly until the screen fills most of the frame. Soft daylight from the left, minimal room, a small plant at the edge of the desk, calm neutral grade.
Audio: a soft whoosh as the camera moves, light room tone.
Duration: 4 seconds.

The flat green screen is for the edit: key it out and place your screen recording inside it, so the camera move lands on your real product. Because the camera moves, the recording has to follow the screen's corners; editors with corner pinning or tracking handle it, and if yours does not, change the prompt to a static camera. Expect to touch up the edges either way.

3. A hook with a cut (Seedance)

Shot 1: Close-up of a phone face down on a wooden table, buzzing and sliding slightly with each vibration. Warm afternoon light.
Shot 2: A woman in her 40s with short curly hair and a mustard cardigan picks up the phone, glances at it and frowns. Medium shot, same warm light.
Camera: static in Shot 1, slow push-in in Shot 2.
Duration: 6 seconds.

Keep the description of the person and the light identical if you regenerate one of the shots on its own, or the two halves will not match.

4. The end card background (Seedance, from your own still)

Upload your hero image or OG background as the start frame, then prompt only the motion:

Motion: the soft shapes drift slowly to the right and the light shifts gently from warm to slightly cooler. Nothing new enters the frame.
Camera: static.
Duration: 5 seconds.

Leave the text out of the prompt. Put your logo, your one line and your URL on top in the editor, where every letter comes out right.

If you would rather start from a line like "a hook shot for a meal planning app, tired parent at 6 pm", Prompt Generator's video prompt generator↗ (opens in a new tab) writes the full prompt for Veo, Seedance and a few other models, with the shot, camera move, light, audio and duration filled in. It is free for 10 prompts a day without an account.

Stitching it together

Any editor works: CapCut, DaVinci Resolve, iMovie, Descript, Premiere. The steps are the same.

1. Lay the screen recording down first. It is the spine of the video. Trim it to the one task and speed up anything that is waiting.
2. Put the generated shots around it. Trim each to its best two or three seconds. Generated clips often wobble at the start or end, and the middle is usually the keeper.
3. Cut on motion. Cutting while something moves (a hand reaching, the camera pushing in) hides the jump between a generated shot and a recording.
4. Match the color. A quick grade that pulls the generated shots and the recording toward the same warmth makes them feel like one video.
5. Lay one sound bed under everything. A single music track, low, plus the few effects that matter. If you keep Veo's generated audio, bring its level down so it sits with the rest.
6. Export a 16:9 master at 1080p, then a 9:16 reframe if you need one. Check the reframe: the important part of each shot has to survive the crop.

Captions

Burn the captions into the video. A lot of feeds start videos muted, and a viewer who does not hear the voiceover still needs the message.

Keep them short: one line on screen at a time, a few words per line, large enough to read on a phone. If there is no voiceover, write captions as the story: "Your inbox at 11 pm." "Paste the thread." "Get the reply drafted." Put them in the lower middle of the frame, clear of the edges where platform buttons sit on vertical video.

Read every caption against the picture before you export. A caption that says one thing while the screen shows another is worse than none.

Before you upload

• 15 to 30 seconds, one task shown, the problem in the first three seconds.
• Every frame with your interface in it is a real recording.
• Captions burned in, readable on a phone.
• A 16:9 master, plus a vertical version if you post to vertical feeds.
• Uploaded to YouTube (public or unlisted) or Vimeo so it plays inside your listing.

Then paste the link into the Demo Video field when you pick your launch week, and set a custom YouTube thumbnail, because the player shows it before anyone presses play.

FAQ

Can Veo or Seedance generate my product's interface?

They will try, and the result will not be your product. Video models invent interfaces, and the text on screen tends to come out garbled. Screen-record your real UI and use the models for everything around it: people, places, objects and motion.

How long should a launch video be?

15 to 30 seconds for the version you post on social and put in your launch gallery. That is enough for a problem, the product doing its main job, and a result. A longer walkthrough can live on your site for people who already want the detail.

Do I need sound in a launch video?

Make it work without sound first, because a lot of feeds start videos muted. Then add sound for the people who turn it on: a music bed, a few effects, or the audio Veo generates with the clip. Burned-in captions carry the message either way.

How do I add a launch video to Nick Launches?

Paste the link into the Demo Video field on the launch form. A YouTube or Vimeo link (or a direct .mp4 file) plays inside your gallery as the first slide, ahead of your screenshots. Other links, Loom included, show as a link to the video.