Guide, updated September 5, 2026
You vibe-coded the app. Now the demo video, without learning After Effects.
The app took a weekend. Cursor wrote the boring parts, Claude fixed the bug at 2 am, Vercel deployed it. Then you needed a demo for the landing page and the whole thing stopped, because Loom looks like a tutorial, CapCut chops it up, and the freelancer wants a brief. Here is the plan that fits the way you already work: describe what you want, let the tool do the mechanics, keep what is good.
What you are making
- One flow, sixty seconds. The thing your app does, from the user's first click to the result.
- Three outputs from one recording. Wide for the landing page, vertical for the launch post, a short loop for the README.
- Real screens only. If it is in the video, it is in the app.
The same move you made with the code
Vibe coding works because you stopped writing every line and started saying what you wanted. The demo works the same way. You do not keyframe zooms and animate a cursor; you record the real thing and let the recorder decide where to zoom, how to draw the cursor, which pauses to cut, what the captions say. Then you look at what it did and fix the two places it got wrong. That is the whole skill.
The tool for that is an AI screen recorder. On Windows, that is what Ciakko is. The rest of this page assumes it, but the plan is the same with any tool that does the editing for you.
Pick the flow, not the app
The mistake everyone makes is the tour: settings, dashboard, every menu. Nobody wants a tour. Pick the one thing someone would sign up for. "Paste a URL, get the summary." "Upload the CSV, see the chart." "Type the prompt, get the image." Start where the user starts, end where the value is, and stop. If your app does three things, that is three short videos, not one long one.
Set up the screen like a screenshot you would be proud of
Real-looking data, not "asdf". A browser with the extra tabs closed and the bookmarks bar hidden. Page zoom at 110 or 125 percent so text is readable before any zoom. Notifications off. Dark or light theme, whichever your landing page uses. If the app has a loading state that takes four seconds, it will get sped up later, so do not fake it.
Record it talking, or record it silent
Talking is better for the landing page: a voice saying "so you paste the link here" while the cursor does it is what makes people feel the product works. Keep the three lines next to you: the problem, the product doing the thing, what the user has at the end. Say them in your own words.
Silent is fine for the README loop and for a tweet. Then the captions become the narration: the recorder can write them from your voice, or you type two or three short ones in the editor.
Either way: one take, mistakes included, under three minutes. Click on purpose, because the zooms land on clicks. When you mess up, say "again" and redo that bit. It gets cut.
Review what the tool did, fix two things
Stop recording and the editor opens with the zooms already on the timeline. Play it once. Usually two things need a hand: a zoom that framed the wrong control (drag it), and a zoom on a click that did not matter (delete it). Cut the "again" moments. Let the app speed up the stretches where the page was loading. Pick a background that matches your site, round the corners, move the webcam bubble if you recorded one. If you spoke, read the captions and click the word it got wrong.
This is ten minutes the first time and five the second. If you find yourself half an hour in, you are polishing, and the launch does not need polish, it needs the video.
Export three things
Landing page: wide, 1080p, with sound. Put it in the hero, muted autoplay with a play button for sound, the way the good pages do it.
Launch post, Product Hunt, Shorts: vertical from the same project. The camera keeps the action centered, so a desktop app becomes a phone video without losing the important half of the screen.
README: the same clip, short, embedded at the top. A twenty-second loop of the tool doing its thing is the best README section anyone has written.
What not to do
Do not try Sora or Runway for this. They will make a lovely video of an app that is not yours, and the first person who tries your product will notice. We explain why on the fake screens page. Do not hire a motion designer for the launch either: the quotes in the founder threads are hundreds of dollars for a minute and weeks of waiting, for a video that shows less of the product than the one you can record tonight.
The bill
Ciakko is $89, once, and the trial is the full product for 14 days with no card and no watermark. That is enough to ship the launch video and decide later. You will have paid more for the domain.
Questions from the launch threads
Is a screen recording enough, or do I need a real video?
A screen recording with the boring parts cut, zooms where the action is, a readable cursor and captions is what a real product video looks like in 2026. The motion-graphics version is a different product for a different budget, and launches do not need it.
Should the demo be on the landing page or on YouTube?
Both, same file. Host it on your own page as the hero so people do not leave, and upload it to YouTube for the search and the embed. The launch post gets the vertical cut.
How do I get the vertical version for Shorts and Reels?
Record once in 16:9, then export 9:16 from the same project. The camera follows the action so the important part stays in frame instead of getting cut off.
What if my app is ugly right now?
Show the one flow that works and frame it tight. A zoom on a working feature beats a wide shot of an unfinished dashboard. Ship the video with the app, improve both.
Can I make the demo video with AI too?
For the editing, yes, that is what an AI screen recorder does. For the footage, no. Generators invent interfaces, and a fake button in a launch video is the one thing your first users will notice.