How to make a product demo video with AI

A demo lives or dies on the script and the screenshots. Here is how to use your own of both, keep the text readable, and spend nothing on visuals.

A demo of the demo: the scene strip and the scene editor are the actual interface, not an illustration.

Write the script first. It is the whole job.

A product demo fails or succeeds on what it says, not on how it looks. Most tools hand that decision to a model and give you something generically enthusiastic. Paste your own script instead and it is narrated verbatim, one scene per sentence.

A demo script that works is usually four movements:

1. The problem, in their words. One sentence. Not your product.
2. What it does. Plainly. No adjectives you would not say aloud.
3. The proof. The one thing competitors cannot copy.
4. What to do next. One action.

At roughly 40 words per 30 seconds, that is about 80 words for a minute. Short is not a constraint here — it is the point.

Use your own screenshots, and hold them still

A demo is mostly screenshots, and screenshots carry text. The default slow Ken Burns pan flatters a photograph and ruins a screenshot: it drifts across the interface and crops the very label the narration is reading out.

So upload your screens under Your own images, and set Movement on still scenes to Hold still. Each image sits perfectly steady, fully in frame, for exactly the length of its line.

Cost

A demo built from your own screenshots generates nothing — no AI images, no Pollen spent. Narration is free in all 14 languages. If you need connecting shots you have no screenshot for, free stock footage and free stock images are unlimited.

In practice a product demo is one of the few videos that costs literally nothing to make here, because you are supplying the pictures.

Then cut it down

First cuts are always too long. Every scene is stored separately, so dropping the two that drag is a re-stitch of clips that already exist: seconds, free, nothing regenerated, and reversible if you change your mind.

Watch it once at full speed. The scene you reach for the scrubber during is the scene to remove.

Where it will be watched

Set the frame before you generate. 16:9 for your site, a docs page or YouTube; 9:16 if it is going on Reels, Shorts or TikTok — captions sit higher there so the platform's own buttons do not cover them. Most people watch muted, and the captions are burned in, so the video has to work with the sound off.

Related: turning a script into a video · vertical for TikTok and Reels.