A product demo video is useful only when it removes a sentence from the sales process. If the viewer still needs an explanation after watching it, the clip failed.
That is the standard we care about: one useful action, one visible result, one clear reason the feature matters.
Example output: a short Impala product moment that shows the page reacting to a user's attention. The goal is not a tour. The goal is one understood interaction.
The real value
The value is not that an assistant can record a screen. The value is that the recording becomes repeatable. When a website changes, the demo can be regenerated instead of manually recreated.
That changes how much quality we can afford. If fixing a clip is expensive, small mistakes survive. If fixing a clip is cheap, the ending frame, cursor path, timing, poster image, and file export all get better.
What we removed
- Live screen recording as a one-off performance.
- Manual cursor movement that changes every take.
- Editing timelines for tiny website interactions.
- Throwaway files with inconsistent names, sizes, and posters.
When a demo is worth making
Not every interaction deserves a video. A good candidate has a before and after that a still image cannot explain quickly.
- A booking flow that proves the next step is obvious.
- A calculator where the result builds trust.
- A catalog search where speed and filtering matter.
- A form or inquiry flow where friction has been removed.
The workflow
We use Playwright to drive the browser and FFmpeg to export web assets. Pi helps us work inside the project: read the page, inspect selectors, update the recorder, run the script, and review the output.
The important part is the written standard around it. A short project skill defines what good means: stable viewport, calm cursor, no accidental UI noise, a poster that matches the first frame, and a contact sheet for review.
We also changed the skill itself after using it. It now separates normal product demos from educational walkthroughs. Product demos stay short, usually 6 to 15 seconds. Tutorial clips can be longer, but they must show the real context first, keep commands readable, hold important frames long enough, and end with a clear result.
That distinction matters. A sales-page demo should make one feature obvious. A workflow tutorial should teach the process without turning into a fast terminal montage.
What the client gets
- A small MP4 for broad compatibility.
- A WebM version for lighter transfer where supported.
- A poster image that makes the page feel stable before playback.
- A repeatable script, so future copy or layout changes do not restart the whole process.
A good demo does not show how much exists. It shows how quickly someone understands one useful thing.
What we learned
The assistant did not replace taste. We still decide what is worth showing. We still judge pacing. We still cut anything that feels like a tour instead of a point.
The win is simpler: less throwaway recording work, more consistent web assets, and a process we can reuse whenever a project needs to be shown clearly.
