Vizard Agent

How to add a chapter progress bar to a demo video

Last updated 2026-10-10 · 2 min read

Upload the screen recording and name the sections. Vizard Agent builds a progress bar that advances with the video and labels each chapter along it, taking the bar's colour from your own interface screenshot so it reads as part of the product. It also crops to the application window and covers long static screens with supporting shots.

What do you type into Vizard Agent?

Name the sections and let it handle placement. Vizard Agent works out the timing from the narration, so the useful instruction is the list of chapters, not where the bar should sit. Keep the labels to two or three words and cap the supporting shots, or they take over.

Prompt

Make a feature introduction video from this screen recording and voiceover. Requirements: 1) The narration and the screen must line up: show whatever is being talked about. Where the screen sits still on one view for a long time, add related supporting shots, each no more than [2–4] seconds. 2) Crop sensibly so viewers can see what matters and the irrelevant parts of the screen are gone. 3) Add a progress bar labelled with each section's topic, advancing as the video plays.

What does Vizard Agent actually do?

Much of the work is the crop, not the bar. A third of the session went on one question: where exactly does the application window sit, and does it stay there? The bar's colour is then sampled from the product, not approximated. In the session behind this article it:

When does this not work well?

A progress bar takes up screen space and makes a promise about structure. Five or six labels is the practical limit, and long section names won't fit under the bar. A moving window means the crop has to follow it or stay wide. On a ninety-second video the bar is clutter, and vertical formats leave less width to label.

Common questions

How many sections work? Five or six. Vizard Agent will say when the labels stop fitting.

Will it match my brand colour? Yes. Vizard Agent samples it from a screenshot rather than approximating.

What if the window moves? Vizard Agent detects the coordinates at several timestamps to check.