Imagine opening a tutorial because you can’t find the export button. A presenter says, “Open the menu in the upper-right corner and choose Export.” You hear the instruction perfectly. The problem is that the presenter is still on screen, and you have no idea which menu they mean.

That’s a useful editing test for anyone making a tutorial with an AI presenter. Look at the moment the script asks the viewer to do something. What should they be looking at right then?

For a short demo, I’d give the presenter one line at the start: “Here’s how to export just one page as a PDF.” A tool like OmniHuman 2.0 is one option to consider for turning an illustration and a voice recording into that opening shot. As soon as the narration gets to “Open the export menu,” I’d cut to the actual screen recording and keep the same voice going over it.

The words “this,” “here,” and “that one” make good edit markers. If the script says “choose this option,” pause on that frame and check whether a first-time viewer could point to the right option. If the menu is tiny on a phone, crop the screen recording closer. If the cursor is already moving on by the time the option is named, hold the shot a little longer.

There’s also a good reason to bring the presenter back. Say the export menu has two options with similar names. After showing both on screen, the presenter can return for one sentence that explains the difference: “Use Current Page for the page you’re on. Use All Pages for the whole file.” If that explanation is easier to follow with the menu still visible, leave the menu up and let the voice do the work.

Before recording the final voice track, try reading the script while clicking through the tutorial yourself. Wherever you have to stop reading to find a button, the viewer probably needs a beat too. Mark that spot in the script. It tells you where to hold the picture, where to pause the voice, and whether the presenter needs to be on screen at all.