“Fold the top corner toward the middle,” says an illustrated craft teacher. She looks down and moves her hand across the desk. The desk is bare.

The instruction leaves the viewer searching for a sheet of paper. Before making a speaking character clip with OmniHuman 2.0, read the script while looking at the character image. When a line says “this corner” or “fold it here,” can the viewer see what those words mean?

If this is meant to be a lesson, put the paper where viewers can see it. Then check the finished clip carefully: does the hand touch the right corner, and is the fold clear enough to follow? A friendly voice cannot make an unclear step useful.

If the image shows only the teacher, change the recording into an invitation instead:

“Our origami club meets on Saturday. Bring a square sheet of paper. We’ll start with a simple fold.”

Now she can speak to the camera with her hands resting on the desk. The paper belongs to the event she is describing; it doesn’t need to appear beside her.

Use a character image and recording you own or have permission to use. Before publishing, play the clip once without sound and follow her hands. Then listen to the words. If she appears to demonstrate a step, make sure the viewer can actually see it.