How to Prompt a Cafe POV Video

The scene needs two roles. Person 1 starts the clip already sitting on the small round tabletop. Person 2 walks in from the right and uses the low bench the way it was meant to be used.

If your model accepts character references, bind each photo to a separate element and name them in the prompt. If it only accepts a single image, describe the second person in words and accept that their face will drift.

Setup: the first person sits confidently on the tabletop with a coffee, and the bench beside it stays clearly visible and empty.

Reveal: the second person enters from the right, sits on the bench, and puts their cup on the round table. That single action tells the audience the table was never a stool.

Recovery: the first person glances down, pauses in visible embarrassment, then slides onto the bench and tries to look natural while sipping.

Ask for one continuous vertical 9:16 shot. Keep the room, furniture, lighting and original audio unchanged. State that no cuts, no close-ups and no camera shake are allowed.

Explicitly say both surfaces stay in frame for the whole shot. Without that line, models tend to tighten the framing after the second person sits down, and the reveal stops reading.

Vertical 9:16 comedy scene in a minimalist cafe. One person sits confidently on a small round table beside a low bench, holding a cup of coffee. The bench stays clearly visible and empty. Another customer enters from the right, sits on the bench, and places a coffee on the table. The first person notices, pauses in embarrassment, then moves to the bench and tries to look casual. Keep both seats visible in one steady wide shot, with natural motion and consistent facial features. No cuts, no text overlays.

Watch the seat switch frame by frame. Hands should not merge with the cup, the table should stay distinct from the bench, and the second person must sit down before the first one moves.

Then check both faces at the moment the first person looks down. That is where likeness most often slips.