Framing for a destination, not for a canvas

The most common avoidable mistake in this product is generating a wide clip and cropping it vertical afterwards. It almost never works, and the reason is not resolution — it is that the model composed for the frame it was given, and the middle third of a good wide shot is rarely a good vertical one.

Generate in the shape it will be watched in

Ask for 9:16 if it is going somewhere vertical. The model puts the subject where a vertical frame wants it, leaves headroom where a vertical frame needs it, and picks a shot size that survives a small screen. None of that is recoverable by cropping.

If a piece genuinely needs both shapes, that is two generations from the same prompt, not one generation and a crop. It costs one more video unit and it is the difference between two good clips and one good clip plus one bad one.

What changes in the prompt

• Vertical wants tighter. A wide shot in 9:16 puts a person at about a tenth of the frame height and gives you a lot of floor and sky. Go one shot size closer than you would horizontally: what would be a wide becomes a medium, a medium becomes a close.
• Vertical wants vertical movement. A pan left to right crosses a narrow frame in half a second. A tilt, a rise, or a push in reads far better and holds attention for the length of the clip.
• Wide wants context. Sixteen by nine has room for a second element, and a wide frame with nothing in the outer thirds looks empty rather than minimal. Name what is on the sides.
• Square wants centre. One subject, centred, with air around it. Square punishes asymmetry more than either of the others, and it is the shape most likely to make a busy scene read as clutter.

Safe areas are real

Anything vertical is going to have interface on top of it — a caption bar, a profile row, a set of buttons down one side. Keep the subject out of roughly the top eighth and the bottom quarter, and keep anything that must be readable away from the right edge.

You cannot ask a model for a safe area directly, but you can ask for the framing that produces one: leave headroom above her, and subject in the upper third both do the job.

Stills are the cheap place to test this

Composition is settled in a still, not in a clip. Generate the frame at the aspect you need, look at it at the size it will actually be watched at, and only then animate it. Two credits to find out that the shot does not work vertically is a very good trade against a video unit.