AI video shot prompt builder

Video models follow a prompt that reads like a shot list: who, doing what, where, and what the camera does. Pick the parts, get one clean prompt to paste into Seedance, Veo, Kling or any text-to-video model.

Why this order works

Text-to-video models weigh the start of a prompt most. Put the subject and action first, then where it happens, then the camera, then light and look. One camera move per shot: two moves in one prompt usually come out as neither. If you need a cut, write two shots and generate them separately.

Sound only matters on models that generate audio in the same pass, such as Seedance 2.5. Name the sounds you would hear, not the mood.