Identity comes from your photo
The subject reference supplies facial appearance, hair, build and clothing. The instruction asks the model to replace only the attacking player and preserve those details throughout the clip. The goalkeeper and defenders remain part of the original reference scene.
Action comes from the reference
A single continuous sequence describes the approach, chip, run around the defence and final touch into the goal. The model receives the original reference video together with the photo. No runtime clip-stitching is needed for the first version.
A useful example instruction
“Replace only the attacking player with the adult person in the photo. Preserve their face, hairstyle, build and clothing. Follow the reference timing: approach the keeper, chip the same ball over the keeper, run around two defenders, reach the ball and finish into the net. Show the ball crossing the goal line. Keep one continuous shot, realistic football physics and a vertical 9:16 frame.”
A prompt is a request, not a guarantee
Even with a detailed instruction, identity, feet and ball motion can drift. We use a fixed template so the action can be evaluated consistently. Standard and HD change output resolution and credit cost; HD does not guarantee perfect anatomy or a stronger resemblance.