01

Copy the starting prompt

Photorealistic full-body shot: the person from the reference photo wearing a red velvet Santa suit with white fur trim and a matching hat, dancing a playful holiday routine in a cozy living room with a decorated tree, warm fairy-light bokeh, light snow outside the window, camera at chest height, full body in frame including feet, consistent face and identity, smooth natural dance motion, photorealistic live-action, no text and no logo.

02

Add the negative list

Extra fingers, extra limbs, face morph, identity drift, cropped feet, floating above the floor, cartoon, 3D render, watermark, text, deformed hands, duplicate person. Most unusable Christmas dance clips fail on hands or feet, so name those failures explicitly rather than hoping the model avoids them.

03

Keep the camera still

A fixed camera at chest height keeps the costume and body proportions readable. If you ask for a whip pan, a zoom and a dance in ten seconds, the model spends its capacity on the camera move and the face drifts. Change one variable at a time so you can tell which change helped.

04

Respect the model’s input format

Photo-reference syntax differs between Seedance, Kling, Runway and open models. Check the provider’s current documentation before pasting tokens or attaching a reference clip. A clip you can watch online is not automatically licensed for processing or for commercial publication.