Make your character move with the reference

Prepare a pose-matched image so Motion Control has a clear identity and a performance it can actually follow.

The source clip has the move. Your image has the character. The best preparation is making those two inputs look like they belong to the same performance.

If the video shows a full-body dancer and your image is a tiny headshot, a longer prompt is not the first fix.

Find a readable starting moment

Open Video’s Motion tab. Choose an available motion model and inspect its input requirements.

Use a clip with one clearly visible performer and a movement you can follow. For the first transfer, skip rapid cuts and crowded choreography. Pick a clear source frame where the needed body parts and pose are visible.

Put your character into that pose

Prepare the character image in Image with an edit-capable model. Attach the source keyframe and your identity reference in a clear order:

Image 1 defines the pose, body placement, crop, and camera viewpoint. Image 2 defines the replacement adult character's face, hairstyle, and outfit. Place that character in the pose and position from Image 1. Match the visible body silhouette and source framing. Keep the face, hands, and feet readable, with believable anatomy.

Compare the prepared still to both sources: face from the character, placement from the performance. Fix a wrong pose before paying for motion.

Let the clip own the choreography

Attach the prepared image and source video in Motion. Choose the options supported by the selected model and inspect the displayed total.

Keep extra direction focused on appearance and scene coherence. Asking a seated reference to sprint away competes with the performance you supplied. For invented action from a still, use an image-to-video workflow instead.

Find the first moment that breaks

Watch from beginning to end. If a hand disappears behind the body or the face changes during a turn, note the earliest bad moment.

A clearer pose, better crop, or shorter source excerpt may solve more than another expensive setting. Check that same moment after the next attempt.

Your starting exercise: one performer, one readable movement, one compatible character frame. You should recognize both the character and the source performance in the result.

Next: direct an original shot with image to video.