Does Seedance 2.5 preserve the face and clothing?
Last updated:
The editing prompt asks Seedance 2.5 to preserve the person’s facial identity, hair and clothing while following the fixed dance. That is an instruction, not a guarantee: details can drift during motion, occlusion or close-ups.
What the workflow provides
One photo is supplied as the appearance reference and the fixed dance video as the motion reference. Seedance 2.5 runs in editing mode at 480p. The request asks the model to change only the dancer’s appearance and retain choreography, framing, timing, background and effects.
Why details can change
A single image does not show every face angle, hidden body part or moving garment. Hands crossing the face, fast turns, small source details and motion blur can force the model to infer information, so identity, hair, clothing and anatomy may vary across frames.
Use a legible reference
Choose one evenly lit adult, fully visible with an unobstructed face, hands and feet. Avoid group photos, heavy filters, very small subjects and loose objects covering the body. Better visibility reduces ambiguity but does not guarantee fidelity.
Evaluate the finished file
Review the full video rather than a single frame. If the completed file is clearly broken, contact support with the video ID for review under the Refund Policy. Normal AI variation may still be present in a technically completed result.
Technical references
Ready to start with your photo?
Choose a photo