Start with the source photo
A small, shadowed or filtered face gives the model less detail. Try a sharp portrait of one consenting adult, facing the camera in even light. Put the clearest portrait in the lead slot, which gets the longest close-up.
Check the cast assignments
Review each uploaded photo against its selected role before generating. The lead is required; sax, keys and the two backing singers are optional. Unfilled roles are played by invented AI adults. AI can still confuse roles, so pause the finished video at both close-ups and wider shots.
Watch the sax transition
All Seedance options generate 0–13 seconds and 13–19 seconds separately and join them at the sax cut. Seedance 2.0 also passes the first part's last frame to the second; Mini, Fast and 2.5 do not. Look for changes in faces, outfits or the room at that cut and listen for an audio change. Continuity is not guaranteed.
What if generation fails?
Credits reserved for a job are returned automatically if generation fails or times out. A completed but clearly broken video is handled under the credits and refunds policy. Check the result before retrying, and change one input at a time so you can tell what helped.