Lock the character with Seedream first, then let video models shoot that face
Seedream generates character lockup and empty-scene images, then Jimeng and Seedance run image-to-video. Free signup. No face drift.
Video models love to lose a face by the second shot. Seedream builds four-view character sheets, close-up portraits, and empty scenes — images that follow every later shot.
After the character design is done, Jimeng (Dreamina) and Seedance take over with image-to-video. Change the outfit or the age by editing the design, not by reshooting the whole episode. Images are billed per piece, signup is free.
常见问题
Can I skip character design and just generate video?
You can, but a serialized drama will drift faces fast. If the same face must appear many times, lock the character first.