ManifoldGen All creative tools API reference
SEEDANCE 2.0 · IMAGE + VIDEO + AUDIO REFERENCES
Direct a video from references.
Cast up to 9 reference images, point Seedance 2.0 at a motion reference video, and lay a soundtrack underneath: the model follows the choreography, cuts and camera of your video while your images set the cast and look. A single reference video up to 60 seconds is split at its cuts into short segments and joined back together.
EXAMPLE OUTPUT
Robot steps into the performance
SAMPLE VIDEOREADY
Robot reference image + source performance video + its own soundtrackSeedance 2.0
01 / IMAGES
Reference images · @Image1–@Image9
PNG · JPG · WebP · up to 90/9
Seedance rejects photos of real people and prompts that name them. Use face-free references (robots, characters, products, scenes) and describe people in words.
02 / VIDEO
Motion reference · @Video1–@Video3
No motion reference yet.
2–15 s total, or one video up to 60 s0/3
03 / AUDIO
Soundtrack · @Audio1–@Audio3
No soundtrack yet.
Audio or video files · audio is extracted0/3
04 / PROMPT
Direct the scene
05 / RUN
Settings and render
Estimate appears when signed in
mini · 720p · match reference
Sign in to generateREFERENCE VIDEOWAITING
No video yet. Add references, direct the scene, then render.
Build this workflow into your product
One JSON request returns an async job that directs a video from reference images, a motion reference video and a soundtrack, with long references split at cuts.