How do I transfer motion from one video onto a character?
Motion transfer needs two inputs: a reference video containing the movement, and a still image of the character who should perform it. The model produces a clip of your character following that motion. This is a distinct capability from general image-to-video — without a reference motion video there is nothing to transfer, so a general animation model is the wrong tool for the job.
Two inputs, both required
Wan 2.2 Animate takes exactly one character image and one reference motion video. Supplying only the character gives it nothing to follow. This constraint is the whole point of the model: rather than inventing plausible movement from a prompt, it copies specific movement that already exists, which is why choreography and gesture come across recognisably rather than approximately.
When to use it instead of image-to-video
Use motion transfer when the movement is the thing you care about — a specific dance, a signature gesture, a physical action that has to match a reference. Use a general image-to-video model when you want plausible motion and are willing to let the model choose it. Reaching for motion transfer for ordinary animation means doing extra work to source a reference video for no gain.
Reusing one motion across characters
Because the motion is an input rather than a property of the output, the same reference video can be applied to several different characters. That makes it efficient for content series where a recurring action needs to appear across a cast, and for testing how a piece of choreography reads on different body types before committing to a full production.
Related questions
Build it on Floniks
Image, video, digital humans, and reusable workflows on one canvas. No card required.
Explore Floniks