AI motion transfer, from your own reference clip
Motion transfer gives an AI video generator a performance to follow. Your reference clip supplies the movement and camera direction; an optional image supplies a new character, outfit, or visual style. Genjutsu AI uses MiniMax H3 Max Reference to interpret those inputs together.
How to use this workflow
- Choose one short MP4 or MOV clip with a clear, continuous action. A single person walking, turning, or dancing is a better starting point than a montage.
- Add an image of the character or look you want. Use a clear view with a pose and framing similar to the source clip.
- Describe what Video 1 should contribute and what Image 1 should replace. Specify the movement, camera, setting, and lighting you want to retain.
- Generate a short variation and review the result. AI may change timing, details, or identity; adjust the prompt before producing another take.
A prompt to start with
Video 1 defines the walking action and slow tracking camera. Image 1 defines the new character. Recreate the walk with that character in a rain-lit city street. Keep a continuous medium shot and natural movement. Sound: light footsteps and distant traffic.
What to expect
Motion is a generation reference, rather than a frame-perfect constraint. Complex interactions, fast cuts, hands, and fine facial details can differ from the source. Genjutsu AI does not promise exact choreography or shot preservation.
Open the studio ↗Updated October 6, 2026 · Genjutsu AI