Genjutsu
Genjutsu changes a clip you already have — it keeps the clip's own motion, camera and timing, and puts what your pictures show into it.
Genjutsu changes a clip you already have. It keeps the clip's own motion, camera movement and timing, and replaces what your pictures show. When Studio offers it, you find it under Revisions on Studio's home and under Revision in Choose a model.
What it takes and makes
Edit — change a clip you already have
| Setting | What it takes |
|---|---|
| Your words | Required — up to 10,000 characters |
| Your clip | One video clip, 4 to 30 seconds long |
| Pictures | 1 to 8 pictures (Image 1, Image 2… in the order you add them) |
| Kind | Motion transfer or Object swap — you choose one |
| Length | Follows your clip |
| Quality | 480p, 720p or 1080p — starts at 720p |
| Shape | Follows your clip |
| Naming your references | Type @ and pick one — one of each kind, read as “the picture”, “the clip” or “the sound” |
| Sound | Silent video |
| Makes | A video |
The price is never written here — prices change. It is on the model's page in Studio: on the Make button and beside each length, size or quality that changes it, before you press.
Made by Higgsfield
Genjutsu is Higgsfield's own model — in Studio its name carries a small Higgsfield tag. What you send to it — your words, and the pictures or clips you add — is processed by Higgsfield under Higgsfield's own terms, which (in force since 27 August 2026) allow Higgsfield to use that content to improve its models. See the Privacy Policy.
Its two kinds
Before the price shows, choose a Kind:
- Motion transfer — your pictures say who or what moves; your clip says how it moves.
- Object swap — one thing in your clip is replaced by what your pictures show; the rest of the frame, and its movement, stay. It needs a clip whose frames are not too small.
What it is good at
Real movement with depth — a body, a camera that travels, something turning in space. Its notes are plain about where it struggles: footage whose movement is flat — shapes, type or layers sliding across one plane, with no depth, no camera and no body to carry the motion.
Writing for it
- Say what changes and into what; your clip already says how everything moves.
- It reads your pictures one of each kind — describe what each one shows in your words rather than numbering them.
- You can start from a template: Start from a template on its page.
For a revision that changes exactly what you name — a face, a jacket, the words on screen, the whole look — while keeping everything else, see Optic, ROUTIX's own revision model.