Higgsfield Genjutsu launched on September 1, 2026 as the company's strongest AI video transformation tool, letting creators recast an entire shot by transferring the exact motion, camera work, and timing of a source clip onto a new cast, location, and visual world. It ships with two modes: Motion Transfer, which rebuilds the scene while keeping the performance intact, and Object Swap, which changes one element and leaves everything else as filmed.
Try It: Recast a Shot Without Reshooting
Upload a reference video between 3 and 30 seconds, add a character, and Genjutsu transfers the acting, lip sync, camera movement, and VFX in a few clicks while holding every detail consistent. You can add up to 40 reference images to lock in characters, wardrobe, products, and locations. For a creator, that turns a single well-blocked take into a template: shoot the motion once, then recast it into different worlds instead of re-filming each variation.
Why It Matters for Creators
Motion is the hardest thing to fake in AI video, which is why most tools still generate from a text prompt and hope the blocking lands. By preserving exactly how a shot was moved and reworking only the world around it, Genjutsu recasts a whole shot rather than just the motion. That is a meaningful shift for ad and short-form creators who need consistent performances across many versions of the same scene.
Key Details
Motion Transfer. Keeps the original motion, camera, and timing, and rebuilds the cast, location, and visual world from reference images.
Object Swap. Changes one targeted element, a character, wardrobe item, product, or location, while leaving the rest of the footage untouched.
Inputs. Reference videos from 3 to 30 seconds, plus up to 40 reference images for visual elements. Higgsfield announced the tool on its official channel with a free trial.
What to Do Next
Open Higgsfield, pick one clean take with strong blocking, and run it through Motion Transfer with a new character to see how well the performance holds. The Genjutsu overview walks through both modes if you want the details before testing.
