01
Choose your reference
Upload a short, clear video of the action and camera movement you want the model to follow. Use footage you have permission to transform.
YOUR MOTION. ANOTHER REALITY.
Genjutsu AI.
A new take on reality.
Start with a video. Recast the character, change the scene,
and let your reference guide the motion.
Reference video + optional image + your direction · MiniMax H3
ONE PERFORMANCE. MORE POSSIBILITIES.01 / REIMAGINE
01 — REFERENCE VIDEO
Start with the movement.
02 — YOUR NEW DIRECTION
Imagine another world.
Workflow illustration · create your own result in StudioTHE WORKFLOWBring the performance.
Bring the performance.
Change the possibilities.
One reference clip, one clear idea, a new generated take.
02
Define the new look
Add an optional character or style image. Describe the setting, appearance, lighting, and what the reference video should contribute.
03
Generate and review
Choose your output settings, check the credit estimate, and start the job. Your completed video appears in your private account history.
MAKE IT YOURSSame inspiration.
Same inspiration.
A different creative direction.
Motion transfer
Use your performance as a motion reference for a new character or style. Try a walk, a turn, or a simple gesture.
Explore motion transferScene restyling
Reinterpret a familiar shot in a cinematic city, a painted landscape, or another setting you describe.
Explore video to videoCreative variations
Make another take for a music clip, concept film, or social post. Compare the result with your original footage before publishing.
Try your ideaA reference guides.
AI reinterprets.
Genjutsu AI regenerates a video using MiniMax H3. It can reinterpret movement, appearance, and scenes; it does not guarantee identical timing or pixel-perfect object replacement. Start with a simple shot and review each result.
How this differs from Higgsfield GenjutsuA LITTLE CLARITY
Before your first take.
What is Genjutsu AI?+
Genjutsu AI is an independent reference-video generator. It uses MiniMax H3 Max Reference through fal to create a new clip from your video, an optional reference image, and a prompt.
Can I transfer motion to a different character?+
You can use a video to guide motion and an image to guide the new character or style. Describe the role of each input. The output is a generated interpretation; exact timing and identity are not guaranteed.
Which videos can I upload?+
Upload one MP4 or MOV, 2–15 seconds long and up to 50 MB, encoded with H.264 or H.265 at 24–60 fps. The minimum dimension is 256 pixels and the supported aspect ratio range is 2:5 to 5:2.
What output settings are available?+
Choose a 5, 10, or 15 second video at 480P or 768P in portrait or landscape. The model generates audio; the original soundtrack is not guaranteed.
Is this the official Higgsfield Genjutsu?+
No. Genjutsu AI is independent of Higgsfield. Our generation model is MiniMax H3 Max Reference, which has a different implementation and limitations. We do not provide Higgsfield's proprietary Genjutsu model.
Can I replace an object without changing the rest of a shot?+
This version regenerates the video from references. It does not offer masked object replacement or promise to preserve every untouched pixel.
Is billing available?+
Billing is not open yet. Generation uses account credits, and early-access credits are arranged through support. No paid subscription or checkout is currently offered.
Who can view my uploads and results?+
Your account's history is private and references are not added to a public gallery. Media is stored in private storage and served to the signed-in owner. The generation provider receives time-limited links to the selected references.