Genjutsu video: turn one clip into a new scene

What Genjutsu video is

Genjutsu is an AI video tool for video-to-video work. You bring a clip, and the model keeps the motion, the camera and the timing while generating a new version of that shot from your reference images and a short prompt. There is no timeline, no masking and no keyframes.

That makes it useful in one specific situation: you already have the take you want, and the only thing wrong with it is who or what is in it.

The two ways to transform a clip

• Motion Transfer rebuilds the whole scene. Movement, timing and camera come from your footage; the cast, the location and the look come from your references. This is the mode for staging a shot somewhere else.
• Object Swap changes one element and preserves the rest of the shot. This is the mode for wardrobe changes, product swaps and single-character replacements.

What you feed it

• A reference video between 3 and 30 seconds. One clean take works far better than a cut sequence.
• Up to 40 reference images per project, covering characters, products, wardrobe and locations.
• A one or two sentence prompt naming the element that changes, the reference to use, and what must stay.

What you get back

• A recast video that keeps the original action, so it cuts together with footage you already have.
• Watermark-free downloads on any active plan.
• Commercial use: what you generate is yours to publish across organic and paid channels, subject to the terms of use.
• A credit cost shown before each run, so a generation never surprises you.

Who Genjutsu video is for

• Editors who have the right shot but the wrong location, and no budget to reshoot it.
• Small brands producing product demos with a stand-in and swapping in the real product.
• Creators building a series around one performer, one character or one visual identity.
• Teams localising a campaign: keep the action, change the people, the wardrobe and the skyline.

Getting started

Sign in with Google, upload a reference video, attach your reference images, choose Motion Transfer or Object Swap, write one or two sentences of direction, and generate.

The example gallery on the homepage shows real source clips next to their recasts, so you can see the ceiling before you spend anything. If this is your first run, the workflow guide covers the prep in the order that avoids wasted credits.