Genjutsu is a Japanese word. Written 幻術, it splits in two. 幻 is the phantom, the thing that is not really there. 術 is the technique. Together they mean an illusion technique: you change what a person sees, and the world underneath stays where it was.
A lot of people met the word in Naruto. A genjutsu does not pick the target up and move them. It replaces the picture. The body keeps doing what it was doing. The scene in front of them is a lie.
Higgsfield borrowed that name for a video tool, and the tool is the clip people are passing around. You give it a video. You give it a character, a product, or a set of clothes. The steps, the camera, and the timing stay. The person performing them changes. That is the illusion. The dance was filmed. The dancer you are looking at was not the one in the original frame.
What the tool actually does
Higgsfield Genjutsu has two moves, and they are easy to mix up because both of them start with one clip.
Motion transfer keeps the performance and rebuilds the shot around your references. The body learns the dance. The room, the clothes, and the face can all be new.
Object swap changes one thing you name and leaves the rest of the footage alone. A person, an outfit, a product, a place. The cut, the camera, and everyone else stay.
The viral version is the feeling of the second one. Same video. Different lead. People are doing it with trending clips because the result looks like a reshoot that never happened.
The reference clip can be 4 to 30 seconds. The tool will take a pile of stills for the new character, the clothes, and the place. A prompt is optional. The references do most of the talking.
We ran it on Seedance 2.5
Seedance 2.5 can do this job. It does not have a button named Genjutsu. You hand it the clip as the performance, and a still as the new person, and you tell it which is which.
This is the reference. A night dance in front of the Eiffel Tower. The man in the black hat leads. Two people stay behind him.
This is the same 13 seconds on Seedance 2.5. The plaza, the tower, the steps, and the people behind him are still there. The lead is a chef, baguette in hand, yellow socks, teal sneakers.
That is the illusion, made with a model you can already call. The filmed motion stayed. The person in front did not.
The trick has two parts
The first part is the edit. You keep the video and replace what should change. Add a character sheet, the other references of that character, and references of the product, or of anything else you want swapped. That is one part. A prompt alone is not the trick.
The second part is the rejection. Seedance 2.5 will refuse the request when a real character is in it. Seedance has a workaround for that. We know it. If you want that on Chromatic, leave your email below.
What you can replicate
Seedance 2.5 takes the performance clip and up to 30 reference images in the same request. Each image can change something in the frame: the character, the product, the clothes, the place, or anything else you want swapped. The steps, the camera, and the timing stay.
One dance can be played by a different person. A product you sell can sit in a shot you already filmed. A brand can turn one hero film into another spokesperson, another product, another market. A creator can put themselves, or a character they keep, into a take they like.
The reference video holds the motion. The images hold what changes.
A Higgsfield Genjutsu alternative
People searching for another way to do this are looking for the same job: one clip, the motion kept, and something in the frame replaced. Seedance 2.5 is that alternative. The video is the performance. Up to 30 reference images decide what changes. A character, a product, clothes, a place, or anything else.
If you want it here
FAQ
What does Genjutsu mean?
Genjutsu (幻術) is Japanese for an illusion technique. 幻 means phantom or illusion. 術 means technique. The word names a trick that changes what someone sees while the real scene stays in place.
What is Higgsfield Genjutsu?
Higgsfield Genjutsu is a video tool that recasts an existing clip. You upload a reference video and images of a character, product, or place. Motion transfer rebuilds the shot around those images. Object swap changes one element and keeps the rest of the footage.
What is a Higgsfield Genjutsu alternative?
Seedance 2.5. Give it the clip for the motion and up to 30 reference images for whatever should change: a person, a product, clothes, or a place. The performance stays. The picture is the part that changes.
Can Seedance 2.5 do the same thing?
Yes. The clip carries the motion. You can add up to 30 reference images beside it, and those images carry whatever should change: a character, a product, clothes, a place, or anything else. The recast comes back with the same performance.
What can you replicate?
Anything in the frame, as long as the performance stays. With up to 30 reference images you are not limited to one swap. A person can step into a dance they did not film. A product, an outfit, or a room can replace what was there.
A brand uses that on a film they already cut: one hero spot becomes another spokesperson, another product, another market. A creator uses it to put themselves, or a character they keep, into a take they like. The video holds the motion. The images hold the change.
