How to Transfer Motion From Any Video Onto Your Own Character With AI

What Motion Transfer Does

Motion transfer takes the movement out of one video and gives it to your own character. You supply a clip with the moves you want (a walk, a dance, a product turned in the hand) and one or two images of who should perform them. The result is a new video of your character performing the same moves, with the same timing and camera move as the original.

On Genfire, motion transfer lives in Genjudo, our video-to-video studio. Genjudo runs ByteDance's Seedance 2.5 in reference-to-video mode: your clip is the reference for motion, your images are the reference for who appears, and the model generates the shot from both. The performer in the result is your character, built from your image, moving the way the person in the clip moved.

This guide walks through a full run, from choosing a source clip to fixing a weak take. The short overview is on the AI motion transfer page.

What You Need Before You Start

• A source clip with the motion you want. Upload a file, or paste an Instagram reel link and Genjudo uses the reel as the source.
• One or two images. Your character, plus an optional second character, garment or location.
• Genfire credits. Genjudo bills per second of output, and the exact cost appears before you run.

No pose skeleton, depth map or keyframes are needed.

Step 1: Choose a Source Clip the Model Can Read

The source clip decides most of the result: the easier it is to see what the body does in every frame, the closer the transfer.

One clear subject. A single person, well separated from the background, transfers cleanly. A busy street where people cross in front of your subject gives the model two jobs at once.

The whole movement in frame. For a dance, a jump or a walk, keep the full body in shot from head to feet. If the feet leave the frame, the model has to guess the footwork, and guessed footwork tends to slide. For a gesture or a product demo, keep the hands and the object visible throughout.

A steady camera helps. Genjudo copies the camera along with the performer, so a tripod shot or a smooth push-in carries over well. Heavy handheld shake is harder to separate from the body's own movement.

Even light and clear contrast. A performer in dark clothes against a dark wall loses their outline. Soft, front-facing light keeps the limbs readable.

Between 4 and 30 seconds. Genjudo's output runs 4 to 30 seconds, so a source clip in that range maps onto one result. One complete action usually reads better than a long string of unrelated moves.

Step 2: Prepare a Character Image That Matches the Framing

Treat the character image like a casting photo.

Match the framing of the clip. If the clip is full body, use a full-body image. If the clip is a waist-up shot of someone talking with their hands, a waist-up image works. When the image shows less of the body than the clip does, the model has to invent the missing parts.

Use a plain background. A clean backdrop makes it obvious which pixels belong to the character; a busy one can leak into the result.

Show everything you want kept. Face, hair, outfit, shoes: anything visible is something the model can hold on to. A neutral, front-facing pose with the arms slightly away from the body reads most clearly.

Start from an original character. If you don't have one yet, design one in the AI Influencer Studio and bring a full-body image across. A character you created is the simplest cast to work with, because nobody else's likeness is involved.

The optional second image is a second performer for a duet, a garment for your character to wear, or a location for the move to happen in.

Step 3: Load the Clip and Pick a Preset

Open Genjudo and choose Motion transfer. Drop in your source clip or paste a reel link, then add your image or images. Next, pick a preset:

• Exact transfer. The same motion, timing, camera move and framing, performed by your character.
• Dance, beat for beat. The same footwork, arm timing and weight shifts, with the framing held.
• Product in hand. The same handling and camera move with your product, its label, shape and materials matched to your image.
• Two characters. Two images cast in two roles, one in the foreground and one behind.
• Garment on model. Your character performs the move wearing the garment from the second image.
• Motion in a new place. Your character performs the move in the location from the second image.

Each preset fills in a plain instruction saying who plays whom and what must not change, and you can edit it before running.

Step 4: Write the Direction, or Let the Director Write It

Let the AI director write it. Leave the prompt box empty and Genjudo's director watches your clip and writes a shot-by-shot direction: what happens in each shot, how the camera moves, and which of your images takes the place of which person or object. You can read and edit it before you run.

Write it yourself. In the prompt, your source clip is @Video1 and your images are @Image1 and @Image2. A good direction does three things: it says who plays whom, describes the action and the camera in plain words, and names what must stay fixed.

Transfer the motion in @Video1 onto @Image1: the same walk toward camera, the half turn at the end and the slow push-in. Keep the face, jacket and proportions of @Image1 unchanged, and keep the timing exactly as filmed.

A few habits help:

• Pin what must not change. Anything you don't mention is open to change. "Keep the identity, clothing and proportions of @Image1 unchanged" is worth including every time.
• Use concrete words. "Her left hand lifts the bottle to shoulder height" works better than "an elegant gesture".
• Say who is where when there are two images, for example "@Image1 in front, @Image2 behind and to the left".

Step 5: Set the Sound, Length and Takes

Sound has three options:

• Clip's sound keeps the audio from your source clip on the result. Use it when the original song or line of dialogue is the point.
• AI sound has Seedance write a new soundtrack for the result.
• Silent returns the video without audio, for when you plan to add music or a voiceover later in the AI video editor.

Length runs from 4 to 30 seconds. Matching it to your source clip keeps the action, and any audio you kept, lined up.

Takes run from 1 to 4 per run. Generative video varies from take to take, so two or three takes of a difficult move give you a choice. Each take is billed for its own seconds of output.

Step 6: Review the Takes and Fix What's Off

Watch each take at full size before you judge it. When something is wrong, the cause is usually in the inputs.

Motion Transfer or Motion Control?

Genjudo also has a separate Motion Control tab running Kling 3.0 Motion Control, a different model for the same kind of job: a character image moves the way a reference video moves. If a move isn't landing in Motion transfer, run the same clip and image through Motion Control and compare. The Kling V3 page covers that model family, and our Genfire vs Viggle breakdown covers how Genfire compares with another motion tool.

Use Footage and Faces You Have the Right To

Use source clips you shot, licensed or have permission to use, and only cast people who have agreed to it. Your own original characters are the safe default.

What It Costs

Genjudo is billed per second of output from the same Genfire credit balance you use for 80+ image, video and audio models, and the exact cost of a run, across all its takes, is shown before you start it. Credit packs start at $19 on the pricing page.

FAQ

What is AI motion transfer?

AI motion transfer copies the movement from one video onto a different character. In Genjudo, you give Seedance 2.5 a source clip and one or two images, and it generates a new video in which your character performs the same moves with the same timing and camera move.

Can I copy motion from a video I found on Instagram?

Genjudo accepts a pasted Instagram reel link as the source clip. You still need the right to use that footage, so stick to your own reels, licensed clips or creators who have given you permission.

How long can a motion transfer video be?

Each result runs between 4 and 30 seconds. If you need the shot to keep going, Extend continues a clip past its last frame.

Do I have to write a prompt?

No. If you leave the prompt box empty, Genjudo's AI director watches your clip and writes a shot-by-shot direction for you. You can read and edit it before running, or write your own using @Video1, @Image1 and @Image2.

How much does motion transfer cost?

Motion transfer is billed per second of output from your Genfire credit balance, and the exact cost is shown before each run. Credit packs start at $19.