Upload the source video
An MP4 up to 32 MB. It supplies the performance, the camera move and the environment.
AI Video Character Swap — Replace the Whole Character, Not Just the Face
Upload a source video and up to three identity reference images. The people in the clip are replaced by your characters — face, hair, outfit and build — while the motion, camera and scene stay exactly as filmed.
Add 1–3 identity images
JPG, PNG or WebP up to 8 MB each, added in the order the people appear in the clip.
Render at 720p or 1080p
Choose resolution and target frame rate, add an optional instruction, and generate.
Whole-character replacement, not a face pasted on a stranger
Most tools in this space are face swappers: the original performer stays in frame and a different face is composited onto their head. The hair, the clothes and the body are still someone else's, which is exactly why those results read as a mask rather than as a person. Character Swap works at the level of the whole subject — your identity references define who is in the shot, from face and hair through to outfit and build, and the source video keeps supplying the motion and the scene around them.
That makes it the right tool when the point is your character appearing in footage: an original design fronting a clip you shot, a cast of three swapped into a scene at once, or a stand-in performance re-skinned into the final look. Add the references in the order the people appear on screen, and use the replacement instruction — up to 2,000 characters — when you need to say which reference belongs to which person.
Output length always matches the source clip, so trimming your MP4 before uploading is the simplest way to control both the render time and the per-second cost. Audio from the source is kept by default. If you want to animate a still image rather than replace someone in existing footage, Motion Transfer is the companion tool, and Talking Avatar handles portraits that need to speak.
Character Swap — FAQ
Is this a face swap?
No, and the difference matters. A face swap keeps the original performer and pastes a new face onto their head — the hair, the outfit and the body all stay as they were, which is why the result often looks like a mask. Character Swap replaces the whole person: face, hair, clothing and build all come from your identity reference images, while the source video keeps supplying the motion, camera and scene. If you only want the face changed and everything else preserved, a dedicated face swap is the closer fit.
What do I need to upload?
A source video as an MP4 of 32 MB or less, plus one to three identity reference images as JPG, PNG or WebP of 8 MB or less each. The source video is what supplies the motion and the scene; the references define who appears in it.
Can I swap more than one person in the same clip?
Yes — that is what the three identity slots are for. Add the reference images in the order the people appear in the source video, and use the optional replacement instruction to be explicit, for example “replace the person on the left with reference 1”.
How long is the output, and at what resolution?
The output matches the length of your source video, measured from the uploaded file on the server. Resolution is 720p or 1080p, target frame rate is Original, 24 or 48, and a Turbo toggle trades a little quality for a faster render.
Does the original audio survive the swap?
Yes. Keep source audio is on by default, so the soundtrack from your MP4 carries into the result. There is also an Ignore audio for motion switch if you would rather the audio did not influence the movement.
How is it billed?
Per second of output, at the rate shown on the cost bar before you press Generate. Because the output length equals the source length, you can see the total before committing — trim your MP4 first if you only need part of it.