Face swap (also called face replacement) is the process of putting the face from a photo onto a person in a video, frame by frame, while everything else — the body, the motion, the background, the light — stays exactly as it was. What once took a studio, a stand-in and hours of compositing now takes two uploads: a clip and a portrait.
How it works
The model finds the face in every frame, reads its angle and expression, and renders the face from your photo in the same pose, relit to match the scene. Because each frame is mapped back to the original performance, the new face turns, speaks and reacts with the timing of the person in the footage. When the camera moves, the swap follows; when the light changes, the new face changes with it.
What it's best at
- Talking-head videos — interviews, monologues and ads where the face carries the whole shot.
- Dance and action clips — fast motion, where the swap has to hold through every turn of the head.
- Scenes with strong lighting — the new face picks up the scene's own light and color instead of looking pasted on.
- Recurring characters — one face across a whole series of clips, without reshooting.
What a good input includes
The cleanest swaps start from two clear sources. In the video, the face should be front-facing, evenly lit and large in frame, with only one person in the shot — profile angles and crowded scenes give the model less to work with. In the photo, use a sharp, high-resolution headshot with nothing covering the face: no sunglasses, no masks, no hands. And only use material you have the rights to — a face is someone's likeness.