P-Video Replace

Takes a video and up to three reference images, and replaces the subject in the footage with the one in your references. The camera move, the timing and the environment all survive; the person or object changes. That is a different job from Animate, which moves your subject the way a video moves. Here the video keeps its own content and only the subject is exchanged.

Both an image and a video are required. Duration follows the source footage up to fifteen seconds. Unlike its siblings it does not force audio generation, so the original soundtrack question stays open.

Three references rather than one is the detail that matters: it is enough to show the model the subject from more than a single angle, which is what makes a replacement hold up across a shot with any camera movement in it. Give it three genuinely different views rather than three frames of the same pose, and expect the result to be only as good as the match between your subject's proportions and the original's.

Describe what you want in your own words and Kuro writes the prompt for you. If you would rather write it yourself, this is what P-Video Replace responds to.

P-Video Replace keeps a source video's motion, timing, camera and scene and replaces the people on camera with the person in one to three identity images.

The clip is as long as the source and keeps its framing. Spend extra image slots on genuinely different views of the same person — distinct views hold a replacement through camera movement, while three frames of one pose add nothing.

The prompt is optional: the source and the images are enough. Add text only to say how the new person should sit in the scene and what must stay unchanged — it cannot re-stage the shot.

The result is only as good as the match in proportions, so pick a source whose subject is built like the one going in. Use clear, front-facing references with little occlusion, and a source with strong, well-lit motion.

The same advice covers Genjutsu Object Swap. Their pages note what differs.

What it can't do

• Up to three images and one source video; no frames and no audio input.
• The source is held to fifteen seconds; trim it first. The output keeps the source's own frame rate.
• Setting, lighting and camera cannot be changed by the prompt. If those should differ, the source has to differ.
• It keeps the source's audio unless you switch it off. 720p or 1080p.

Use something else when…

• Your character should perform the source's movement in its own setting: Kling Motion Control.
• Only the mouth should change, to new words: Kling Lip Sync.
• No source footage exists: Seedance 2.5.