Question 1

What can I make with P-Video Replace?

Accepted Answer

One shot, many presenters. Film a spot once and issue it with a different face for each market. Put a creator into a UGC clip you already own. Test a casting choice against footage you have shot. Replace a stand-in with the person who was meant to be in frame. The performance is already in the source video, so you only change who is giving it.

Question 2

What inputs does P-Video Replace need?

Accepted Answer

A source video and one to three photos of a person. The video supplies the performance, the scene, and the camera. The photos supply the person. An optional instruction tells the model how to place that person in the scene. Write something like "keep the black jacket" or "the person is seated at the desk". When more than one person is on camera, the instruction is the only way to say which one to change.

Question 3

What makes a good set of reference photos?

Accepted Answer

Sharp, well lit, and front facing. One photo works. Three work better, because the model has more of the face to go on. A steep angle, sunglasses, or anything that hides the face costs you the likeness, so leave the profile shot out. Show the person at the framing the source video uses, so a full body photo for a full body clip.

Question 4

What kind of source footage works best?

Accepted Answer

Steady, well lit footage of normal movement. Walking, a talking head, a presenter, and a product demo are what the model is tuned for. Very fast action, a subject that keeps passing behind things, and a hard camera move all make the result less consistent. Frame the source the way you want the output framed, because the output keeps the aspect ratio of the source.

Question 5

What do the resolution, FPS, and Turbo settings change?

Accepted Answer

Resolution sets the output size. 720p is the tier for testing photos and instructions. 1080p holds more detail in the face and is the tier for delivery. FPS follows the source by default, and you can force 24 or 48 instead. Turbo shortens the run and gives up a little quality, which is a fair trade while you iterate. Both tiers take the same inputs, so lock the take at 720p and re-run it at 1080p without changing anything else.

Question 6

How long can the source video be?

Accepted Answer

Thirty seconds, as an MP4 or a MOV of up to 200 MB. The output is as long as the source. Visual Sandbox checks the length before the run starts and tells you plainly when a clip runs long. For a longer piece, cut it into parts, run each part with the same photos, and join them in an editor.

Question 7

Does the output keep the audio from the source video?

Accepted Answer

Yes. The source audio carries into the output unchanged and stays aligned with the picture. That matters here, because the voice in the clip is still the voice of the original performer. Replace the track in post when the new person should be heard as well as seen.

Question 8

How is P-Video Replace different from P-Video Animate and Kling 3.0 Motion Control?

Accepted Answer

Ask which file defines the world. P-Video Replace lets the clip define it, so the scene, the camera, the lighting, and the performance survive and only the person changes. P-Video Animate lets the still define it. The look, the lighting, and the wardrobe come from your image, and only the motion comes from the clip. Kling 3.0 Motion Control does the same job as Animate. It adds an orientation control. That control decides whether the character faces as the photo does or as the driving video does.

Question 9

Do I need the right to use the person's face?

Accepted Answer

Yes. You are responsible for having permission to use the likeness of every person in the photos and in the source video. Get that permission in writing before you run a real person through this model, and check the rules of the platform you publish on. Do not use it to make somebody appear to say or do something they did not. Pruna AI's terms allow commercial use of the output, but the consent is yours to obtain.

Question 10

How much does a P-Video Replace run cost?

Accepted Answer

It is billed per second of the delivered video, at the rate the pricing table above shows for the resolution you pick. The request never states how long the source is. Visual Sandbox holds credit for thirty seconds before the run. It then charges only for the seconds the model delivers. A ten second clip costs a third of what was held. There is no subscription and no monthly minimum.

Tier	Price
720p	$0.04/s
1080p	$0.08/s

Create with P-Video Replace

What is P-Video Replace AI person swap?

P-Video Replace pricing