AI body swap video: replacing the whole person, not just the face

Guide · Updated 21 October 2026 · 6 min read

Why a face-only swap looks wrong when the original performer doesn’t look like you, and how whole-person body swap models fix skin tone, hair and outfit.

The problem with face-only swaps

A face swap replaces the face and leaves everything else. That works when the person in the clip shares your skin tone, hair and build. When they don’t, you get your features on someone else’s body: a different skin colour at the neckline, hair that isn’t yours, hands that don’t match your face.

We saw this first-hand. In early testing, a tester uploaded a photo of a white woman into a trend performed by a Black woman. The face came through, but on the original skin tone. To the tester it looked like neither of them.

How a body swap works

A body swap (also called character swap or motion transfer) doesn’t paste anything. The model reads the motion, camera and timing from the clip, then regenerates the person from your photos: face, skin, hair, body and, if you give it one, the outfit. Some models can change the background too.

Models that do it in 2026

• Higgsfield Genjutsu. Motion transfer that takes a video and up to eight reference images. This is what vtrends runs on, at 720p.
• Viggle. Animates a character or photo with the motion from a video. Strong on stylised and meme content.
• Wan Animate. An open-source model from Alibaba’s Wan family that can animate or replace a character; you can run it yourself with enough GPU.
• Runway Act-Two. Transfers a performance from a driving video to a character, aimed at filmmakers.

What you can control on vtrends

• Your face. The main photo. We crop it to head and shoulders so the model takes your identity, not the clothes you happened to wear in the photo.
• Outfits. Up to three outfit photos. Without one, the original outfit stays.
• Locations. Up to three place photos. Without one, the original background stays.
• Other people. Up to three more faces for duets and group clips; see multiple face swap.

Where body swaps still go wrong

Regenerating a whole person is harder than pasting a face, and the errors are different. Hands can come out soft or with the wrong number of fingers in fast moves. Very loose or layered clothing can flicker. Text on clothing is rarely legible. And because the model reinterprets the body, a clip that depends on a specific prop or costume detail may lose it.

For most trend clips (a walk, a dance, a transition) these are small. For a product shot or anything where a detail must be exact, don’t rely on it.

Ready to try it? Pick a clip from this week’s trends, or read how to face swap a video for the photo tips first.

Questions

Keep reading

• How to face swap a video with AI (and get a result that holds up)What a video face swap actually does, which photo and which clip give the best result, and the shots where every tool still struggles.
• Multiple face swap: putting several people into one videoHow to swap two, three or four people into a duet, couple or group trend, how the model decides who goes where, and how to avoid the wrong face on the wrong person.
• Viggle AI alternatives in 2026What Viggle AI is good at, where people outgrow it, and the alternatives for realistic swaps, real TikTok clips, more control or running it yourself.