Face-swap deepfake

A deepfake that replaces one person's face with another's in a video or photo, while keeping the original body and background.

The footage looks like the original scene, but the face belongs to someone else, such as a person the fraudster wants to impersonate.

A deepfake subtype that superimposes a target's facial features and expressions onto an existing video or image using generative models, most commonly built on GANs trained on footage of the target.

A face-swap typically works in three steps:

1. The model learns the target's face from photos or video.
2. It maps the target's face onto the face of a person in the source footage.
3. It blends the result so the lighting, movement, and expressions match the scene.

Some systems do this in real time, which lets a fraudster appear as someone else on a live video call.

Face-swap deepfakes sit next to terms that are easy to mix up:

• Deepfake is the broader category of AI-generated media.
• Synthetic avatar deepfake renders an entirely fabricated person, while a face-swap changes the face on real footage.
• Morphing attack blends two faces into one still image, while a face-swap places a face onto a video.

For more, see The 3 Most Common Types of Facial Deepfakes Explained.