ByteDance, best known as the parent company of TikTok, has just introduced OmniHuman-1, a brand new AI model that can generate lifelike human videos from just a single image. This innovation goes beyond most existing “deepfake” methods by animating the entire body—down to gestures, head movements, and hand interactions—instead of focusing solely on facial expressions.

OmniHuman-1 can capture natural motion from minimal inputs such as static images, audio signals, or reference videos. The model opens up possibilities for content creators, filmmakers, and marketers to produce digital avatars or fictional videos with relative ease. Previous AI approaches fell short in generating complete body movement, typically animating only the face or upper body.