ByteDance has launched OmniHuman-1, an AI tool capable of generating realistic videos from a single photo. This technology performs well under weak signal inputs (such as audio) and supports image inputs of any ratio (portrait, half-body, or full-body), producing high-quality realistic videos. OmniHuman-1 can also recreate historical figures, such as videos of Einstein speaking in front of a blackboard, showcasing natural facial expressions and gestures. Freddy Tran Nager, a professor at the Annenberg School for Communication at the University of Southern California, described the technology as "very impressive," especially when used on small screens like smartphones.

PANews reported on February 6 that according to Forbes, TikTok's parent company ByteDance launched the AI tool OmniHuman-1 , which can generate highly realistic videos from a single photo, including dynamic images of people talking, gesturing, playing musical instruments, etc. The research paper points out that the technology performs well under weak signal input (such as audio), supports image input of any ratio (headshot, half-body or full-body), and generates high-quality realistic videos. On the display page, OmniHuman-1 also reproduces historical figures, such as videos of Einstein speaking in front of a blackboard, showing natural facial expressions and gestures. Freddy Tran Nager, a professor at the University of Southern California's School of Communication, called the technology "very impressive" and said the effect is particularly realistic if used on a small screen such as a mobile phone.

ByteDance Launches OmniHuman-1: An AI Tool that Generates Realistic Videos from a Single Photo