AI video generator OmniHuman-1 from TikTok owners creates hyper-realistic videos from any photo

China continues to make great strides in AI services. ByteDance, the parent company of TikTok, has unveiled a new AI framework called OmniHuman-1, which can transform a single image and audio recording into a realistic video with detailed facial and body movements.

OmniHuman-1’s technology is based on data blending, which allows it to generate videos with a high degree of realism. The developers trained the model on 19,000 hours of video footage, which significantly improved the quality of facial expressions, gestures and hand movements. The tool supports images of any format – from portrait shots to full-size images.

OmniHuman-1 can be used in the creation of virtual characters, educational videos and other digital projects. So far, it is in the research stage, but in the future it may become a competitor to solutions such as Sora from OpenAI and Veo from Google.