OmniHuman-1: Rethinking the Scaling-Up of One-Stage Conditioned Human Animation Models
2025-02-04
OmniHuman-1 is an AI model capable of generating realistic human videos. It supports various visual and audio styles, generating videos at any aspect ratio and body proportion (portrait, half-body, full-body). Realism stems from comprehensive motion, lighting, and texture detail. The model handles diverse inputs, including singing, talking, and various poses, even from cartoons or challenging viewpoints. It leverages audio and video driving signals for precise control. Ethical considerations are addressed, with images and audio sourced from public domains or generated models.
Read more