DreaMoving

DreaMoving is a diffusion-based controllable video generation framework designed to produce high-quality customized human videos. It leverages a Video ControlNet for motion control and a Content Guider for identity preservation, enabling the generation of videos where the target identity performs actions driven by posture sequences. The framework can adapt to various stylized diffusion models to produce diverse results, demonstrating proficiency in identity control through face reference images, precise motion manipulation via pose sequences, and comprehensive video appearance control using text prompts.