Human Motion Generation with Diffusion Models

Human motion modeling is important for many modern graphics applications, which typically require professional skills. In order to remove the skill barriers for laymen, recent motion generation methods can directly generate human motions conditioned on natural languages, speech, and music. However, it remains challenging to achieve diverse and fine-grained motion generation with comprehensive condition signals. Inspired by the success in image generation, recent works attempt to apply diffusion models to motion generation tasks (Motion Diffusion Models) and achieve impressive progress in aspects of realness, fidelity, diversity, and controllability. In this talk, I will introduce several works related to motion diffusion models. In the first part, I will use MotionDiffuse to showcase this research trend and introduce our efforts to push the quality and efficiency of text-driven motion generation based on the diffusion model. In the second part, I will summarize the core ideas of other recently published works, which enhance the generation quality and extend applications of motion diffusion models.