HUMAN MOTION GENERATION

Summary from the TLDR AI newsletter

FlowMDM is a new model for generating long, continuous human motion sequences from text descriptions. This first-of-its-kind diffusion-based model uses Blended Positional Encodings for realistic motion creation without extra denoising steps, excelling in accuracy and realism on key datasets.