minWM: An open-source framework for real-time interactive video world models

ShengShu AI's minWM converts bidirectional text-to-video and text-and-image-to-video foundation models into interactive world models. It uses Wan2.1-T2V-1.3B and HY1.5-TI2V-8B backbones. The Causal Forcing++ pipeline combines autoregressive diffusion training, causal ODE distillation, and asymmetric DMD for low-latency generation. Code, checkpoints, and inference scripts are available on GitHub under CC BY 4.0.

Visit original source