Runway releases its first world model, adds native audio to latest video model

Runway has launched its first world model, GWM-1, marking a significant step in AI image and video generation. This model operates via frame-by-frame prediction, simulating the physical world with an understanding of physics and temporal dynamics. Unlike prior models, GWM-1 is described as more general than competitors like Google’s Genie-3, capable of creating simulations useful for training AI agents across various domains such as robotics and life sciences. Runway has introduced specialized versions of this model—GWM-Worlds, GWM-Robotics, and GWM-Avatars—each tailored for interactive world creation, synthetic data generation for robotics under varied conditions, and realistic human behavior simulation, respectively. In addition to the world model, Runway updated its foundational video generation model to Gen 4.5, which now supports native audio, long-form multi-shot video generation, character consistency, and complex scene composition with dialogue and background sound. This update brings the company closer to competitors offering