Particle.news
Download on the App Store

Runway Launches GWM-1 World Models, Adds Native Audio to Gen 4.5

The move signals a push beyond video into physics-aware models for robotics, simulation, avatars.

Overview

  • GWM-1 debuts as a family of models—Worlds, Robotics, Avatars—built on frame-by-frame prediction to model physics and temporal behavior.
  • GWM-Worlds generates interactive scenes with geometry, physics, and lighting consistency for uses like gaming and training agents.
  • GWM-Robotics will be offered through an SDK, with Runway in active discussions with robotics firms and enterprises about deployments.
  • Gen 4.5 now supports native audio and long-form, multi-shot creation, enabling one-minute videos with character consistency and dialogue.
  • Runway disclosed a CoreWeave deal to use Nvidia GB300 NVL72 racks for future training and inference, as it positions GWM-1 against rivals such as Google’s Genie-3.