GWM-1

Runway has announced GWM-1, the first iteration of its General World Model. This release marks a transition from standard AI video generation toward systems that understand and simulate the underlying physics of the real world. For filmmakers and visual effects artists, this means moving away from unpredictable pixel-morphing and toward scenes that respect spatial consistency and object permanence.

What's new

GWM-1 is built to represent the physical world through a unified internal logic rather than just predicting the next frame based on visual patterns. This architecture allows the model to simulate how objects move, how lighting interacts with different surfaces, and how humans navigate 3D environments. Unlike previous iterations of Gen-2 or Gen-3, this model prioritizes the "rules" of a scene, aiming to reduce the visual artifacts common in generative video.

The system focuses on several core capabilities:

• Consistent environmental physics that maintain the integrity of objects during movement.
• Improved spatial reasoning for complex camera paths and character interactions.
• A foundational framework designed to eventually support interactive, real-time world simulation.

You can find more technical details on the official Runway research page (see the provider's announcement).

How it fits your workflow

For directors and cinematographers, GWM-1 addresses the primary hurdle of AI video: the lack of control over physical logic. In a traditional VFX workflow, maintaining a consistent environment across multiple shots requires manual tracking and asset management. GWM-1 aims to automate this by ensuring that if a camera moves through a room, the furniture and lighting remain logically placed and stable. This makes it a viable tool for pre-visualization and even final-pixel background plates where environmental accuracy is non-negotiable.

Editors and animators can use GWM-1 to generate b-roll or environmental transitions that feel grounded in reality. While tools like Sora or Kling focus on high-fidelity visual output, Runway is positioning this model as a simulation engine. This puts it in a unique space between traditional game engines like Unreal Engine and standard diffusion-based video generators. It augments the creative process by providing a digital sandbox where physics are already baked in, reducing the need for manual physics simulations in post-production.

What it costs / how to try it

Runway has integrated GWM-1 into its research roadmap, with capabilities expected to roll out to its creative suite over time. Access typically follows Runway's standard subscription tiers, though specific availability for this model may initially be limited to enterprise users or research partners. Check the Runway website for current plan details and model availability.