Google DeepMind Announces Genie 4 World Model

Google DeepMind announced Genie 4, a new world model designed for real-time interactive simulation. Authors Raia Hadsell, Diego Rivas, and Hang Qi detailed how the model incorporates capabilities from the broader Gemini family. Gemini Omni anchors the simulations in real-world appearance and behavior, while other Gemini models supply complex reasoning to help maintain a simulation's coherence and continuity over time.

Genie 4 operates as a natively multimodal model, generating audio-like music, ambient noises, and interactive dialogue simultaneously with visual environments. Users can steer and modify a simulation in real-time using natural language text prompts and standard keyboard controls for navigation and camera placement. Prompts can change weather conditions, move the camera, or introduce new people and objects. Between user inputs, Gemini monitors and steers the scene to keep the simulated world progressing cohesively.

According to Google DeepMind, the model simulates real-world physics scenarios, including aerodynamics and collisions, alongside interactive characters that move naturally and respond organically to their environments. Genie 4 models how environments sound, producing ambient audio like wind in trees or waves at a beach, alongside real-time interactive human dialogue that adapts to user inputs and in-world events. Google DeepMind notes that simulating the world provides training environments for robots and agents while unlocking new interactive experiences.

Google DeepMind has begun rolling out Genie 4 as a research preview for trusted testers before a broader release to consumers and developers. Access is starting with Google AI Ultra subscribers in Project Genie and early access partners of the Gemini API and Google AI Studio, alongside an open waitlist for Project Genie access.