Google DeepMind unveils Genie 4 interactive world model

On October 27, 2026, Google DeepMind announced Genie 4, an experimental model for interactive simulations. Users can change the scene with text prompts, while the model generates in real time how the world looks and sounds, and how characters move, react, and speak.

Prompts can change the weather, move the camera, or add characters and objects. According to the company, Genie 4 also models physical phenomena, including collisions and aerodynamics.

Genie 4 is currently available to trusted testers in a research preview. Google DeepMind is collecting feedback before expanding access to Project Genie for Google AI Ultra subscribers and to the Gemini API and Google AI Studio for early access partners. The company notes that complex physical interactions, real-time audio accuracy, spatial accuracy, and object permanence remain areas for further work.

Image: JPxG Public domain

Translation editor: Салтанов А.