Google DeepMind unveils Genie 4, a photorealistic world model with real‑time audio and interactive editing

Google DeepMind released Genie 4 in research preview, the latest iteration of its world‑model series. The model can generate photorealistic, interactive 3‑D scenes that last up to two minutes, double the duration of its predecessor. Users can type prompts mid‑simulation to alter weather, camera angle, or add new objects, and the model produces matching real‑time audio—including music, ambient sounds, and dialogue—that syncs with the visual environment. Genie 4 is tightly integrated with Google’s Gemini AI, which plans actions and verifies frames while Genie renders the visuals. In a research demo, a Gemini 4 Argon robot‑planning model broke tasks into steps that Genie 4 simulated in real time, suggesting a pathway for testing robotic scenarios in virtual environments before deployment. Potential commercial applications span gaming, interactive learning, robotics, and rapid‑ideation workflows such as film prototyping and storyboarding.