Google DeepMind, AI research laboratory which is a subsidiary of Alphabet Inc. has introduced Genie 4, its next-generation foundation model engineered to simulate interactive, photorealistic worlds in real time. The model natively incorporates multimodal capabilities, generating synchronized ambient audio, acoustics, and character dialogue alongside video as the environment unfolds.
Genie 4 integrates components from the Gemini model family: Gemini Omni anchors the visual appearance and physical behavior of the environment, while additional Gemini reasoning systems maintain spatial continuity and scene coherence over extended sequences.
In the announcement, they share that the system allows users to navigate scenes using standard keyboard controls while directing simulation events, weather, and characters on the fly using natural language inputs. DeepMind positions the architecture both as a foundation for interactive digital experiences and as an infinite, safe simulation environment for training robotics and autonomous agents.
