Google DeepMind unveils Genie 4, a world model that generates explorable 3D worlds with sound
Users can change a scene mid-simulation by typing, and Google says scenes stay coherent for up to two minutes.
Google has announced Genie 4, the next generation of its DeepMind world model, which generates interactive 3D environments in real time that users can move through like a video game. The Verge reports that Google lists four main updates: more realistic worlds, the ability to change an experience in real time with a natural-language prompt, audio generation, and the use of Gemini 3.8 Flash behind the scenes to help run the simulation. Trending Topics says Google calls Genie 4 its first real-time world model to generate audio natively, producing ambient sound, music and interactive dialogue in sync with the picture. Users can switch the weather, move the camera or add people and objects partway through a session. Citing TechCrunch, Trending Topics reports that scenes remain consistent for up to two minutes, twice as long as with Genie 3, which came out a little over a year ago. Google also says the model simulates aerodynamics and collisions, and that characters move more naturally and react to their surroundings. World models like Genie are used for interactive media and as simulated training grounds for AI agents, so longer coherence and directable scenes widen the range of tasks they can support.