Google DeepMind is releasing Genie 4, its most advanced world model for creating interactive 3D spaces. The Project Genie update makes the model available to Google AI Ultra subscribers; previously, it was accessible only in a research version. Users can change worlds with text prompts in real time—for example, adjusting the weather, characters and plot. Genie 4 also generates audio that changes with the events in a scene. Genie 3 now supports sessions of up to three minutes, 360-degree panoramas and the ability to link worlds. Project Genie also adds an AI agent that creates worlds from voice descriptions, and a gallery where users can publish and adapt work made by the community.