THE CRUNCH
Google DeepMind has started rolling out Genie 4, its most realistic and interactive world model yet, to eligible Google AI Ultra subscribers as part of a revamped Project Genie experience. The browser-based service generates explorable worlds, and Genie 4 adds the ability to type a new prompt mid-session to edit environments in real time, whether that means shifting the weather, dropping in a character or redirecting the action. It is also the first Genie model to pair its visuals with real-time native audio that adapts as the scene unfolds, from ambient soundscapes to character dialogue.
The update bundles more than the new model. A Project Genie agent turns a natural language description into a full world, orchestrating Google's multimodal models to produce initial visuals and custom background scores, and can bring Gemini models into the session to observe the screen, suggest where to go next or trigger in-world events. A community gallery lets users publish their creations for others to explore and remix, with curated examples including a room-by-room gothic manor mystery built with creative technologist Will aka BLVCKL!GHT and an ant's-eye colony tour made with filmmaker Anand Gandhi.
Genie 3 remains available in the model picker, and Google positions it for cases where you want longer runs or panorama-seeded spaces. New Genie 3 updates include real-time promptable events that connect separate worlds into one continuous journey, and 360-degree panorama grounding to keep locations consistent as you look around.
Access is tiered: Ultra subscribers get in now, while Google AI Pro members can join a waitlist. Google says all video and audio generated in Project Genie carries SynthID watermarking, its tool for marking AI-generated content.
Earlier on this story
Crunchy Tech briefs on this story.


