What changed
Google DeepMind has released Genie 4, the latest iteration of its interactive world model, to eligible Google AI Ultra users. This update introduces real-time audio that adapts as scenes unfold, along with the ability to modify environments during exploration. Genie 4 also pairs with Gemini models to enhance the interactive experience.
What this means in practice
Genie 4 allows users to prompt mid-session changes to their created worlds, enabling real-time edits such as altering weather, introducing characters, or redirecting actions. The model also supports real-time native audio, from ambient soundscapes to character dialogue. Genie 3, which remains available, now includes extended sessions, 360-degree panorama image grounding, and real-time in-world events. Users can explore for up to 3 minutes per generation or link multiple worlds for longer journeys.
Limitations and availability
Genie 4 is available to eligible Google AI Ultra users starting November 18, 2026. Google AI Pro members can join a waitlist. The new Project Genie gallery includes curated and community-built worlds, such as 'Villa Giallo' and 'The Life of Ants,' which users can explore or remix. All generated content includes SynthID watermarking, and model inputs and outputs undergo safety screenings. The release remains in the frontier research space, with ongoing efforts to address challenges like camera latency and real-time audio precision.
