Google has made Genie 4, its latest interactive world model, available to eligible Google AI Ultra subscribers through a revamped Project Genie experience, according to a November 18 post on Google's blog by product leads from Google DeepMind and Google Labs. Google AI Pro members can join a waitlist.
Genie 4 generates explorable worlds from prompts, and Google says users can now prompt mid-session to change the world they created, for example by shifting the weather, adding a character or redirecting the action. Google says this is the first time Genie pairs its visuals with real-time native audio that adapts as the scene unfolds.
The update also improves Genie 3: users can explore for up to three minutes per generation, up from 60 seconds, and Google added world chaining, grounding worlds in 360-degree panorama images and real-time in-world events. A new Project Genie agent builds worlds through conversation by orchestrating Google's multimodal models for visuals and music, and can bring Gemini into a session to watch the screen and suggest next steps. A community gallery lets users share worlds.
Google describes Genie 3 and Genie 4 as world models still in the frontier research space, and lists camera latency, intricate physical interactions, real-time audio precision and long-term scene permanence as open challenges. All generated video and audio is watermarked with SynthID, and inputs and outputs go through safety screening.
World models like Genie are being explored for games, simulation and training environments for AI agents. The Ultra rollout gives paying users hands-on access to Google's most capable version, while the stated limits on physics and scene persistence indicate it is not yet a substitute for a game engine or a production simulator.