Google expands access to Genie 4 world-building AI
The model adds audio and lets users alter interactive environments during a session, while Google cautions that details can still drift.

Google is giving eligible Google AI Ultra subscribers access to Genie 4 through its browser-based Project Genie prototype, the company said. The rollout includes a conversational creation agent and a gallery for sharing worlds. AI Pro subscribers can register for a waitlist.
According to Google, users can explore generated environments and modify them while a session is underway. Prompts can introduce characters, change weather or steer events. Genie 4 also introduces audio, including ambient sounds and dialogue between characters.
The creation agent builds environments through conversation, while the gallery allows people to share their creations and remix those made by others. Google DeepMind initially demonstrated Genie 4 to testers in October. Some Project Genie users can already explore simulated cities generated from Street View.
Genie 3 remains an option for longer sessions, with each generation now supporting up to three minutes rather than 60 seconds. Google describes both models as research systems: visual details may drift during extended use, while legible text and complex physics remain difficult. Google says all Project Genie creations carry its invisible SynthID watermark identifying AI content.




