What happened
Google DeepMind has made Genie 4, its most capable model yet for producing interactive 3D environments, available to people who subscribe to Google AI Ultra, The Decoder reported. The release forms part of a broad update to Project Genie. Until now, Genie 4 had only been offered as a research preview. Users can alter a generated world with text prompts while it runs, adjusting elements such as the weather, the characters or the plot.
According to the report, Genie 4 also produces its own sound in real time, a first for the model, with audio that shifts to match what is happening in the scene. The earlier Genie 3 model is receiving upgrades as well, including sessions lasting as long as three minutes, support for 360-degree panoramas and the ability to link worlds together in a chain.
Project Genie is also adding an AI agent that assembles worlds from spoken descriptions, plus a community gallery where users can publish their creations and remix those made by others. Every piece of generated content is marked with SynthID watermarks and passes through safety checks. Members of the Google AI Pro tier can sign up for a waitlist.
Context
Genie 4 follows Genie 3, which Google DeepMind is extending rather than retiring, with longer sessions and new formats. Before this update, Genie 4 was available only as a research preview. Access to Project Genie is now split by subscription tier: Google AI Ultra subscribers get the new model, while Google AI Pro members can only join a waitlist. All output carries SynthID watermarks and passes through safety checks.
The briefing
The week in tech, in one email.
The stories worth your time, every week. Unsubscribe anytime.
