Google's Project Genie adds Genie 4 with real-time world editing
Google has launched a new Project Genie experience for eligible Google AI Ultra subscribers. It opens Genie 4 for the first time and combines world models with Gemini models. Users can create and shape virtual worlds in the browser.
Genie 4 lets users type prompts during a session to edit the world in real time, such as changing weather, adding characters or shifting actions. It also adds real-time native audio for visuals. Genie 3 now supports up to 3 minutes of exploration per generation, up from 60 seconds, plus 360-degree panorama input, world chaining and real-time events triggered by natural language.
The new agent can call multimodal models to build starting visuals and custom music from plain language. Gemini can join a session, watch the screen and offer suggestions. Users can post creations to a community gallery for remixing. All generated video and audio carry a SynthID watermark, and Genie 3 and Genie 4 remain research-stage world models.
Comments
Loading comments…