What happened
Google has made Genie 4, its latest model for generating interactive worlds, available to eligible Google AI Ultra subscribers through Project Genie, an experimental world-building tool that runs in the browser, TNW | Artificial-intelligence reported. The company announced the rollout on Wednesday. Members of the AI Pro tier can sign up for a waitlist in the meantime.
Genie 4 lets users alter a world while moving through it. A prompt typed during a session can change the weather, introduce a character or steer events in a new direction. The model also produces audio for the first time, ranging from ambient sound to spoken lines from characters. The release bundles an AI agent that assembles worlds through conversation, plus a gallery where users can publish and remix one another's creations.
Google said both Genie 4 and the older Genie 3 remain research systems. Details can drift over long sessions, and legible text and complex physics are still difficult. Genie 3 stays on offer for longer runs and now supports up to three minutes per generation, compared with 60 seconds previously. All content produced in Project Genie is marked with SynthID, the invisible watermark Google applies to AI-generated material.
Context
Testers saw Genie 4 from Google DeepMind in October, and some Project Genie users can already roam simulated Street View cities. DeepMind argues such models could help train robots and AI agents before real-world use, while competing lab Odyssey claims its Odyssey-3 model can already control robots and cars. Major game studios are still hesitant, with Take-Two stating that generative AI played no part in GTA 6.
The briefing
The week in tech, in one email.
The stories worth your time, every week. Unsubscribe anytime.
