Google opens Genie 4 interactive worlds to AI Ultra subscribers
The Project Genie update adds conversational world-building and a sharing gallery, while Google retains warnings about the technology's limits.
Ledger: AdoptedAI is taken up

Google has made its Genie 4 world model available through an updated Project Genie experience for eligible Google AI Ultra subscribers, moving the system from a research preview into a consumer-facing browser tool. The release adds a conversational creation agent and a community gallery where users can publish, explore and remix interactive worlds.
Project Genie combines Google's world models and Gemini models to let users generate, enter and alter simulated environments in a browser. During a session, Genie 4 can respond to natural-language instructions to change weather, add characters, move the camera or redirect action, while generating adaptive native audio including ambient sound and character dialogue.
As we reported in October, Google had introduced Genie 4 to trusted testers as a real-time world model with native audio. Its model card describes output at 720p and up to 24 frames per second, generated from text, images and user actions; Google says the system is built on Gemini Omni Flash, with Gemini used for reasoning and real-time steering.
The new Project Genie agent is intended to build more complex interactive environments through conversation by orchestrating Google's multimodal models. Google said early testers had asked for greater control over events inside worlds and simpler ways to build and share them. The gallery includes curated and community work such as Living Memory, Dino Field Guide and Rock Paper Scissors, as well as Villa Giallo, made with Will, also known as BLVCKL!GHT, and The Life of Ants, made with filmmaker Anand Gandhi.
Genie 3 remains selectable for longer runs or spaces seeded by panoramic images. The Decoder reported that its update includes sessions lasting up to three minutes, 360-degree panorama support and chainable worlds.
Google AI Pro members can join a waitlist for Project Genie. All video and audio made through the service carry SynthID watermarks, and both inputs and outputs undergo safety screening.
Google still characterises Genie 3 and Genie 4 as frontier research systems. It says camera latency, complex physical interactions, real-time audio precision and long-term scene permanence remain challenges; the model card also warns that visual details and object permanence can drift in longer simulations, after complex edits or while moving through a scene.
- Media Video, Voice and audio, Games
- Company Google