Genie 4 and Project Genie agent launch for AI Ultra
- Source
- x.com
- Date
The all-new Project Genie experience is now available. Moving beyond the initial preview, this release launches a full creative ecosystem: — Genie 4 is live: Experience our most realistic, interactive world model yet, now with real-time editing and native audio. Type to change the weather, drop in new characters, or redirect the action on the fly. — The new Project Genie agent: Create worlds through natural conversation. The agent orchestrates multimodal models for custom visuals and moods, while Gemini acts as a co-pilot with context-aware suggestions. — Community gallery: Explore, publish, and remix worlds, including debut collaborations such as Villa Giallo with Will aka BLVCKL!GHT and The Life of Ants with Anand Gandhi. — Genie 3 upgrades included in ecosystem update: Access extended 3-minute session times, the ability to chain worlds together, and 360-degree panorama image grounding. Available now to eligible Google AI Ultra subscribers (waitlist open for Google AI Pro).
Context
Genie 4 has moved from a closed test to something paying customers can use. Google DeepMind announced the model on October 27 as a research preview for trusted testers, and Google AI's launch post now makes it available inside Project Genie to eligible Google AI Ultra subscribers, Google's top consumer AI plan. Subscribers to the cheaper Google AI Pro plan can only join a waitlist.
Genie is a world model: instead of rendering a finished video, it generates an explorable scene in real time and keeps responding as you move through it or change it. With Genie 4, a typed request can change the weather, add characters or redirect events while the scene is running, and the scene now comes with generated sound. DeepMind's announcement explains that a Gemini model works alongside Genie, breaking a request into steps, watching the scene and adjusting its plan; earlier coverage of that design goes into more detail.
The rest of the release is about making worlds easier to build and share. A new Project Genie lets people describe a world in conversation instead of writing one prompt, and a community gallery lets them publish worlds and remix other people's. Genie 3, which Project Genie opened to Ultra subscribers in January, also gets longer sessions of up to three minutes, the ability to chain worlds together and in 360-degree panorama images.
Google's post does not say how long a Genie 4 session can last, and the three-minute figure applies to Genie 3. DeepMind's own announcement also warns that visual details and object permanence can drift over longer sessions or after many edits, so these worlds are best understood as short interactive scenes rather than persistent places.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- grounding — Tying a model's answers to checkable sources — retrieved documents, live data, tool results — instead of letting it answer from memory alone.
Checking sign-in…
Loading comments…





