Google Opens Project Genie to AI Ultra Subscribers in the U.S.

Google is rolling out Project Genie, an experimental research prototype that lets users create, explore and remix interactive worlds, to Google AI Ultra subscribers in the U.S. aged 18 and over, with expansion to more territories planned "in due course."

The prototype is a web app powered by three Google systems: Genie 3, the general-purpose world model the company previewed in August; Nano Banana Pro, its image generation model; and Gemini. Google announced the rollout in a blog post describing the project as the next step after months of work with trusted testers, who the company says "were able to create an impressive range of fascinating worlds and experiences, and uncovered entirely new ways to use it."

The launch matters because world models sit at the center of Google DeepMind's stated path to artificial general intelligence. A world model simulates the dynamics of an environment, predicting how it evolves and how actions affect it. DeepMind has historically built agents for narrow, fixed environments like Chess and Go, but the company argues that "building AGI requires systems that navigate the diversity of the real world." Genie 3 is its answer: a model that generates the environment in real time as a user moves through it, rather than presenting a static, pre-built 3D scene.

That distinction defines the product. Unlike explorable experiences built from static 3D snapshots, Genie 3 "generates the path ahead in real time as you move and interact with the world," according to Google. The company says the model simulates physics and interactions for dynamic worlds, and that its "breakthrough consistency enables the simulation of any real-world scenario — from robotics and modelling animation and fiction, to exploring locations and historical settings."

Three core capabilities

Project Genie organizes the experience around three functions.

World sketching. Users prompt with text and generated or uploaded images to create a living, expanding environment. They define a character, a world and a mode of exploration — "from walking to riding, flying to driving, and anything beyond," in Google's phrasing. For finer control, Google has integrated World Sketching with Nano Banana Pro, which lets users preview the world and modify the image before entering it. Users can also set the camera perspective, choosing first-person or third-person view before they start.

World exploration. The generated world is a navigable environment. As the user moves, Project Genie generates the path ahead in real time based on their actions, with adjustable camera controls during traversal.

World remixing. Users can remix existing worlds into new interpretations by building on their prompts, browse curated worlds in a gallery, or use a randomizer icon for inspiration. Finished explorations can be exported as downloadable videos.

Known limitations

Google frames the prototype as a research effort inside Google Labs, and it is candid about the model's current constraints. Genie 3 is an early research model, and the company lists three known areas for improvement: generated worlds "might not look completely true-to-life or always adhere closely to prompts or images, or real-world physics"; characters "can sometimes be less controllable, or experience higher latency in control"; and generations are limited to 60 seconds.

Some capabilities Google announced for Genie 3 in August are also missing from this first release. Promptable events — features that change the world as you explore it — are not yet included in the prototype.

Why Google is shipping it now

The rollout follows testing with trusted testers "from across industries and domains," and Google says the goal is to understand demand and usage patterns before scaling further. The company says it wants to share the prototype with users of its most advanced AI subscription tier "to better understand how people will use world models in many areas of both AI research and generative media."

The choice of distribution channel is itself a signal. Google AI Ultra is the company's top-tier subscription, which means the prototype reaches a small, paying, presumably tolerant audience — a pattern consistent with how Google Labs typically tests early products before wider releases.

Google's stated goal is eventual broad access. "In time, our goal is to make these experiences and technology accessible to more users," the company says.

For now, Project Genie doubles as both a product experiment and a research instrument: every world U.S. Ultra subscribers sketch, explore and remix will feed Google's understanding of where generative world models fit between AI research, robotics simulation and generative media.