Google Opens Access to Project Genie, an AI That Generates Infinite Interactive Worlds

Google has officially taken a major step toward AI-generated virtual worlds. The company has opened access to Project Genie, an experimental prototype capable of generating interactive 3D environments in real time. Powered by Genie 3, the model introduced by Google DeepMind in August 2025, Project Genie allows users to explore living worlds created from nothing more than a text prompt or an image.

There’s a catch, though. For now, access is extremely limited: only Google AI Ultra subscribers in the United States can try it. Still, even in its early form, Project Genie offers a fascinating glimpse into what the future of games, simulations, and digital creativity might look like.

From Genie 3 to Project Genie: Leaving the Research Lab

When Google DeepMind unveiled Genie 3 last August, it was presented strictly as a research model—impressive on paper, but inaccessible to the public. The system demonstrated the ability to generate interactive environments, but testing was limited to internal teams and a handful of selected partners.

Six months later, that research has evolved into something tangible.

Project Genie is now a web-based experiment available through Google Labs, built on top of Genie 3, Nano Banana Pro, and Gemini. The core idea remains unchanged: generate explorable worlds dynamically. What’s new is an interface designed for real users rather than researchers.

The most striking aspect of Project Genie is that nothing exists ahead of time. As you move forward, the AI generates the world in front of you on the fly. There are no preloaded maps, no loading screens, and no fixed environments. Everything—from terrain to objects to interactions—is created in real time, complete with physics simulation and dynamic behavior.

Three Core Features for Creating and Exploring Worlds

Project Genie revolves around three main capabilities, each focused on a different stage of creation and exploration.

1. World Sketching: Creating a World From Scratch

The first feature, World Sketching, lets users generate a world using either a text prompt or an uploaded image. You can define:

• The environment (forest, city, alien planet, abstract world, etc.)
• Your character
• The movement style (walking, flying, driving, or something more unconventional)

This step sets the foundations of the world before you enter it.

With Nano Banana Pro, Google adds extra visual control. Users can preview the generated environment, tweak visual details, refine composition, and choose between first-person or third-person perspectives before starting exploration.

2. World Exploration: Real-Time, Infinite Navigation

Once the world is created, World Exploration allows you to freely move through it. As you advance, the AI generates new areas based on your actions and direction.

The camera dynamically follows your movement, and you can adjust it at any time. There are no boundaries in the traditional sense—only the limits imposed by the current session duration.

This approach creates the illusion of infinite space, even though everything is generated moment by moment.

3. World Remixing: Modifying and Reusing Worlds

The third feature, World Remixing, focuses on iteration and discovery. Users can:

• Start from worlds created by others
• Browse a gallery of shared environments
• Use a random world generator for inspiration

At the end of an exploration session, users can export videos showcasing their journey through the generated world, making Project Genie useful not just for experimentation but also for content creation.

Current Limitations You Should Know About

Google is very clear that Project Genie is still highly experimental, and the limitations are significant.

• Visual fidelity is not photorealistic
• Generated worlds don’t always strictly follow prompts
• Physics simulation can be inconsistent
• Character controls may feel imprecise or slightly laggy

One major restriction stands out: sessions are limited to 60 seconds. Long-form exploration is not yet possible, which greatly limits practical use cases.

Additionally, several advanced concepts mentioned during the original Genie 3 presentation—such as world-changing events during exploration—are not implemented in this prototype.

Google emphasizes that Project Genie exists primarily to study real-world usage, both for research purposes and for future generative content tools.

Extremely Limited Availability—for Now

Access to Project Genie officially begins today, but only under very specific conditions:

• Google AI Ultra subscription
• United States only
• Users must be 18 or older

Google says the rollout will expand gradually, with other countries coming “soon,” but provides no timeline for Europe or France. There’s also no indication that Project Genie will be included in cheaper subscription tiers.

At the moment, the pricing reflects its experimental and premium positioning:

• €139.99 for the first three months
• €274.99 per month afterward
• A US IP address is required to access the tool

This approach follows Google’s usual strategy: test with a small, high-paying audience before broader deployment.

Conclusion:

Project Genie isn’t a game engine, and it’s not ready to replace traditional 3D tools. But it doesn’t need to be. What it offers instead is a preview of a future where virtual worlds are generated instantly, without pre-built maps, assets, or scripts.

Despite its rough edges, time limits, and restricted access, Project Genie shows how AI could transform gaming, simulation, education, and digital storytelling. The technology is still immature—but the direction is clear.

For now, Project Genie is less about practical use and more about exploration and experimentation. If Google continues developing it, this prototype could eventually become the foundation for an entirely new class of interactive experiences.