Skip to main content
MindStudio
Pricing
BlogAbout
My Workspace
Project Genie GoogleGoogle world modelGenie AI game generator

Google Project Genie: What Is the AI World Model and How to Try It

Google's Project Genie turns text prompts into playable 3D worlds generated in real time. Here's what it is and how to access it.

Edited by Luis Chavez-Mattos, Director of Product RSS
Google Project Genie: What Is the AI World Model and How to Try It

What is Project Genie?

Project Genie is an experimental AI system from Google that generates interactive 3D environments in real time, rather than producing a fixed video or image. Instead of rendering a pre-built game world, it invents the environment as you move through it, generating what’s ahead, behind, and to the sides on the fly. Google calls it a simulation rather than a game, and that distinction matters: nothing in these worlds exists until the model generates it in response to your movement.

TL;DR

  • Project Genie generates 3D worlds on the fly, creating terrain, objects, and characters as the user moves rather than loading a pre-made environment.
  • It lives inside Google Labs, at labs.google/fx/projectgenie, which is where Google tests early, unpolished AI experiments before wider release.
  • Access currently requires a Google AI Ultra subscription, which starts at $100, and is limited to certain regions including the US.
  • Early demos range from dinosaur safaris to horror-style environments and an “infinite doom scroll” video feed, showing the breadth of what a world model can produce rather than a single polished use case.
  • The visuals are rough and choppy right now, closer to an early tech demo than a finished game, which is normal for a first public version of a new kind of AI model.
  • The likely long-term uses involve simulation, such as practicing social situations, training scenarios, or experiential learning, though Google hasn’t locked in a specific product direction yet.

Live class · Wed, Dec 2

$40,000+/mo

in SaaS spend the Remy team cut by building its own apps with AI.

We'll show you how.

Bring an idea.
Build it live.
Leave certified.

How does Project Genie actually work?

Project Genie is built on the idea of a “world model,” an AI system trained to predict and generate a consistent environment rather than just the next frame of a video or the next word in a sentence. When a user moves forward, turns, or interacts with an object, the model generates new visual content consistent with the physics and logic of the scene so far. It’s drawing on patterns from how real environments, objects, and movement behave, then synthesizing new 3D space in response to user input.

This is different from standard AI video generation, which produces a clip from a prompt and stops there. In Project Genie, the environment responds continuously to navigation. Walk toward a building and the model has to generate what that building looks like up close. Turn around and it generates what’s now behind you, even though that space didn’t technically exist a moment earlier. The system is improvising the world in real time, scene by scene, based on your actions.

What do the early demos actually show?

Public demos accessible through Google Labs include a range of different simulated scenarios. One lets users walk through a dinosaur safari, complete with AI-voiced dinosaurs narrating the experience and bizarre, glitchy transformations as the model struggles to keep objects consistent. Another resembles a horror game environment. There’s also a demo built around an “infinite doom scroll,” simulating an endless swipeable video feed generated live, and a social simulation themed around a holiday dinner where users pick a virtual companion.

None of these are polished products. Movement is choppy, objects warp unexpectedly, and the fidelity is nowhere near a traditional video game, let alone something like a AAA commercial release. That’s expected for a first public version of a genuinely new category of AI system. The point of these demos isn’t to compete with game studios, it’s to show what becomes possible when an AI model can generate explorable space instead of static output.

Why does a world model matter if the demos look rough?

The rough visuals are the least important part of this release. What’s notable is the underlying capability: an AI system that simulates physics, movement, and environment consistency well enough to let a person walk around inside a world it invented moments earlier. That’s a meaningfully different capability from generating an image or a video clip, because it requires the model to track spatial consistency, object permanence, and interactive responses over time.

Early, clunky versions of transformative AI tools are common. Large language models looked far less capable in their first public releases than they do now. A rough-looking demo today doesn’t predict where the technology ends up in a year or two. The relevant question isn’t “does this look like a finished game,” it’s “can an AI model maintain a coherent, interactive 3D space in response to user input,” and Project Genie is an early, public answer of “yes, partially.”

What could world models like Project Genie be used for?

Build enterprise applications with AI that are
Secure.Role: Adjuster✕ Export PHI
Regulation-compliant.✓ HIPAA✓ SOC 2
Auditable.09:14M. Chen approved claim #4471
Scalable.12,480 users online

Nobody, including Google, has fully defined the end use case yet. But a few plausible directions follow from what the technology does well: generating consistent, responsive environments on demand.

One likely application is training and simulation. A world model that can generate realistic scenarios could let people practice situations before encountering them for real, such as social interactions, workplace scenarios, or physical environments that would be expensive or risky to replicate otherwise. Education is another candidate: instead of reading about a business scenario or historical event, a learner could walk through a simulated version of it and apply knowledge interactively rather than passively.

Entertainment is an obvious near-term use too, given that the current demos already lean toward game-like and video-feed-like experiences. It’s also plausible that, similar to how AI models for text and image generation eventually became tools regular people could build with, future versions of world-model technology could let users construct their own interactive simulations from a prompt, rather than only exploring pre-made ones.

Is Project Genie worth trying right now?

For most people, not yet in a practical sense, but it’s worth knowing it exists. Access currently requires a Google AI Ultra subscription, which starts at $100, and is limited to specific regions. That’s a steep cost for an experimental demo with choppy visuals and no clear everyday use case. If your interest is purely practical, there’s no workflow here that replaces an existing tool today.

If you’re interested in where AI is heading structurally, rather than what it can do for you this week, it’s one of the more interesting public releases of the year. Few companies have put a real-time generative world model in front of regular users at all, and seeing the rough edges firsthand, like inconsistent object physics or warped scenery when you turn around, gives a clearer sense of the technology’s current limits than reading about it secondhand.

How do I access Project Genie?

Project Genie is available through Google Labs at labs.google/fx/projectgenie. Google Labs is the company’s testing ground for early AI experiments, so features here are generally less stable and less polished than tools released to the general public. Access currently requires a Google AI Ultra subscription, which starts at $100 a month, and is restricted to certain regions, including the United States. Availability is likely to expand over time, as is typical for Labs experiments that Google decides to invest further in.

Frequently Asked Questions

What is a world model in AI?

A world model is an AI system designed to generate and maintain a consistent, explorable environment, predicting what should appear in a space as a user moves through or interacts with it, rather than just generating a single static image or video clip.

Is Project Genie a video game?

Not exactly. It looks and behaves somewhat like a game, with movement and interaction, but nothing in the environment is pre-built. Every part of the scene is generated by the AI model in real time as the user explores it.

How much does Project Genie cost to access?

Remy AI Engineer Certification

Companies want to build their own software.

Few people can build it right.

Become the one they trust.

  • Access
  • Audit trails
  • Security
  • Compliance

It currently requires a Google AI Ultra subscription, which starts at $100, and access is limited to certain regions, including the United States, through Google Labs.

Why do the Project Genie demos look choppy and glitchy?

This is an early public version of a new kind of AI model. Generating a coherent 3D environment in real time is a hard technical problem, and current visual fidelity reflects the early stage of the technology rather than its ceiling.

What is Project Genie actually useful for right now?

There’s no clearly defined everyday use case yet. Plausible future directions include training simulations, education, and interactive entertainment, but Google has presented it as an experimental demo rather than a finished product.

Editorial standards

Presented by MindStudio

No spam. Unsubscribe anytime.