Google DeepMind has revealed Genie 3, a major upgrade to its AI-powered “world model” that can create interactive 3D environments from a simple text prompt. Unlike previous versions, users can now spend a few uninterrupted minutes exploring these AI-generated spaces in real time, and the model remembers what’s in the world even after you look away.
What are world models, and why does it matter?
World models are a type of AI designed to build virtual environments. These are used in gaming, education, training simulations, or for testing AI agents and robots. The user gives a prompt, and the AI creates a fully explorable space, much like a video game level, but without human-made 3D assets. Everything is AI-generated on the fly.
Also Read: How to preload Battlefield 6’s open beta: Dates, maps, and what to expect
This is part of Google’s push into AI-powered environments. DeepMind had earlier released Genie 2 in December 2024, which could generate worlds from static images. That version, however, had limits, users could interact for only 10–20 seconds, and the worlds were short-lived and visually inconsistent.
Genie 3 upgrades: longer playtime, better visuals, smarter memory
Genie 3 now supports continuous interaction for several minutes, a big leap from Genie 2’s brief windows. It also comes with short-term visual memory: if you turn away from an object and look back, it stays the same. Details like wall paint or chalkboard writing won’t randomly disappear or change anymore.
In terms of quality, the generated worlds run at 720p resolution and 24 frames per second, making them more playable and smoother than earlier versions.
Also Read: New Demon Slayer game releasing. Check date, price, platforms, and other details
World events and weather changes, all through prompts
Genie 3 also adds a new feature: promptable world events. You can now use text commands to modify the environment, such as changing the weather or dropping new characters into the scene. This makes interaction feel more dynamic, similar to triggering events in a sandbox game.
You probably won’t get to try it, yet
Don’t expect to play around with Genie 3 just yet. Google is rolling it out as a limited research preview to a small group of academics and creators. The idea is to test the system safely and understand its risks before going public.
There are still some known issues. Users have limited control over interactions, and legible text only appears if it’s provided in the prompt. Google says it’s exploring wider testing access in the future.
The big picture: AI is getting closer to real-time game world creation
While it’s still early, Genie 3 shows how fast AI is evolving in building digital worlds. It brings us closer to a future where AI could generate game levels, virtual classrooms, or entire simulation zones, all from a few lines of text. If you thought AI could only make images and videos, this is the next level.
AQs: Genie 3 by Google DeepMind
Q1. What is Genie 3?
Genie 3 is the latest version of Google DeepMind’s AI “world model” that generates interactive 3D environments based on simple text prompts. It allows users and AI agents to explore these virtual spaces in real time.
Q2. How is Genie 3 different from Genie 2?
Unlike Genie 2, which only allowed 10–20 seconds of interaction, Genie 3 supports several minutes of continuous play. It also remembers the position and details of objects even after you look away and back.
Q3. What are “promptable world events”?
Promptable world events let users change things in the environment using text. For example, you can switch the weather, add characters, or trigger actions in the world, just by writing a prompt.
Q4. What’s the resolution and frame rate of Genie 3 worlds?
Genie 3 generates worlds at 720p resolution and 24 frames per second, offering a smoother and clearer experience than earlier versions.
Q5. Can anyone try Genie 3 right now?
Not yet. Google has released it as a limited research preview for a small group of academics and creators. They’re testing it for safety and performance before a broader release.
Q6. Does Genie 3 generate readable text inside the world?
Only sometimes. Google says text inside worlds is often only readable if it’s included in the initial prompt or description.
Q7. Will Genie 3 come to gamers or creators soon?
Possibly. Google is still exploring wider access, but it’s not available for public use just yet. Future testing could open it up to more users.
