Google DeepMind’s AI world model Genie 4 can now build interactive worlds that you can explore and reshape as they run. It also generates ambient sounds, music, and dialogue that respond to your actions.
The new world model builds on Genie 3 with improved realism, more natural human movement, and better simulations of real-world physics. It could eventually help with everything from game development to training robots.
How does Genie 4 make AI-generated worlds more realistic?
Genie 4 can build interactive, two minute long worlds from a simple text prompt. Demo videos from Google show how you can change the world midstream by asking for new weather, objects, or camera angles, and the AI adapts without starting over.
Genie 4 relies on Google’s Gemini 3.8 Flash model to keep simulations on track, breaking your goals into steps and checking each frame as it generates. That makes the physics more convincing, with better handling of aerodynamics and collisions. People inside these worlds also move more naturally, with less of the stiffness seen in earlier AI generated worlds.
Native audio is a bigger deal than it sounds, since realistic dialogue was a major turning point for AI video, and something similar may now be headed for interactive worlds too. Google admits these simulations can still drift after a couple of minutes, losing track of details or object permanence.
Google sees potential in game prototyping, filmmaking, and training environments for robots and AI agents, letting developers test how AI systems respond in simulation before deploying them in the real world.
How to access Genie 4?
Genie 4 is rolling out as a research preview to trusted testers first, including filmmakers doing early previsualization work, game developers prototyping ideas, and robotics teams building training scenarios. Google plans to bring it to developers through the Gemini API and AI Studio, though the pricing isn’t confirmed yet.
On the consumer end, access will arrive through Project Genie, starting with Google’s pricier AI Ultra subscribers who pay $100 a month, with AI Pro and AI Plus plans possibly following later. There’s a waitlist open now if you want to get in early, though Google hasn’t shared a firm timeline for wider access just yet.
RELATED COVERAGE:
- Google’s Genie AI will let you create a whole visual world of your imaginations
- Google AI Plus is losing Pro models, and keeping them means a pricier subscription
- You can now build your own Android apps with AI and install them through Google Play
- Google is replacing Gems with Skills, but the upgrade comes with a rather expensive catch
- Gemini’s free tier is getting a major downgrade
