Google DeepMind’s new AI model Genie 3 can create 3D interactive worlds in real time

Google DeepMind has introduced Genie 3, its latest AI model capable of generating interactive 3D worlds from simple text inputs. The environments render in real-time at 24 FPS, maintaining visual consistency in 720p for several minutes—a major leap from Genie 2’s 10–20 seconds.

The model enables users to explore dynamic, persistent spaces. If you leave a location and return, it still looks the same—an advancement in AI memory and spatial awareness.

According to DeepMind, world models like Genie are vital for training AI agents in simulated environments. They help agents predict environmental changes and consequences of their actions, making them a foundational step toward Artificial General Intelligence (AGI).

Other world modeling tools like Veo 2 and Veo 3 are also gaining attention for their ability to understand and recreate realistic physics in video generation.

Currently, Genie 3 is not publicly available and will be rolled out to a select group of creators for testing.

Source: The Hindu

Voice of Osiz

Genie 3 marks a groundbreaking step toward more immersive and persistent AI-generated environments. We see this as a strong indicator of how text-to-world models are reshaping simulation, gaming, and training systems. Our take? These advancements accelerate the path to AGI and open up real opportunities for building dynamic user experiences. It's a bold glimpse into the future of AI interactivity.