Google launches Project Genie, an AI that builds interactive 3D worlds for video games

Google DeepMind has launched Project Genie, providing public access to its Genie 3 artificial intelligence system for the first time. This AI generates interactive 3D worlds that adapt in real time to user movement and input, making it feel closer to a video game than a static 3D scene generator. While originally developed as a research tool for AI agent training, Project Genie marks its debut outside Google.

Project Genie is available to US users aged 18 or older with a Google AI Ultra subscription priced at 250 dollars per month. It launches with three modes: world sketching, exploration, and remixing. In world sketching, Google’s Nano Banana Pro creates an initial image that users can tweak before Genie 3 generates the interactive world. Users can also define a character, pick a camera view such as first person or isometric, and choose how they want to move through the environment. Prompts can be reuses and adapted from other worlds, although generations are currently limited to 60 seconds, capped at 720p and 24 frames per second.

Google says Genie 3 can simulate physics and interactions, but it is not a traditional game engine. Still, the announcement was followed by a selloff in several gaming related stocks the next day, a reaction widely interpreted as investors pricing in a future where AI generated interactive experiences could disrupt parts of the gaming pipeline, with Take Two down 7.93% to 220.30 dollars, Roblox down 13.17% to 65.76 dollars, and Unity down 24.22% to 29.10 dollars, taking an outsized hit given its role as a widely used game engine and tooling layer rather than a single game publisher.