Google introduces Project Genie, a new AI tool that allows users to create interactive three-dimensional environments with simple text instructions.
Google has launched an experimental research prototype called Project Genie, which allows users to generate virtual worlds via prompts.
Genie 3
The tool runs on Genie 3, a so-called world model that Google presented last year. This generates virtual environments in real time with a resolution of up to 1280×720 pixels and a maximum of 24 frames per second. Users can navigate a generated world for up to 60 seconds per session, although the underlying technology could handle longer interactions.
Users describe both the environment and the character moving through the world via two input fields. Camera position and visual style are also adjustable. Project Genie first creates a visual sketch using Google’s Nano Banana Pro-model, after which the environment can be further refined with additional prompts or preset templates.
Building worlds with text and sketches
Interactions can be saved as video, which makes Project Genie interesting for content creation and prototyping. Google also sees potential for AI training, as the generated virtual environments can serve as synthetic visual data. The company is investigating whether Project Genie will be available to developers via an API in the future.
Project Genie is initially available in the US for subscribers of Google AI Ultra, the $250 per month subscription. This package offers higher AI limits, 30 TB of cloud storage, and access to faster development tools, including Google’s Antigravity-coding agent. Google writes that the international rollout will follow later.