Google prepares Genie 3 public experiment with AI-generated worlds

Google is preparing a public experiment for its Genie 3 model, allowing users to generate explorable virtual worlds from simple prompts. Genie 3, introduced earlier this year as an AI world model, enables the creation of digital environments that users can walk through and interact with. The upcoming experiment addresses earlier feedback, as the initial Genie 3 reveal did not include a way for users to try the technology directly.

The new experiment will reportedly allow users to describe both the world and the character, then navigate and interact within the generated environment. No release timeline has been specified. The feature is expected to appear as a web-based playground or possibly in Google Labs, targeting early adopters, researchers, and developers interested in generative AI and interactive simulations.

If released, this would extend Google’s current strategy of offering hands-on demos for advanced models, similar to its NotebookLM and ImageFX projects. It would also serve as a public test for generative world modelling, a field with applications in education, entertainment, and rapid prototyping of virtual spaces. Details about performance, limitations, or export options are not yet available. Discovery of the experiment is based on preview materials and code references, not on any official launch.