Google's 3D world generator now creates real places from Street View
Google's Project Genie is becoming more widely available and can now generate interactive 3D environments from Street View imagery.
(Image: Google)
At its I/O 2026 developer conference, Google unveiled a new capability of Project Genie: the 3D world generator can now access images from Google Street View and create explorable 3D environments of real places from them. This brings AI-generated worlds and real-world locations closer together.
To achieve this, users select a location in the USA in Project Genie via a map icon, optionally set a style such as “desert world” or “stone age,” and then describe a game character. Genie then creates a fantasy world based on Street View imagery. In its blog, Google uses the Golden Gate Bridge as an example, depicted as an underwater world to be explored by divers.
Technically, the function is based on the “Maps Imagery Grounding” imaging tool introduced in April, which allows generative AI images to be anchored in real street views from Google Street View. According to Google, only locations in the USA are currently available, with more regions to follow later.
Videos by heise
Project Genie was previously reserved for US users. With yesterday's announcement, Google is now gradually making Project Genie available outside the USA. However, users will still need the premium subscription “Google AI Ultra.” The previous plan for $250 per month was replaced at the developer conference by two new Ultra plans for $100 and $200 per month. Project Genie remains reserved for the more expensive $200 plan, but it will at least become slightly cheaper.
No competition for game developers yet
Project Genie is based on Genie 3, Google's general world model, which can generate diverse and partially interactive environments. Google is researching such world models for, among other things, the virtual training of AI agents, robots, and autonomous systems such as self-driving vehicles.
In January, Google first made the model accessible to users and referred to this area of research as Project Genie. While Genie is developed by Google DeepMind, Google Labs is responsible for Project Genie.
Google describes Project Genie as an experimental prototype because its functions are still significantly limited: generated 3D environments can currently only be explored for about a minute, and interactivity remains highly limited. To create a world, text inputs, images, and now also real locations are sufficient.
At the announcement in January, the stocks of major game companies came under pressure because investors saw AI models like Project Genie as potential competition for classic game development. However, our practical test showed that the industry has no reason to worry for now.
(afl)