The Genie 3 is a significant upgrade over its predecessor, the Genie 2. The new model lets users interact with AI-generated 3D style worlds for “several minutes” instead of just 10-20 seconds, features an improved visual memory that remembers an object’s location for about a minute, runs at 720p / 24fps, and includes “prompt world events” for dynamic changes like weather or the addition of characters.
However, Genie 3 is only available as a limited research preview to a small cohort of academics and developers while Google investigates potential risks, and the world models still face challenges like image quality and object consistency, which make them appear less complete than traditional video games.
It’s an area Google is putting a lot of effort into. The company showed off Genie 2 in December, 2024, and it could create interactive worlds based off of an image, and it’s building a world models team led by a former co-lead of OpenAI’s Sora video generation tool.
Google says it is “exploring” how Genie 3 could be tested by “additional testers” in the future. DeepMind is also adding what it calls “promptable world events” to Genie 3, too.
