Google Genie 2

About the product

Genie 2 is a foundation world model from Google DeepMind that generates an endless variety of action-controllable, playable 3D environments for training and evaluating embodied agents. Given a single prompt image, it creates a world that a human or an AI agent can play using keyboard and mouse inputs. As a world model it simulates the consequences of actions, including object interactions, character animation, physics and the behavior of other agents, and it remembers parts of the world that are no longer in view. Genie 2 can generate consistent worlds for up to a minute, with most examples lasting 10-20 seconds, and can render different perspectives such as first-person, isometric and third-person views.

Genie 2 pros

• Creates diverse, action-controllable 3D worlds from a single image
• Simulates physics, object interactions and other agents
• Provides a limitless curriculum of novel worlds for training embodied agents

Genie 2 cons

• A research model, not a public product you can sign up for
• Generated worlds stay consistent for up to about a minute