Key Capabilities of Genie 2
Infinite procedural world generation from text or image inputs
Real-time rigid-body and soft-body physics engine
Context-aware object affordances (e.g., doors open, levers activate mechanisms)
Biomechanically grounded character locomotion and animation
Cross-episode memory for coherent world evolution and continuity
Real-World Applications
Accelerate game design iteration—from concept to playable prototype in under a minute
Benchmark and train embodied AI agents across procedurally varied, physically grounded scenarios
Frequently Asked Questions
-
Is Genie 2 publicly available?
-
Can Genie 2 generate worlds from images alone?
-
About Genie 2 by Google DeepMind
Developed by Google DeepMind, Genie 2 represents a major leap in foundation models for interactive simulation. Official platform: playgenieai.com.
Learn more about its architecture, research origins, and vision on the About page.
-
Genie 2 by Google DeepMind on YouTube
Watch live demos, technical deep dives, and behind-the-scenes insights on the official channel: https://www.youtube.com/@google_deepmind.
-
Genie 2 by Google DeepMind on GitHub
Explore open research artifacts, evaluation frameworks, and model cards at https://github.com/google-deepmind.