Google Unveils Genie AI That Creates Interactive Video Games from Text and Images
-
Genie is a new AI model from Google DeepMind that can generate interactive video games from text or image prompts, without needing game mechanics training.
-
Genie has 11B parameters and uses a video tokenizer, dynamics model, and action model to generate controllable, playable environments.
-
Genie can take a single image prompt and bring it to life, allowing people to step into and interact with imagined virtual worlds.
-
Genie learns controls and consistent behaviors exclusively from unlabeled internet videos, with no labels or domain-specific training.
-
Genie lets anyone easily create and step into their own fully imagined virtual worlds from sketches or AI-generated images.