# Genie 2

- Date: 2024-12-04
- Type: model
- Credit: Google DeepMind
- URL: https://aifilmhistory.org/entry/genie-2

Google DeepMind showed Genie 2, a world model that turns a single image into a playable 3D environment a person or an AI agent can move through with a keyboard and mouse. Each demo began with a picture from Imagen 3; the model then generated the world frame by frame as the player moved, with objects you could interact with, characters that animated, physics, and parts of the scene that came back intact after leaving view. It was a research post, not a product, framed as an endless supply of training worlds for embodied agents.

For filmmakers it was the first sign that a generative model could hold a place rather than a shot: the idea [Genie 3](https://aifilmhistory.org/entry/genie-3) made real-time eight months later.

Links:
- Google DeepMind: Genie 2, a large-scale foundation world model: https://deepmind.google/discover/blog/genie-2-a-large-scale-foundation-world-model/
- Engadget: Genie 2 can generate interactive 3D worlds: https://www.engadget.com/ai/google-deepminds-genie-2-can-generate-interactive-3d-worlds-200708207.html

---
Source: https://aifilmhistory.org/entry/genie-2
Era: 2024 · The Great Convergence
Last updated: 2026-10-17