Google DeepMind updates Project Genie with Genie 4 and chainable worlds
Google DeepMind has released Genie 4, bringing real-time interactive 3D world generation to subscribers.
2 min read, Wednesday, 18 November

Google DeepMind is expanding its interactive environment generation platform with the release of Genie 4. Previously limited to a research preview, the new model is now available to Google AI Ultra subscribers.
Genie 4 is a world model — an artificial intelligence system that learns the rules of a physical environment to simulate interactive 3D spaces. Users can modify these virtual spaces using text prompts in real time, altering weather conditions, characters, or narratives on the fly. The system also generates synchronized audio that responds dynamically to changes in the scene.
Key facts
- World Model
- Genie 4
- Project
- Project Genie
Updates to previous versions
Alongside the flagship release, Google DeepMind is updating the previous Genie 3 model with several new features. Sessions can now run for up to three minutes, and the system adds support for 360-degree panoramas.
A notable addition is the ability to chain multiple worlds together into longer sequences. To make world-building more accessible, a new assistant agent can construct environments directly from spoken descriptions.
Safety and community sharing
To encourage collaboration, the update includes a community gallery where users can publish and remix each other's virtual worlds.
Content moderation remains a central focus for the platform. All media generated through Project Genie undergoes automated safety checks and is embedded with SynthID watermarks, a technology designed to invisibly label AI-generated content. While Google AI Ultra subscribers have immediate access to Genie 4, Google AI Pro members can currently sign up for a waitlist to access the new features.
Source: Original announcement
Worth a read?



Comments