Google DeepMind has unveiled Genie 4, an updated world model capable of generating interactive virtual environments in real time. The new version introduces significant improvements in physical realism, human character appearance, and the integration of audio support for the first time.
The system utilizes the Gemini 3.8 Flash model as a control agent to manage user objectives, evaluate frames, and maintain context across different environments. Despite these enhancements, the generated worlds remain stable for only a few minutes before losing their initial conditions.
Developers have improved the simulation of aerodynamics and collision physics. Characters now exhibit more natural movements and reactions, while the model can generate background sounds and interactive, adaptive speech. The technology is currently positioned as a research project.
Access for developers is provided via the Gemini API and AI Studio. Consumers can access the model through Google AI Ultra plans, starting at $100 per month, with potential future availability in AI Pro and AI Plus tiers. A limited version remains accessible via the Project Genie web interface.
