Google Opens Genie 4 World Model to Developers via Gemini API

Google has opened up Genie 4, its interactive world-generation model, to developers through the Gemini API. The model can produce 720p environments that respond to real-time controls, with native audio baked in rather than added as a post-processing step. That combination - spatial coherence, interactivity, and synchronized sound - marks a meaningful step beyond earlier versions, which were primarily research demonstrations.
For developers, the API access means Genie 4 can be integrated into games, simulations, and interactive experiences without needing to run the underlying model independently. The 720p output resolution is a practical improvement over prior iterations, which operated at lower fidelity and were slower to respond to user input.
Separately, Google is rolling out a consumer-facing Project Genie experience to eligible Google AI Ultra subscribers. That interface is distinct from the developer API and appears aimed at letting users explore generated worlds directly, without writing code.
Genie 4 sits in an unusual category - part generative video model, part interactive simulation engine. Its availability on a commercial API suggests Google is moving to position it as infrastructure for interactive media rather than keeping it confined to research contexts.

