Google Launches Genie 4 World Simulation Model

Google announced the release of Genie 4, a world simulation model capable of generating real-time interactive environments. The system integrates native audio, including dialogue, music, and ambient noise, and allows users to modify simulations in real-time using natural language commands to adjust weather, camera angles, or objects.

Built on the Gemini family of AI systems, the model utilizes Gemini Omni for appearance and behavior simulation and Gemini 3.8 Flash as a guiding agent to track goals and progress. Genie 4 supports physics simulations and can create experiences up to two minutes long, which doubles the capacity of the previous Genie 3 model.

The company has released the model as a research preview to trusted testers in engineering, gaming, and filmmaking. Google plans to expand access to Gemini API partners and Google AI Ultra subscribers. To identify AI-generated content, all output includes SynthID digital watermarking.

Google acknowledged several current limitations of the technology. These include challenges with text rendering, complex physical interactions, and visual drift that occurs over extended sessions.