Google DeepMind on Tuesday (October 27) released a research preview of its latest world model, Genie 4, which has been made available to "trusted testers." The model is capable of generating more realistic and interactive simulated worlds, and for the first time supports real-time audio (including music, ambient sounds, and dialogue). Users can change scenes, weather, or perspectives in real-time through text prompts during the simulation, and add new characters or objects. Genie 4 can maintain a consistent scene for up to two minutes, double that of its predecessor Genie 3, and has improved physics and character anthropomorphism. The model is also more tightly integrated with Google's AI model Gemini and can be used in commercial applications such as gaming, interactive learning, and robotics testing. Its real-time editing capabilities have also been noted by testers for use in film prototyping.