Google Unveils Genie 4 World Model With Real-Time Steering and Audio

27 Oct 2026, 10:00 pm · 4h ago · 1 min read · TechCrunch AI

Google has announced Genie 4, the latest iteration of its generative world model designed to simulate interactive environments. The new model doubles the scene consistency duration of its predecessor, maintaining coherent virtual worlds for up to two minutes. Genie 4 introduces real-time user steering, integrated audio generation, enhanced physics simulation, and more natural humanlike character behaviors. Unlike standard video generators, world models allow users to act within generated spaces, opening up applications across game development, robotic training, and synthetic simulation. Google aims to advance autonomous agent research and interactive media with this architecture.