Google DeepMind announced Genie 4 on Tuesday, introducing a new world model designed for real-time interactive simulations with improved photorealism, native audio generation, and advanced capabilities for dynamic scene control.

The model builds on the Gemini family of systems. Gemini Omni anchors simulations in realistic appearances and behaviors, while additional Gemini models, including integration with Gemini 3.8 Flash, support reasoning to maintain scene coherence during extended interactions. This setup allows the system to plan multi-step events and evaluate output frames for continuity.

Key capabilities include real-time steering through natural language prompts, enabling users to alter weather conditions, adjust camera angles, introduce new characters or objects, and interact with elements mid-simulation. Genie 4 generates audio natively on the fly, including ambient sounds like wind or waves, music, background noise, and interactive dialogue that adapts to changes in the scene. Characters exhibit more natural movement and responses, with improved physics simulation covering aerodynamics, collisions, and human-like behaviors.

Simulations maintain consistency for up to two minutes, doubling the limit of the prior version. The system offers greater photorealism across spaces and subjects, reducing uncanny valley effects in human figures. Applications span safe training environments for robots and agents, as well as creative tools for filmmaking, gaming, journalism, and manufacturing. These features represent progress toward more sophisticated virtual environments that support testing of complex scenarios.

Google DeepMind has begun a research preview rollout to a limited group of trusted testers. Broader access is planned soon for Google AI Ultra subscribers through Project Genie and for early access partners via the Gemini API and AI Studio. A waitlist for Project Genie is now open.

This development follows the release of Genie 3 more than a year ago and better positions the company amid competition in simulation technologies from other firms.