Google DeepMind Unveils Genie 4, an AI That Simulates Living Worlds
Google DeepMind released Genie 4, its most advanced AI world model to date, as a research preview on October 27, 2026. The model marks a major leap in real-time interactive simulation, generating fully explorable, photorealistic environments with native audio—a first for the Genie series.
Genie 4 is natively multimodal, generating music, ambient sounds, and interactive dialogue alongside its simulated environments. Users can steer scenes in real time with natural-language commands, such as:
- Changing the weather or camera angle
- Adding characters
- Triggering events during a simulation
The model is built on the Gemini family. Gemini Omni anchors the visual realism, while other Gemini models help maintain scene coherence and logical continuity.
Compared with its predecessor, Genie 4 can maintain a consistent scene for up to two minutes—double Genie 3’s one-minute limit. It also offers improved physics modeling for aerodynamics, collisions, and humanlike character movement.
DeepMind says the technology could support applications in:
- Robotics training
- Gaming
- Filmmaking
- AGI research
The model is currently rolling out to trusted testers. Wider access is planned for Google AI Ultra subscribers through Project Genie and for developer partners through the Gemini API.
Source: TechCrunch — Google’s Genie 4 world model adds audio, more realistic worlds, and real-time steering