Google DeepMind Unveils Genie 4 For Real-Time AI Worlds With Built-In Audio: All Details

Google DeepMind has unveiled Genie 4, its most realistic and interactive world model to date, in a research preview. Users can type text or image prompts to create a photorealistic world they can explore, and even change it live while it is running. This is Genie 4's first real-time world model with built-in audio, which means music, ambient sounds, and interactive dialogue are generated alongside the world.

Users can alter various elements of the simulation mid-simulation, such as changing the weather, camera angle, adding a person or an object, without disrupting the ongoing scene. DeepMind claims that the visuals are built on Gemini Omni and that the physics has improved, with more accurate aerodynamics, collisions, and natural human movement.

A scene remains consistent for up to two minutes, which is double the limit of Genie 3, which was released in August 2025.

The introduction of Genie 4 is part of DeepMind's Gemini Robotics initiative, where a Gemini 4 Argon model broke down tasks into steps and simulated them in real-time to rehearse robot scenarios before attempting them in the real world. Furthermore, worlds like these could potentially train agents such as DeepMind's SIMA game-playing system. Companies like Waymo, which adopted Genie 3 in February 2026 to build a specialized world model for autonomous-driving simulation, have already explored similar applications.

Genie 4's intended uses include gaming, interactive learning, and robotics. However, it is currently not open to the public, with only trusted testers having access. Google AI Ultra subscribers will be able to access it through Project Genie, Google's tool for generating interactive worlds from prompts, and early-access partners of the Gemini API and Google AI Studio.

Google has not announced a wider release date or availability in India. The competition in this area has intensified since the launch, with AI video startup Runway releasing GWM Worlds 2, Alibaba announcing a world model called Happy Oyster, and the technology's potential applications extending to games, robots, and self-driving cars.