Google DeepMind just unveiled Genie 4, and it's a wild leap for real-time world models. You can navigate a generated 3D world with keyboard controls and steer the scene on the fly with plain text: change the weather, move the camera, or drop in a new character mid-simulation. For me, the biggest addition is native real-time audio. Earlier world models were silent video generators. Genie 4 generates ambient sound, music, and interactive character dialogue on the fly as the scene unfolds, with Gemini running in the loop like a live director to keep the physics and story coherent for up to two minutes. This is where gaming, filmmaking previz, interactive music experiences, and robotics training start blurring into the same engine. When a world model scores its own soundtrack and dialogue as you walk through it, what happens to how we build games and virtual worlds?

Genie 4: A new generation of world simulationGenie 4 introduces real-time interactive world simulation with native audio, advanced physics, natural human motion, and steering powered by Gemini models.deepmind.google