Google unveils Genie 4, its most realistic world model powered by GeminiGoogle Deep

Mind on Tuesday, October 27, 2026, unveiled Genie 4, its latest world model described as its most realistic and interactive world model yet.

Genie 4 is a real-time, interactive world simulation system that allows users to generate and navigate dynamic environments on the fly, with the ability to steer scenes by typing prompts mid-simulation – changing weather, camera angles, or adding new people and objects.

For the first time, a Google world model generates audio natively within the same loop as visuals. That includes ambient sounds like crashing waves, music, and characters who can converse and react in real time, all causally tied to events in the scene.

The model is powered by multiple models from the Gemini family. According to Google, Gemini Omni anchors simulations in real-world appearance, physics and visual continuity, while Gemini 3.8 Flash acts as a guiding agent in the loop – observing frames, breaking goals into steps, and adapting plans as the world shifts.

Google said the most dramatic leap is in how Genie 4 simulates people, creating characters that move naturally and respond organically.

The model can maintain scene consistency for up to two minutes, double the limit of its predecessor Genie 3, which was launched about a year ago.

Google highlighted use cases including creative direction for games and movies, historical scenario building, immersive experiences, and developing training environments for robots and AI agents.

Genie 4 is currently in research preview for trusted testers, with rollout planned for Google AI Ultra subscribers in Project Genie and early access partners via the Gemini API and Google AI Studio. All video and audio generated will include Google’s SynthID watermark.