Google DeepMind unveils Genie 4 world model with native audio and more realistic people

As of 6:57 a.m. ET, Monday, Nov. 2, 2026

Google DeepMind on Tuesday announced Genie 4, the latest version of its AI "world model," which generates interactive, explorable environments frame by frame in real time. The company calls it its "most realistic and interactive world model yet."

Genie 4 is starting as a research preview for a limited group of trusted testers, DeepMind said in a blog post. The company said it plans to bring it to Google AI Ultra subscribers in its Project Genie app and to early-access partners in the Gemini API and Google AI Studio "soon." It gave no specific date.

What DeepMind says is new

According to DeepMind, Genie 4 is its first real-time world model to generate audio natively, including ambient sound, music and interactive dialogue in which characters "converse and react to your inputs." The company says the model simulates physics such as aerodynamics and collisions more realistically, and that "the most dramatic leap" is in how it simulates people who "move naturally" and can be steered in real time.

Users can type prompts mid-simulation to change the weather, move the camera or add a person or object, DeepMind said. The model builds on Google's Gemini family: Gemini Omni "anchors the simulations in real-world appearance and behavior," and Gemini 3.8 Flash acts as a guiding agent that plans multi-step events and checks incoming frames, according to the post.

DeepMind frames world models as a step toward artificial general intelligence and as a way to build training environments for robots and AI agents. It says all video and audio from Genie 4 will carry SynthID, its invisible watermark for AI-generated content.

DeepMind's stated limits

DeepMind describes Genie 4 as "an early, experimental model." In a section on current limitations, the company says that while it "can maintain coherent environments over short sessions, visual details and object permanence can still shift or drift over time," especially in longer simulations, after multiple complex prompt edits or while moving through a scene. It lists complex physical interactions, real-time audio accuracy, legible text, spatial accuracy and long-term consistency as areas still under research.

Ars Technica reported that DeepMind says Genie 4 worlds stay consistent for only a couple of minutes. TechCrunch reported the limit as up to two minutes, double that of its predecessor, Genie 3.

What is not known

Pricing and a release date for wider access have not been announced. DeepMind's capability claims are based on its own demos and testing; GROKsiNEWS.com has not tested the model.