Google DeepMind unveils Genie 4: a real-time world model that also generates sound and dialogue

Google DeepMind announced Genie 4 on 27 October 2026, the latest version of its “world model” — an AI system that generates an explorable, interactive environment frame by frame in real time, which users navigate with keyboard controls and change with text prompts.

What is new

• Native audio: DeepMind calls it its first real-time world model to generate sound natively — ambient noise, music and interactive dialogue from characters, produced together with the picture.
• Real-time steering: users can change the weather, move the camera or add people and objects mid-simulation; a Gemini model adapts the prompt to the scene.
• Physics and people: the company says simulation of collisions, aerodynamics and especially natural human movement has improved.
• Gemini in the loop: Gemini 3.8 Flash watches the generated frames, breaks goals into steps and checks progress, which DeepMind says enables multi-step events. The lab also showed Gemini 4 Argon planning a simulated humanoid robot playing checkers.

Availability and limits

Genie 4 is rolling out as a research preview to trusted testers, with access “coming soon” for Google AI Ultra subscribers in Project Genie and early-access partners of the Gemini API and AI Studio. All output carries the SynthID watermark. DeepMind itself lists limitations: details and object permanence can drift in longer sessions or after many edits, and complex physics, audio accuracy, legible text and long-term consistency remain open research problems. Demonstrations so far are the company’s own selected recordings.