Google DeepMind unveils Genie 4: interactive AI worlds with sound, speech and real-time steering
Type a world and walk into it. Google DeepMind announced Genie 4 on Tuesday, Oct 27, a “world model” that generates interactive, photorealistic scenes in real time, now with its own sound (Google DeepMind).
What’s new
- Native audio: each world comes with its own ambient sound, music, and interactive dialogue from characters who react to you. DeepMind calls it its first real-time world model to generate audio.
- Steer it live: move with keyboard controls, then type prompts like “make it rain” or add a person or object, and the scene adapts.
- More realistic physics and people, including collisions, aerodynamics and characters that move naturally.
- Gemini in the loop: Gemini 3.8 Flash watches the scene and plans multi-step events so the world stays coherent.
Who can use it
For now it’s a research preview for trusted testers. DeepMind says it’s coming soon to Google AI Ultra subscribers in Project Genie and to early-access partners in the Gemini API and Google AI Studio, with a waitlist open. No date or price was given.
The limits
DeepMind admits details can drift in longer sessions, and that text, spatial accuracy and long-term consistency still need work. Everything Genie 4 generates carries Google’s SynthID watermark.
Koffee take: game developers are already testing it for quick prototypes. It won’t replace a real game engine yet, but it’s a wild peek at where this is going.
Filed Oct 27, 2026 ~12:30 PM CT — Midday pass. Sources: Google DeepMind — Genie 4: A new generation of world simulation (Oct 27, 2026) · TechCrunch coverage (Oct 27, 2026).