News
Breaking: Google DeepMind Unveils Genie 4, a World Model With Sound You Can Steer Live
Google DeepMind announced Genie 4 on October 27, 2026: a real-time world model with native audio and live text steering, in research preview.
Google DeepMind announced Genie 4 on Tuesday, October 27, 2026, a new world model that generates interactive, photorealistic simulations in real time — and, for the first time in the Genie line, generates the sound to go with them. Music, ambient noise, and characters who talk back are produced in the same generative pass as the visuals, not layered on afterward. You can also steer a running simulation with plain-language prompts: change the weather, move the camera, or add a person or object while the scene keeps playing.
What we know
- Genie 4 builds a place you can move through, rather than a clip you watch. It takes a text or image prompt and simulates what happens next as you act, using keyboard controls and text prompts.
- Native audio is the headline change. DeepMind says every world gets its own sound — wind, waves, music, and interactive dialogue — generated simultaneously with the visuals.
- Two Gemini models work inside the loop: Gemini Omni anchors how objects look and behave in the real world, while Gemini 3.8 Flash acts as a guiding agent, breaking a goal into steps and adjusting the plan as the user intervenes.
- Scenes now hold together for up to about two minutes, roughly double Genie 3's limit, according to launch reporting. DeepMind itself cautions that details and object permanence can still drift over longer sessions or after many complex edits.
- Access is limited for now. Genie 4 is a research preview for trusted testers — filmmakers, game developers, and robotics teams — with rollout to follow for Google AI Ultra subscribers in Project Genie and early-access partners of the Gemini API and Google AI Studio. A waitlist is open; DeepMind has given no firm timeline or pricing.
What's unconfirmed
DeepMind has published no figures for resolution, frame rate, or API pricing, and the two-minute coherence figure comes from launch reporting rather than DeepMind's own post. Real-world quality beyond the company's curated demos — especially for longer, heavily edited sessions — remains untested by independent users.
Why it matters
World models are the bridge between chatbots and machines that act in physical or virtual spaces: the same simulation tech trains robots and agents and prototypes games and films. Genie 4's native audio and live steering make those simulations meaningfully more complete, but for regular users nothing changes today — this is a research preview, not a product you can open yet. The practical signal is the direction: DeepMind is pushing world simulation toward something you can direct like a scene, not just watch like a video. If you work in games, film previsualization, or robotics training, the waitlist is the concrete next step.
Sources: Google DeepMind's Genie 4 announcement (October 27, 2026), as reported by TechCrunch, The Verge, and TechTimes. Specs beyond DeepMind's post are attributed to reporting where noted.