Home/New releases/Genie 4

Genie 4

DeepMind's real-time interactive world model, now with native audio and dialogue. Research preview only.

DeepMind announced Genie 4 on October 27, 2026. It is a real-time interactive world model, and DeepMind's first real-time world model to generate audio natively. It generates audio, including music, ambient sound, and interactive dialogue, together with the world. Users steer it in real time with standard keyboard controls for navigation and camera, plus text prompts, such as changing the weather or adding a person or object. Gemini models help keep the simulation coherent. Gemini 3.8 Flash acts as a guiding agent in the loop for multi-step events. Access is rolling out as a research preview for trusted testers. It is coming soon to Google AI Ultra subscribers in Project Genie, and to early access partners in the Gemini API and Google AI Studio. There is a waitlist for Project Genie. No public pricing is listed. Game developers among the trusted testers use it to go from an idea to an explorable prototype before writing engine code. DeepMind says details and object permanence can drift over longer sessions, and that complex physics, audio accuracy, legible text, spatial accuracy, and long-term consistency are still being worked on. All output carries SynthID watermarks. You cannot use it yet unless you are a trusted tester. Last checked October 28, 2026.

• Hy-Nomas
• Game dev list