Sig AdvisorySig Advisory
models

Google DeepMind's Genie 4 world model generates audio with the scene, but only testers can use it

Genie 4 builds interactive worlds you steer with text (move the camera, change the weather, add a character) and generates sound, music and dialogue as it goes. It's a research preview for trusted testers, with AI Ultra subscribers and Gemini API early-access partners next.

Written by Mara at Sig · Oct 28, 2026

Google DeepMind announced Genie 4 on October 27. You set a scene in motion and steer it in real time with natural language: move the camera, change the weather, bring in new characters. Each world also generates its own audio on the fly, including ambient sound, music and spoken dialogue. DeepMind says Gemini Omni grounds how things look and behave, while other Gemini models keep the simulation coherent.

Access is narrow for now. It's rolling out to a limited group of trusted testers, then to Google AI Ultra subscribers in Project Genie and to early-access partners in the Gemini API and AI Studio. There's a waitlist.

DeepMind lists the limits itself: visual details and object permanence can drift in longer sessions or after several complex prompt edits, and physics, audio accuracy, legible text and long-term consistency are still open research problems.

Why it matters

Google pitches world models as training grounds for robots and agents, and Gemini API access is on the roadmap. For now there's no API, no price and no date, so this is one to watch rather than build on.

Sources

Back to all newsGet Sig Weekly every Monday