Google adds sound and live scene control to Genie 4
The world model is a research preview for trusted testers. Google still lists drifting objects and unreliable physical interactions among its limits.
Verified 12:17 AM PDT · 2 original sources
Google DeepMind announced Genie 4 on October 27, adding generated sound and dialogue to interactive simulated worlds. Users can change a scene with text while moving through it.
Google says Gemini 3.8 Flash watches the generated frames, divides goals into steps, and checks progress. Genie draws the world as that process runs. The company is testing uses in creative planning and simulated environments for robots and agents.
Access remains a research preview for trusted testers. Google says broader access is coming to Project Genie for Google AI Ultra subscribers and selected developer partners. A waitlist is not general availability.
These are Google’s claims, not independently verified robot-training results. Google says objects and visual details can drift during longer sessions or repeated edits. Physical interactions, sound accuracy, readable text, and spatial consistency remain research problems. The announcement does not establish that robots trained in these worlds perform better in real settings.
Live changes let testers examine a sequence of events instead of accepting a finished video. That could help them test an idea before building a scene or physical setup.
Look for repeatable task tests, published failure cases, and results on physical robots. Broader access will let people test more than the scenes Google selected.
Audit the story
Original sources
Company claims remain company claims. Follow the reporting and judge the evidence directly.
Continue the edition