DeepMind Unveils Genie 4: Real-Time Worlds With Sound

Google's AI division, DeepMind, has released Genie 4, a generator of interactive environments. It is the first version in the Genie line to include real-time sound: sound effects, background music, and dialogue between characters. The system uses Gemini 3.8 Flash to keep track of a goal and adjust its plan as the action unfolds. For now, the tool is available only to a small group of selected users. It will soon open up to Google AI Ultra subscribers through Project Genie, as well as to early-access users of the Gemini API.

DeepMind itself admits that scene detail and object memory still drift over time. So this is more of a research prototype than a finished product. This is not about writing code in a command-line interface and then checking the result. It is about training agents or robots in a safe environment, with no real consequences. So yes, there is something to this, as long as DeepMind eventually opens it up to everyone.