Google Unveils Genie 4 World Model With Native Real-Time Audio and Steering

Google has launched Genie 4, a new generative world simulation model featuring native real-time audio and live prompt steering. Currently available as a research preview to trusted testers and select industry partners, the system doubles consistent scene simulation duration to two minutes compared to Genie 3. Powered by Google's Gemini models, Genie 4 supports photorealistic rendering, dynamic environmental adjustments, and robotic simulations. Google plans to roll out Genie 4 to Google AI Ultra subscribers and developers via the Gemini API in the coming weeks, with all generated media watermarked using SynthID.