Google said on October 27 that it had begun rolling out Genie 4 to a limited group of trusted testers. The world model lets users change scenes in real time with natural-language prompts and generates audio natively, including ambient sound, music, and dialog.

Access will soon be available through Project Genie for Google AI Ultra subscribers, and through the Gemini application programming interface (API) and AI Studio for early access partners.

The announcement extends the Genie product line. Google DeepMind previously described Genie 2 as a foundation world model and later described Genie 3 as a new frontier for world models.

Google AI Ultra is a paid subscription that Google says offers access to more capable Gemini models. Generative media tools raise broader copyright questions about training data and AI-generated outputs.

🔗 Source: Google