Google DeepMind Says Genie 4 Builds Worlds with Live Audio and Mid-Scene Prompts
Summary
Google DeepMind released Genie 4 in research preview, a world model it says generates photorealistic interactive scenes with matching real-time audio and accepts typed prompts mid-simulation. It holds a consistent scene for up to two minutes. Trusted testers get access now; Google AI Ultra subscribers follow in Project Genie.
Key Points
- Research director Tejas Kulkarni says future systems will be "hybrid world models" combining Gemini writing code and text with Genie simulating reality in real time.
- Google acknowledges that details and objects drift during longer simulations, and that audio accuracy and in-world text legibility remain unsolved.
- Early access partners of the Gemini API and Google AI Studio get Genie 4 access in coming weeks.