Google unveils Genie 4, a world model that builds interactive scenes from text
The model is in a limited research preview, with wider access for AI Ultra subscribers and early partners expected in the coming weeks.
Google has unveiled Genie 4, a world model that generates interactive environments in real time from natural-language instructions, and it is being rolled out first to a small set of trusted testers as a research preview. Compared with Genie 3, it produces more photorealistic scenes, generates synchronized audio such as music, ambient sound and dialogue, and can hold a consistent scene for up to two minutes instead of one. Users can change weather, camera angle, characters or objects mid-simulation, and Gemini models supply the planning and frame-checking that keep the scene coherent. Google positions the technology for robot training, film pre-visualization, game prototyping and education, though it concedes that details can drift over longer runs and that audio accuracy and legible in-world text remain unsolved. Access is expected to expand to Google AI Ultra subscribers in Project Genie and to early partners on the Gemini API and AI Studio in the coming weeks.
Why it matters
For people following AI, the release marks a step in generating interactive environments rather than fixed clips: scenes now hold together for up to two minutes and carry synchronized sound. Google's own caveats matter just as much, since details can drift over longer runs and audio accuracy and in-world text are still unresolved. Until wider access arrives, the claims remain Google's to demonstrate.
Matters or hype?
Does this story matter, or is it hype? Vote before you see what everyone else said.