Google DeepMind began a research preview of a new world model, Genie 4, on Tuesday, generating interactive scenes with audio for the first time, the company said in a blog post.
The model generates sound directly as it builds a scene and can hold a consistent scene for up to two minutes, twice as long as its predecessor, according to DeepMind. Users can steer the simulation frame by frame as it runs.
Google is also testing its Gemini 4 Argon model alongside Genie 4 for robotics simulation.
Access is currently limited to select testers. Google AI Ultra subscribers will get it later through the Project Genie web app, and some partners will receive early Gemini API access, the company said.
“What we'll increasingly see in the future are these hybrid world models, where you have a combination of Gemini writing code and text, and Genie ingesting that in real time and simulating reality at different levels of abstraction,” said Tejas Kulkarni, a research director at Google.
Diego Rivas, a Google DeepMind engineer, said the real-time aspect “becomes most valuable in the areas where keeping a state of flow in the creativity space becomes incredibly useful.”
DeepMind acknowledged that objects and details can drift during longer simulations and that audio accuracy and in-world text legibility remain unresolved. The company did not disclose performance benchmarks, technical architecture details, subscription pricing or a timeline for general availability.
DeepMind showed the previous version, Genie 3, in August 2025.
