Google DeepMind has released its latest world model, "Genie 4", as a research preview, TechCrunch reported on Oct. 27 (local time).

Genie 4 focuses on creating realistic, highly interactive virtual worlds, the report said. Users can enter prompts during a simulation to change a scene. It also generates scene-appropriate audio in real time, including music, ambient noise and dialogue.

It supports flexible editing such as changing weather and camera angles in real time and adding new characters and objects. It maintains scene consistency for up to 2 minutes, twice as long as Genie 3 released last year. The company said its physics representation has improved and characters move more like humans.

Genie 4 also integrates closely with Google's AI model Gemini. If Gemini takes the role of director by planning steps and checking frames, Genie can create the visuals.

Tejas Kulkarni (테자스 쿨카르니), a Google research director, said, "Going forward, there will be more hybrid world models where Gemini writes code and text and Genie receives it in real time to simulate reality across different levels of abstraction."

Google presented games, interactive learning and robotics as key uses. "Trusted testers" who try the new product first are also mentioning a new use case of producing film prototypes, TechCrunch reported.

Diego Rivas (디에고 리바스), a Google DeepMind engineer, said, "Real-time features deliver the greatest value when rapid ideation is needed, such as during concept development or the storyboard stage."

Rivas stressed that Genie 4 is still in its early stages. As simulations run longer, details and objects can become distorted, and issues with audio accuracy and the readability of text in the virtual world remain unresolved, TechCrunch reported.

Genie 4 is currently provided as a research preview to trusted testers. It is set to be rolled out within weeks to Google AI Ultra subscribers through Project Genie. Gemini API users and Google AI Studio early-access partners will also be able to use it.

Competition in the world model market also appears to be intensifying. A few months ago, AI video startup Runway released "GWM Worlds 2", and AMD announced it would acquire World Labs founded by Fei-Fei Li.