Google's Genie 4 world model gains audio, more realistic worlds, and real-time control
The Gemini Robotics experiment demonstrates a potential way to test robot actions in simulation before executing them in the real world
Google DeepMind has released Genie 4 as a research preview. The model generates interactive scenes with real-time audio—including music, background noises, and character dialogues. Users can modify the scene using text commands directly during the simulation.
Compared to Genie 3, the new version allows on-the-fly changes to weather and camera angles, as well as the addition of people and objects. It maintains scene consistency for up to 2 minutes, improves physics modeling, and creates more human-like characters. Among the stated use cases are gaming, interactive learning, and robotics.
Genie 4 still has limitations: during prolonged simulation, details and objects may change, and developers have yet to resolve issues with audio accuracy and text readability within scenes. Currently, the model is available to trusted testers, and in the coming weeks, it will be provided to Google AI Ultra subscribers in Project Genie and early-access partners of the Gemini API and Google AI Studio.
