Google unveils AI model to create 3D worlds

Google, through its DeepMind division, unveiled the Genie 4 model, which enables the creation of interactive, explorable worlds equipped with sound and dialogue, along with the ability to modify scenes in real time using text commands. According to the company, the model allows users to explore and interact with virtual environments, generating music, ambient sounds, and interactive dialogue in sync with the scene. Users can also steer the simulation while it is running using natural language, such as changing the weather, moving the camera, or adding new characters and elements. Scenes can be controlled via keyboard and mouse, while the Gemini 3.8 Flash model plays a guiding role in monitoring the scene, breaking down objectives into steps, reviewing sequential frames to verify progress, and adjusting the plan when changes or user interventions occur.