Google DeepMind unveils Genie 4, a real time world simulation model

Google DeepMind announced Genie 4, a new world model that generates interactive, explorable environments in real time, complete with native audio such as ambient sound and dialogue generated alongside the visuals. The model builds on Gemini: Gemini Omni anchors simulations in realistic appearance and behavior, while other Gemini models provide reasoning to keep a simulation coherent over time.

Users steer a Genie 4 world with standard keyboard controls for movement and camera, plus text prompts that can change the weather, move the camera, or add new people or objects on the fly. DeepMind says Gemini also monitors the scene between user inputs to keep it progressing naturally. The company highlights advances in simulating physics, such as collisions and aerodynamics, and especially in simulating people, who it says move and react more naturally than before.

For more complex, multi-step scenarios, DeepMind pairs Genie 4 with Gemini 3.8 Flash acting as a guiding agent: it breaks a goal into steps, checks incoming frames to track progress, and adapts the plan as the world or the user changes things.

Genie 4 is rolling out as a research preview to trusted testers first, starting with Google AI Ultra subscribers in Project Genie and early access partners of the Gemini API and Google AI Studio, ahead of a wider consumer and developer release.

Why it matters

A capable, controllable world model like this could become a practical training ground for robots and AI agents, and a new kind of interactive media format for developers to build on.

Source: Google DeepMind