Google DeepMind releases Genie 4 world model in research preview

Genie 4 adds real-time audio, more photorealistic scenes and edits made by typing prompts during a simulation. It is available to trusted testers, with a wider rollout planned in the coming weeks.

Google DeepMind has released Genie 4, its newest world model, as a research preview for trusted testers. Users can type prompts during a simulation to change the scene, and the model generates matching audio in real time, including music, ambient sound and dialogue.

Compared with Genie 3, released a little over a year ago, Genie 4 allows free-form edits such as changing the weather or camera angle, or adding people and objects. It keeps a scene consistent for up to two minutes, double the earlier limit. Gemini acts as a director that plans steps and checks frames, while Genie draws the world. DeepMind said details and objects can drift over time, and the team has not yet worked out how to ensure accurate audio and legible text inside the worlds.

In a Gemini Robotics experiment, a Gemini 4 Argon model broke tasks into steps that Genie 4 simulated, which points to a way of testing robot tasks in simulation before trying them in the real world. Genie 4 is planned to reach Google AI Ultra subscribers in Project Genie in the coming weeks, along with early access partners of the Gemini API and Google AI Studio.

This is a short summary by Zaxvix based on the source above. We write our summaries with the help of AI and check them before publishing. For the full story, read the original. Spotted an error? Email hello@zaxvix.com and we will review the article and correct or remove it if needed.