Google (opens in new tab) DeepMind announced Genie 4 on Tuesday, a world model that generates interactive, explorable environments from prompts, and released it in research preview to a limited group of trusted testers. Google (opens in new tab) says users can steer a scene in real time with natural language, such as moving the camera, changing the weather or introducing new characters, and that every world generates audio on the fly, including ambient sounds, music and interactive dialogue.
The Verge (opens in new tab) reported that Google (opens in new tab) lists four main updates over Genie 3: more realistic worlds, real-time edits by natural-language prompt, audio, and use of Gemini 3.8 Flash to help with simulations. Experiences run about two minutes, up from one minute for Genie 3, according to The Verge (opens in new tab) and TechCrunch (opens in new tab). Ars Technica (opens in new tab) reported that Genie 4 worlds still stay consistent for only a couple of minutes before they forget how things started.
Google (opens in new tab) says simulating the world is “a crucial step toward AGI, enabling safe training environments for robots and agents, and amplifying creative tools across industries like filmmaking, journalism, gaming, and manufacturing.” The Verge (opens in new tab) said Google (opens in new tab) is pitching the model for film pre-visualization, prototyping game ideas and creating training scenarios for robots.
Google (opens in new tab) acknowledged limits. As simulations run longer, details and object permanence can shift, The Verge (opens in new tab) reported. TechCrunch (opens in new tab) reported that Google (opens in new tab)’s team also has not worked out how to ensure the audio will be accurate or any text inside a simulated world will be legible, and said it is not yet clear what the end-to-end product will look like.
On access, Google (opens in new tab) says Genie 4 is “coming soon” to Google (opens in new tab) AI Ultra subscribers in Project Genie and to early access partners in the Gemini API and Google (opens in new tab) AI Studio. TechCrunch (opens in new tab) reported the rollout would come in the coming weeks, while Ars Technica (opens in new tab) said there is no specific timeline. Ars Technica (opens in new tab) reported that Ultra plans start at $100 per month and that pricing for developers is not yet available. A waitlist is open for people who want early access, The Verge (opens in new tab) said.
The Verge (opens in new tab)’s reporter said he had not been able to try Genie 4, and the examples in his report came from Google (opens in new tab).