Google’s Genie 4 Adds Audio And Real-time World Editing
Google’s new world model generates more realistic interactive scenes that users can change as they run, though access is limited and simulations can still lose consistency.
Google has introduced Genie 4, a world model that creates interactive AI-generated environments with more realistic visuals, accompanying audio and the ability to change a scene while it runs. The model is launching as a research preview for trusted testers, according to TechCrunch and The Verge.
Table of Contents
What changes in Genie 4
Users can enter natural-language prompts during a simulation to alter the experience in real time. TechCrunch reports that they can change the weather or camera angle, or add people and objects. The model can also generate audio that fits the scene, including music, ambient sound and dialogue.
Both outlets say Genie 4 can keep an experience running for about two minutes, compared with roughly one minute for Genie 3. TechCrunch also reports improvements in physics and the appearance of characters. Those advances do not eliminate consistency problems: Google cautioned, according to both reports, that details and objects may shift as a simulation continues. TechCrunch says accurate audio and readable text within generated worlds remain challenges, too.
How Google is positioning the model
Google is presenting Genie 4 as a way to explore game ideas, plan scenes for films and create simulated situations for robotics, the outlets report. TechCrunch says trusted testers have pointed to its potential for early film work, when creators need to try ideas and storyboard quickly. These are proposed uses rather than established end-to-end products; a Google DeepMind engineer told TechCrunch that the eventual product experience is still unclear.
Gemini plays a role in the new model, although the reports describe different details. The Verge says Genie 4 uses Gemini 3.8 Flash to help with simulations. TechCrunch describes Gemini as planning steps and checking frames while Genie produces visuals. It also reports a separate research experiment in which a Gemini 4 Argon model broke tasks into steps that Genie 4 simulated in real time, suggesting a possible approach to testing robot scenarios before trying them outside a simulation.
Access and potential uses
For now, access is limited to trusted testers. TechCrunch says Genie 4 is due in Project Genie for Google AI Ultra subscribers in the coming weeks, as well as for early-access partners using the Gemini API and Google AI Studio. The Verge says Google describes subscriber access as coming soon and has opened a waitlist for people seeking to try the model in Project Genie.
For professionals working on games, film production or robotics, the proposed applications center on faster prototyping and simulated testing. Whether Genie 4 can support those workflows reliably remains an open question given its limited release and the consistency issues Google has acknowledged.
Sources
This story was compiled by AI from the reports below. Read the originals for the full details.