Google DeepMind unveils Genie 4 with sound and world editing — TechCrunch
Google DeepMind has unveiled the Genie 4 model for generating interactive virtual worlds in a research preview. It creates synchronized audio — music, background sounds and dialogue — and allows users to enter text commands directly during a simulation to change the scene, TechCrunch reports.
Changing the scene during a simulation
With Genie 4, users can change the weather and camera viewing angle in real time, as well as add people or objects to the virtual environment. The model can maintain the consistency of a single scene for up to two minutes — twice as long as its predecessor, Genie 3. Developers also report improved simulation physics and more human-like characters.
The Gemini model has been more closely integrated with Genie 4: it plans a sequence of actions and checks frames, while Genie generates the visual component. In a research experiment related to Gemini Robotics, the Gemini 4 Argon model broke tasks down into steps, while Genie 4 simulated them in real time. This approach can be used to test scenarios for robots in simulation before they are performed in a real environment.
More current news is available on the UA.News Telegram channel Telegram.
Potential areas of application
Among the possible uses for Genie 4, Google names games, interactive learning and robotics. Testers who were among the first to gain access to the product also view the model as a tool for creating film prototypes, including during idea development and storyboard preparation.
At the same time, Google DeepMind acknowledged the model's limitations: during long simulations, details and objects may gradually lose stability, while audio accuracy and the readability of text within the virtual world still require improvement. Genie 4 is currently available to trusted testers, and in the coming weeks it is planned to be opened to Google AI Ultra subscribers in Project Genie, as well as partners with early access to the Gemini API and Google AI Studio.
