Google DeepMind released a research preview of its Genie 4 world model on October 27, 2026. It creates interactive environments with sound and lets users change scenes with text prompts during a simulation. It is currently available to selected trusted testers.
Real-time scene edits and sound
Compared with Genie 3, the new model offers more photorealistic visuals and adds sound. It generates audio in real time to match the simulated environment, including music, ambient noise and interactive dialogue.
Users can edit a scene as it runs, using text prompts to change the weather or camera angle, for example, or to add people and other objects.
Two-minute scene consistency and Gemini integration
Genie 4 can maintain a consistent scene for up to two minutes, twice the limit of its predecessor. It also simulates physics better and creates more human-like characters.
The model is more closely integrated with Gemini, which acts as a director: it plans individual steps and checks frames while Genie creates the visuals.
The development team cautioned that details and objects can change during longer simulations. It has also yet to resolve how to ensure accurate audio and readable text within generated environments.
Robot scenarios and film prototypes
In a research experiment tied to Gemini Robotics, Gemini 4 Argon broke tasks into individual steps that Genie 4 simulated in real time. The experiment points to a possible way to test robot scenarios in simulation before carrying them out in the real world.
Film prototyping is another proposed use. Google DeepMind engineer Diego Rivas said working in real time could help with rapid ideation and storyboarding. He also stressed that development remains at an early stage and the final shape of the product is unclear.
Who gets access
Following its initial release to selected trusted testers, Genie 4 is set to reach Google AI Ultra subscribers in Project Genie in the coming weeks. The Google tool creates interactive worlds from text or image prompts.
The model is also set to reach partners with early access to the Gemini API and Google AI Studio.
