Google DeepMind has introduced Genie 4, a research-preview world model that generates interactive environments with more realistic visuals, synchronized audio and real-time scene controls.
Users can alter a simulation while it is running by changing elements such as the weather or camera angle, or by adding people and objects. Genie 4 can keep a scene consistent for up to two minutes--twice as long as its predecessor--and offers improved physics and more lifelike characters. Its generated sound can include music, ambient effects and interactive dialogue.
From simulation to creative prototyping
Google's Gemini model can help plan actions and assess frames while Genie 4 generates the simulated world. In a research experiment connected to Gemini Robotics, a model divided tasks into steps that Genie 4 then simulated in real time, pointing to a potential way to explore robot scenarios before they are attempted in physical settings.
Possible applications include games, interactive learning and robotics. The ability to make changes as a world unfolds could also support early creative work, such as quickly exploring ideas and storyboards for films.
Genie 4 is currently available to trusted testers. Google plans to extend access through Project Genie and selected Gemini API and Google AI Studio partners. As interactive simulations become easier to shape, they could offer creators and researchers a more immediate way to test ideas before bringing them into the real world.
