Google Unveils Genie 4, Longer Interactive Experiences

Google officially announces Genie 4, an AI model designed to generate longer and more realistic interactive experiences. The launch will begin with a research preview reserved for trusted testers, followed by an announced availability for subscribers of Google AI Ultra. A waitlist is now open. However, Google warns that the stability of details may evolve during extended simulations.

Restricted Access Initially and Stability Warning

Genie 4 will first be offered as a research preview for trusted testers. Google plans to make it "available soon" for subscribers of Google AI Ultra. A waitlist is accessible for those who wish to try the Genie project. The company clarifies that when simulations are prolonged, the details and permanence of objects may change over time.

Announced Features: Audio and Real-Time Editing

Google states that Genie 4 allows for the creation of more realistic worlds and enables modifications to an experience in real-time through natural language instructions. The model is also capable of generating audio. It can produce experiences lasting about two minutes, whereas Genie 3 was limited to one minute.

Embedded Technology and Targeted Uses

Genie 4 relies on Gemini 3.8 Flash to perform its simulations. According to Google, this model is suitable for previews aimed at filmmakers, prototyping game ideas, as well as developing training scenarios for robots. The company has presented examples to illustrate these uses.