Genie 3 It was released as early as August 5, 2025, though at the time it was in the form of a research preview, provided to a limited number of scholars and creators for testing, just earlier… Google Project Genie has finally been officially launched, allowing eligible users to experience generating interactive 3D virtual environments through text or images.
Google Genie 3 Officially Launches: Ushering in a New Era of Interactive 3D Virtual Worlds
Genie 3 is a general-purpose world model whose core function is transforming text or image prompts into interactive 3D virtual environments. Unlike AI video generation tools like Sora or Veo, Genie 3 doesn’t generate fixed videos—it creates “living” worlds that users can explore freely like playing a game.
The interface will have two prompt sections—one for environment and one for character—that you can configure. For example:
- Realistic style high mountain meadow scene, scattered with colorful wildflowers, surrounded by evergreen pine trees and distant snow-capped peaks. A rustic wooden log cabin with a front porch sits beside the meadow, a path leads to the front of the cabin, wooden split-rail fences extend along the meadow. Warm afternoon sunlight, soft shadows.
- Character prompt: “A shiba inu in the center of the frame, with a third-person game camera perspective, featuring a highly sensitive and responsive control feel.”
The system generates a dynamic 720p 24 fps environment like the interactive 3D environment shown below, where you can control a Shiba dog to walk, turn its head, and even interact with objects.
You can also upload images for Genie 3 to generate matching interactive 3D virtual environments:
, or even creating your own virtual 3D character:
Using demo video:
Genie 3 allows users to truly “step into the world,” which not only enhances immersion but is also applicable for training AI agents (such as robots learning environmental adaptation). Foreign media outlet TechCrunch once called it “a key step toward AGI,” because it simulates real-world causal relationships, enabling AI to understand “if…then…” logic.
Currently, generation has the following limitations:
- Each interaction lasts about 60 seconds
- The resolution and frame rate are fixed at 720p / 24 fps and cannot reach 4K or higher.
Google also mentioned in their blog that Genie 3 is still an early-stage model with several known limitations:
- The generated world doesn’t necessarily fully comply with real-world physics rules, and may also have slight deviations from the prompt content.
- Character control is sometimes unstable and may experience delays.
- Some advanced features announced last year (such as the ability to trigger events during exploration and change the world) have not yet been added to this prototype.
Project Genie is currently first available to Google AI Ultra users in the US, with plans to expand to more countries and regions.
Google hasn’t mentioned any additional options, which shows that even if more countries are opened up in the near future, it’s likely still limited to AI Ultra users only, with free users, AI Pro, and AI Plus users still unable to experience it for now.
Source: KOCPC Chinese