World Labs, an innovative startup specializing in artificial intelligence, introduced its first AI system on Monday. This yet-to-be-named system is capable of creating interactive 3D environments from image inputs. Users can navigate these generated worlds, transforming a static 2D visual asset into a dynamic 3D scene using a keyboard and mouse. Currently, the AI is in preview mode and has not yet been released to the public. The company, founded by noted computer scientist Fei-Fei Li, has indicated plans for a full launch in the near future.
World Labs Introduces AI with 3D World Generation Abilities
In a recent blog post, the San Francisco-based World Labs detailed the features of its AI model. The company pointed out that while many generative AI tools are capable of producing 2D content, such as images and videos, the ability to create 3D models from 2D images or text inputs remains quite limited. Recently, only Google DeepMind has released an AI model specifically designed to create distinctive 2D video game levels.
However, according to the interactive examples shared by World Labs, the unnamed AI system significantly enhances the generative capabilities currently available. Users can input any image representing a scene, and the AI can then create an interactive 3D version of that scene. This functionality allows users to explore the generated environment by moving in various directions.
Notably, the AI not only reproduces three-dimensional representations of the objects featured in the input image but also invents additional details, including hidden pathways, decorative ceiling elements, and new objects. World Labs asserts that aside from the original image, all other elements are entirely generated by the AI system.
Moreover, users have the ability to modify the generated scenes. They can adjust camera angles, depth, and zoom levels, along with adding 3D effects to both the background and foreground elements.
The AI system from World Labs is designed to integrate with other AI tools, enabling creators to generate an initial image using established text-to-image generators like Ideogram, DALL-E, or Midjourney before transforming it into a 3D world with this latest tool. The startup is currently collaborating with a select group of creators to beta test the system’s capabilities and refine its 3D-native generative workflow.
As of now, access to the AI system is restricted, and World Labs is actively working on enhancing the size and detail of the generated worlds. Interested parties are invited to join the company’s waitlist to receive updates on the forthcoming release of the AI system.
