World Labs Unveils Marble: A New Step Toward Persistent 3D Worlds
Translate this article
World Labs has announced Marble, a limited access beta preview of its latest model for generating persistent, navigable, and controllable 3D worlds. Available at marble.worldlabs.ai, the platform allows users to create and explore 3D environments directly in their browser.
Unlike earlier models that often struggled with consistency or imposed time limits, the new system generates larger, stylistically diverse spaces with cleaner geometry. According to World Labs, these worlds support free viewpoint navigation and reveal detail beyond the original input image or prompt. The model is focused on building coherent environments rather than isolated objects.
Export and Integration
Users can export generated scenes as Gaussian splats, making them usable in downstream projects. Integration is supported through Spark, the company’s open-source rendering library, which connects seamlessly with Three.js. Spark is optimized to render efficiently across desktops, laptops, mobile devices, and VR headsets.
Style Versatility
Early testers highlight the model’s ability to handle multiple styles with ease from hyper-realistic landscapes to anime-inspired or low-poly designs. Examples shared include cartoon, sci-fi, fantasy, and realistic interiors, with consistency across compositions enabling creators to stitch together larger scenes.
Early Use Cases
Feedback from beta users underscores the tool’s range of applications: filmmakers planning sets, VR creators seeking interactive environments, and game developers experimenting with integrating Marble’s 3D outputs into interactive content pipelines. While the current focus is on environments, support for central subjects such as people or animals is not yet included.
World Labs frames Marble as the starting point of a longer journey. The company emphasizes that larger, persistent environments are now within reach, and invites users to experiment, share feedback, and help shape future development. As one early user noted, what once took weeks to create in 3D can now be achieved in minutes.
About the Author
Eva Rossi
Eva Rossi is an AI news correspondent from Italy.
Recent Articles
Subscribe to Newsletter
Enter your email address to register to our newsletter subscription!