Marble, the multimodal world model from Marble Labs, is now generally available, marking a significant step toward accessible spatial AI. The platform moves beyond simple text and image-to-3D generation, offering robust tools for iterative editing, expansion, and composition of virtual environments. This release positions Marble as a serious contender in the race to build foundational models capable of understanding and simulating the physical world.
The core advancement lies in Marble’s expanded multimodality. Users can now generate 3D scenes not just from text or single images, but also from multiple input views or video, allowing for far greater control over the resulting geometry. Crucially, the introduction of "Chisel," an AI-native 3D sculpting mode, decouples scene structure from visual style. Users can define the coarse layout using basic 3D primitives or imported assets, and then apply a text prompt to dictate the aesthetic—a powerful workflow for designers needing structural fidelity alongside creative flair.
