World Labs Unleashes ‘Marble’: The AI Revolution Generating 3D Worlds from Text and Images, Redefining Digital Creation Forever

In a groundbreaking leap forward for artificial intelligence, World Labs, a pioneering company co-founded by renowned AI expert Fei-Fei Li, has officially launched “Marble” to the public. This innovative multimodal world model is set to revolutionize how we conceive, create, and interact with three-dimensional environments, transforming simple text prompts, images, or videos into rich, explorable 3D worlds. The public release, announced on November 12, 2025, marks a pivotal moment in the pursuit of “spatial intelligence,” a concept Li believes will define the next decade of AI development.

What is Marble and Why Does it Matter?

At its core, Marble is a first-in-class generative multimodal world model designed to reconstruct, generate, and simulate 3D worlds. Unlike previous generative AI that primarily focused on 2D content like images and videos, Marble transcends these limitations by producing fully explorable, persistent, and interactive 3D environments. This capability represents a significant evolution in AI, moving beyond mere perception to a deeper understanding and generation of physical space and its inherent properties. World Labs emphasizes that just as humans integrate vision, hearing, touch, and language to build a mental model of the world, Marble is built to synthesize diverse input signals into cohesive 3D worlds.

Fei-Fei Li, often dubbed the “godmother of AI,” views spatial intelligence—the ability for AI systems to perceive, model, reason about, and interact within physical or geometric space—as the “new frontier” of AI. Marble is World Labs’ inaugural commercial product and a direct manifestation of this vision, promising to transform a wide array of industries.

Unpacking Marble’s Advanced Capabilities

Marble’s robust feature set empowers creators with unprecedented flexibility and control:

• Multimodal Input Mastery: Users can initiate 3D world generation from a wide range of inputs, including descriptive text prompts, single images, multiple images, videos, or even coarse 3D layouts. This flexibility allows for combining Marble with other AI tools, enabling users to generate images with their preferred image model and then lift them into a full 3D world.
• Interactive Editing and Expansion: Beyond initial generation, Marble provides interactive editing tools. Users can refine, extend, and combine generated 3D worlds with fine-grained control. This includes modifying objects, changing styles, or making structural alterations to the environment.
• “Chisel” – The 3D Sculpting Interface: For advanced users seeking more precise control over spatial layouts, Marble introduces “Chisel.” This experimental 3D sculpting mode allows users to design rough layouts and then apply styles through natural language prompts, effectively decoupling structure from visual style.
• High-Quality Export Options: Generated 3D worlds can be exported in various industry-standard formats, including Gaussian splats (for high-quality rendering), traditional meshes (like triangle meshes and collider meshes), and videos with pixel-level camera control. This ensures seamless integration with existing creative pipelines, game engines like Unreal Engine and Unity, and VFX software.
• Real-World Reconstruction: Marble can even take multiple photos or short videos of real-world spaces and intelligently stitch them together to create immersive, realistic virtual replicas, allowing users to explore digital versions of actual locations.

Revolutionizing Industries: Who Benefits from Marble?

The applications for Marble are vast and span across numerous sectors:

• Gaming and VR/AR: Game developers, especially indie studios, can rapidly prototype and create detailed environments, saving significant time and resources. It also offers new possibilities for building immersive virtual and augmented reality experiences.
• Filmmaking and Visual Effects (VFX): Filmmakers and VFX artists can quickly generate complex scenes, environments, and reference materials, accelerating pre-production and post-production workflows.
• Architecture and Design: Architects and designers can visualize and iterate on complex spatial concepts with unprecedented speed and realism, transforming conceptual sketches into explorable 3D models.
• Robotics and Simulation: Marble’s ability to generate predictive simulations of the real world and learn physical properties makes it invaluable for creating simulated environments for robotics training and testing.
• Education and E-commerce: Interactive 3D simulations can enhance learning experiences, while virtual try-ons and immersive product showcases could redefine online retail.

By democratizing access to 3D content creation, Marble aims to address the challenges of traditional 3D world-building, which often requires extensive time, budget, and specialized expertise.

The Road Ahead: Marble Labs and the Future of Spatial AI

World Labs has also introduced “Marble Labs,” a dedicated creative hub designed to foster collaboration and experimentation. It serves as a workspace for artists, engineers, and designers to explore workflows, case studies, tutorials, and documentation, pushing the boundaries of what’s possible with generative world models.

While World Labs has secured significant funding, including $230 million last year, and is making Marble available through freemium and paid tiers (starting from a Free tier up to Max at $95/month for extensive use), the journey of spatial intelligence is just beginning. Competitors are emerging in this space, including other startups like Decart and Odyssey, as well as research initiatives from tech giants like Google’s Genie and Nvidia’s Cosmos.

Fei-Fei Li’s assertion that spatial intelligence will be the “defining challenge of the next decade” underscores the immense potential and ongoing development in this field. Marble represents a bold stride towards an AI future where machines not only understand our language and images but also grasp the complexities of the physical world, opening up entirely new dimensions of human-computer interaction and creativity.