Meet LCC 2.0: World Models Get Their Photoshop

News2026.06.165 min read

The world’s first editor built for world models — making digital worlds truly usable.

For the past few years, the work around world models has centered on reconstruction. Photos become models, video becomes space, and high-fidelity 3D capture is now within reach of almost anyone. But a finished reconstruction is raw material, not a usable asset. There is stray geometry to clear, separate captures to merge, scenes to reorganize, assets to maintain and reuse.

Reconstruction is where a world model begins, not where it ends. It becomes useful only once it can be edited, refined, and put to work, until it earns its place in production as a digital asset.

That is why we are releasing LCC 2.0, with the first editor built for world models that carry collision, geometry, and depth. It gives world models a complete path from production, to editing, to application, and sets up LCC2 as the infrastructure between the physical world and spatial intelligence.

The World Model Editor: A Photoshop Moment for 3D

Photoshop turned the photograph from a record into something you could work on. LCC 2.0 does the same for captured 3D. The editor has two parts: LCC Model Editor for the model itself, and LCC Scene Editor for the experience around it.

LCC Model Editor: Edit a 3D Model as Easily as a Photo

LCC Model Editor works directly on Gaussian data. Merge multiple captures into one, duplicate content, crop and clean away clutter, color grade entire scenes, or separate individual objects.

Aerial 3D Gaussian Splatting model of a dense city block at dusk edited in LCC 2.0

The newly upgraded Crop tool delivers more precise spatial trimming, producing cleaner cuts and sharper boundaries while minimizing residual artifacts, creating a stronger foundation for model stitching and downstream editing workflows.

LCC Model Editor crop tool producing cleaner edges than other tools on a reconstructed interior

Repairing scan artifacts, removing unwanted content, or recombining captured spaces into something entirely new now happens in one place, on the Gaussian layer, without round-tripping between tools.

Reconstructed corridor lined with chairs edited on the Gaussian layer in the world model editor

For the first time, a world model gains a true post-production workflow: editing 3D space becomes as direct as editing an image.

Night forest scene with a dinosaur skeleton exhibit as an editable 3D capture

LCC Scene Editor: Design the Whole Spatial Experience

Where the Model Editor edits the model, the Scene Editor edits the scene.

LCC Scene Editor interface editing a baroque garden fountain captured in 3D

Add annotations, group and manage assets, configure interaction logic, and set flythrough routes.

Living room digital showroom scene with annotations and interaction points set in LCC 2.0

In a digital showroom, a cultural exhibition, a training program, or an industrial simulation, you are designing a space people can walk into, explore, and act on.

Dining set highlighted while grouping assets in a navigable 3D interior scene

LCC2: The Foundation That Makes It Scale

As world models grow, how their data is organized decides what is possible. A single campus, city block, or working factory can hold tens of millions, even hundreds of millions, of Gaussian points. LCC2, the new format underneath LCC 2.0, is built for that.

Worlds Without Limits

LCC2 rebuilds the data structure and compression stack from the ground up, preserving fine detail while dramatically reducing storage and transmission costs. It breaks large-scale world models into schedulable units and streams them based on viewpoint and distance: distant regions retain structure and silhouette, while nearby regions progressively reveal detail. With its dual-track LOD architecture, model scale is no longer a constraint—from localized spaces to world-scale environments, everything can be loaded, managed, and rendered within a single system, enabling smooth real-time performance at virtually any scale.

Flythrough zooming into fine surface detail of a large-scale model using LOD streaming

The Physical World, Inside the Model

LCC2 stores more than the Gaussians. It holds colliders, geometry, and depth in the same place, so a scene can be analyzed by software, read by AI, and sensed by a robot, not only rendered. That is what takes a captured space from something you look at to something a machine can reason about.

Quadruped robot walking through a reconstructed street scene that carries collision and depth

No Model Is an Island

LCC2 imports PLY, SPZ, and SOG and converts cleanly to and from them, so existing assets join the workflow without re-capture.

Indoor scene with Olympic rings imported from PLY and SPZ into the LCC2 format

It is open source, with a public spec and developer docs.

Aerial digital twin of a campus with location labels in an open 3D world model

A world model proves its worth by circulating: the same asset opening in different software and running on different platforms, across digital twins, film, XR, robotics, and AI.

From Looking Good to Working

LCC 2.0 moves world models past the demo stage: scenes you can edit like an image, manage like a database, and move between systems like content on the web. That is the shape we think content infrastructure for spatial intelligence should take.