LCC2 Format Explained: XGRIDS's Next-Gen 3DGS Container for AEC and Surveying

XGRIDS just released the LCC2 Data Organization Format — a Level-of-Detail-native, open-specification 3D Gaussian Splatting container built for large-scale AEC, surveying and digital twin workflows. Full technical breakdown from Australia's XGRIDS National Distributor.

XGRIDS has released the LCC2 Data Organization Format White Paper — and for anyone shipping 3D Gaussian Splatting on real construction, survey or digital twin projects, this quiet release is one of the most consequential 3DGS updates of 2026.

"LCC2 is designed for how 3D Gaussian Splatting actually gets used in production — large scenes, long iteration cycles, multi-team workflows."

What Is LCC2?

Lixel CyberColor 2 (LCC2) is the next-generation data organization format for 3D Gaussian Splatting developed by XGRIDS, the global leader in SLAM-based 3DGS capture technology. It is the successor to the LCC (.lcc) format that has powered XGRIDS's Lixel CyberColor ecosystem since 2024.

Unlike LCC1 — which was a tightly-coupled proprietary binary container designed for single-scene reconstruction — LCC2 is built around a Level-of-Detail (LoD)–native N-ary tree structure, with a decoupled storage layer that supports multiple splat formats (.ply, .spz, .sog), co-located mesh data, and BVH (Bounding Volume Hierarchy) support for ray-cast workflows.

LCC1 vs LCC2: Ten Key Differences

• →Storage architecture — LCC1: tightly-coupled binary decoder. LCC2: decoupled data organization + storage layers.
• →Level of Detail — LCC1: baked into storage. LCC2: first-class N-ary tree, independent of splat format.
• →Splat formats — LCC1: single proprietary binary. LCC2: .ply, .spz, .sog within the same LCC2 folder.
• →Mesh co-location — LCC1: external, separate pipeline. LCC2: native data/mesh/* alongside data/3dgs/*.
• →BVH support — LCC1: not supported. LCC2: optional native ray-cast acceleration structure.
• →Scene editing — LCC1: heavy refactor for edits. LCC2: designed for long-term iteration.
• →EPSG coordinate — LCC1: external handling. LCC2: first-class metadata field.
• →Semantic attributes — LCC1: ad-hoc extensions. LCC2: splatExtraAttributes as first-class extension point.
• →Version discipline — LCC1: implicit. LCC2: explicit version field with forward-compatible reader rules.
• →Licence — LCC1: proprietary. LCC2: open specification, derivatives must be open-sourced.

LCC2 Architecture in Plain English

An LCC2 dataset is a folder, not a single file. The root contains a scene.lcc2 JSON metadata file, plus a data/ directory split into 3dgs/ (required, containing the splat files in .ply, .spz, or .sog format) and mesh/ (optional, containing triangular mesh files).

The scene.lcc2 JSON file is the scene brain. It describes scene metadata (name, description, GUID, data source, EPSG coordinate code), LoD hierarchy (how many levels, how many Gaussians per level, total splats), storage configuration (splat file type, rendering hints), and the N-ary tree of nodes where each node has an axis-aligned bounding box and points to data in /data/3dgs/ and /data/mesh/.

Node Structure

• →Root node — holds global file lists (splatFiles[], meshFiles[], bvhFiles[]) and scene-level environment data
• →Branch node (childNum > 0) — contains a child{} object with serialised children (keyed 0, 1, 2…)
• →Leaf node (childNum = 0) — contains the actual data references (data.3dgs required, data.mesh and data.bvh optional)

This is the same hierarchical pattern used by OGC 3D Tiles, Cesium, NVIDIA Omniverse's USD scene graph, and Unreal Engine's World Partition — which is precisely why LCC2 integrates cleanly into those pipelines.

Why LCC2 Matters for AEC and Surveying Workflows

1. Level-of-Detail Streaming for Large Scenes

If you're scanning a 10-hectare construction site, a 5-kilometre infrastructure corridor, or a multi-storey facility, you cannot load every Gaussian at once. LCC2's LoD tree lets a viewer or renderer stream only the detail level required for the current camera distance — enabling site-scale digital twins that render in a browser or on mobile devices without requiring a GPU workstation.

2. Mesh Co-Located with 3DGS

A single LCC2 container can carry both 3DGS for photorealistic visualisation and mesh for measurement, physics simulation, collision detection, and CAD export. Real AEC deliverables are not just pretty — they need to support measurement tools, clash detection, and engine physics. LCC2 eliminates the need to juggle parallel file trees for the same scene.

3. EPSG as First-Class Metadata

LCC2 has an explicit epsg integer field in the scene metadata. In Australia, that means your 3DGS can carry MGA Zone 54/55/56 or GDA2020 coordinates directly into Esri ArcGIS, Autodesk Civil 3D, or Cesium — without a separate georeferencing transformation step.

"For council and government work — where deliverables live in Esri or Autodesk ecosystems — this is the difference between a demonstration and a production deliverable."

4. Semantic / Thermal / Instance Attributes

The splatExtraAttributes field is the sleeper feature. It allows LCC2 to carry instance IDs (each door, window, pipe or asset as a tagged object), semantic classification ("this Gaussian is a wall / floor / rebar / ceiling"), and thermal values (infrared overlay baked directly into the splat data).

For Scan-to-BIM workflows, this is the bridge from "we have a pretty point cloud" to "we have a semantic model ready for automated Revit export." XGRIDS's LCC4Revit plugin — and AECT's AI-driven Scan-to-BIM workflow — both benefit directly from this extension point.

5. Open Specification, Open Tooling

The LCC2 licence requires derivatives of the Data Organization Format to be open-sourced. That is a significant commitment from XGRIDS and — importantly — a strong signal to ecosystem partners (Unity, Unreal, Cesium, NVIDIA Omniverse, Autodesk, Esri) that LCC2 is safe to build against.

For Australian clients asking the "what happens if XGRIDS disappears?" question, this answers it directly: the format is documented, the licence is permissive, and derivative tooling must remain open. That is not something Matterport or most proprietary 3D capture formats can claim.

What This Changes for AECT Clients in Australia

If you already run XGRIDS L2 Pro, K1, or PortalCam with LCC Studio

• →LCC1 files continue to work — no forced migration required.
• →LCC Studio v1.12.0 (Premium tier) now exports .lcc2 alongside .lcc.
• →For new long-lifecycle projects — site-scale digital twin, multi-phase construction QA, council asset management — we recommend exporting both formats.

If you are evaluating 3DGS pipelines against Matterport or proprietary alternatives

• →LCC2's EPSG-native, LoD-native, mesh-co-located architecture narrows the gap between proof-of-concept 3DGS demos and production AEC deliverables.
• →Versus Matterport or closed formats: LCC2 is a documented open specification.
• →Versus raw .ply or research-code 3DGS: LCC2 has the production plumbing (LoD, metadata, versioning) already built in.

What AECT Solutions Is Doing With LCC2 Right Now

As XGRIDS's Australian National Distributor, we are running three parallel initiatives in production:

• →ArcGIS Online integration pilot — leveraging the native EPSG field to push LCC2 scenes directly into Esri Scene Viewer with no intermediate transformation layer.
• →Scan-to-BIM pipeline upgrade — experimenting with the splatExtraAttributes semantic field for automated Revit family recognition.

We will publish benchmarks and integration findings in the coming weeks. If you are working on a large-scale or long-lifecycle 3DGS project in Australia and want early access to the LCC2 workflow, get in touch.

Frequently Asked Questions

What is the LCC2 format?

LCC2 (Lixel CyberColor 2) is XGRIDS's next-generation 3D Gaussian Splatting data organization format. It is an open specification designed for large-scale, high-complexity spatial data using a Level-of-Detail (LoD) N-ary tree structure. LCC2 decouples data organization from storage format and supports .ply, .spz, and .sog splat containers, mesh data, and BVH structures within a single folder-based container.

How is LCC2 different from LCC1?

LCC1 used tightly-coupled binary storage with LoD baked into the format. LCC2 separates those layers: Level-of-Detail becomes a first-class N-ary tree structure, storage formats are swappable, mesh and BVH data live inside the same container, and the entire specification is published openly. LCC2 is designed for long-term iteration and large-scene editing — use cases where LCC1's tight coupling became a bottleneck.

Is LCC2 production-ready?

LCC2 is currently at version 0.0.3 (Beta) as of April 2026. It is production-ready within XGRIDS's own ecosystem (LCC Studio v1.12.0 Premium) and forward-compatible by design. Third-party integrations (Unity, Unreal, Cesium, ArcGIS) are catching up. For production-critical workflows today, exporting both .lcc and .lcc2 is recommended.

Can LCC2 be used for Scan-to-BIM workflows?

Yes. LCC2's splatExtraAttributes field supports semantic, instance, and thermal classification — enabling automated identification of walls, doors, windows, and other building elements. Combined with XGRIDS's LCC4Revit plugin, this accelerates the path from raw scan to BIM-ready model. AECT Solutions offers AI-driven Scan-to-BIM services leveraging this workflow.

Is LCC2 better than Matterport for real estate or AEC?

LCC2 offers photorealistic 3D Gaussian Splatting rendering (versus Matterport's mesh-based approach), native EPSG georeferencing, open file format, no hosting subscription lock-in, and Apple Vision Pro / WebXR support. For AEC, surveying, and technical deliverables, LCC2 is substantially more capable than Matterport's proprietary ecosystem.

Where can I read the LCC2 specification?

The full LCC2 Data Organization Format White Paper is published at developer.xgrids.com as of April 2026.

See LCC2 Running on a Real Australian Project

If you want to see LCC2 working on a live Australian project — or understand how it changes your current 3DGS workflow — AECT Solutions is running live demonstrations and workflow walkthroughs.

• →Sydney Build Expo · Booth C-19 · 29–30 April 2026 — PortalCam, K1, and L2 Pro running live with LCC Studio v1.12.0 demonstrating LCC1/LCC2 export side-by-side.
• →3DGS Masterclass · Launching June 2026 — 3-day hands-on training covering capture → LCC2 processing → publishing to Esri, Unity, Unreal, and Omniverse pipelines.
• →One-on-one technical walkthrough — 30-minute video call. Book via the contact page.

"The gap in the Australian AEC market isn't capability. It's workflow and trust. LCC2 is architected for the integration points Australian AEC actually demands."

Talk to Our Team

Every article is backed by real project experience. Book a discovery call and let's discuss your specific project.