9.1 Understanding the LCC Studio Pipeline
Meet the LCC software family, then follow a job from field capture to a finished, shareable 3DGS model: the reconstruction modes, which device feeds each one, and the hardware that sets processing time.
What LCC Is
LCC, short for Lixel CyberColor, turns the raw LiDAR and image data from an XGRIDS scanner into a photorealistic 3D Gaussian Splatting model that you can navigate, edit, measure, export, and publish as a web link. The desktop application that runs the reconstruction is LCC Studio, and it works locally on a Windows workstation.
LixelStudio, covered in Module 8, produces point clouds for survey and BIM. LCC produces 3DGS models for visualization, virtual tours, and client delivery. Many jobs run through both. This page is the map of the module, and each stage below links to the page that covers it in depth.
The LCC Software Family
LCC is not one program. It is a small family of applications that hand work down a chain: reconstruct the model, edit it, view it, and, when the deliverable is BIM, model from it. Knowing which app owns which job is the fastest way to avoid doing the right work in the wrong tool.
| Application | What it does | When to use it | Who it is for |
|---|---|---|---|
| LCC Studio | Reconstructs raw scanner data into a 3DGS model, manages your model library, publishes shareable links, and batch-converts finished models between formats through its Converter. | First, on every job. Import capture data, pick a reconstruction mode, and process. | Anyone running a capture-to-model job. |
| LCC Scene Editor | Builds the experience around a finished model: viewpoints, guided tours, annotations, measurements, floor-plan layouts, and scene reports. It arranges the scene and does not alter the underlying points. | After reconstruction, to turn a raw model into a client-ready walkthrough or presentation. | Whoever prepares the deliverable. |
| LCC Model Editor | Works at the point level: select, crop, separate, clone, and color grade the Gaussian model itself. | When the model data needs cleanup, stray points removed, a section isolated, or color balanced. | Whoever cleans and finishes the model data. |
| LCC Viewer | Opens a published LCC model in a web browser for navigation, viewpoints, and basic measurement. No install, no editing. | To share a finished model by link, or to host it on your own site for clients and reviewers. | Clients, reviewers, and anyone you send a link to. |
| LCC for Revit | Brings an LCC model into Revit as a live modeling reference, with real-time sync, AI-assisted levels, walls, doors, and windows, semantic point-cloud overlay, and issue tracking. | When the deliverable is a BIM model built from the scan. | BIM modelers doing scan-to-BIM in Revit. |
LCC Studio comes first and holds the model library. From there the work forks by intent: the Scene Editor shapes what a viewer experiences and leaves the points untouched, while the Model Editor edits the points themselves. Both open from the same library and share one sign-in, so you move between them without re-importing.
Stay on the current release. XGRIDS adds device support, reconstruction improvements, and export formats between releases, and an older build can behave differently or miss a feature a job needs. Download the latest build from xgrids.com/intl/support/download. One installer covers LCC Studio, the Scene Editor, and the Model Editor. LCC for Revit installs separately as a Revit plugin, and the LCC Viewer runs in the browser with nothing to install.
The Pipeline, End to End
Every LCC Studio job follows the same five stages. The work you do in the field determines what is possible at every stage after it, so the pipeline starts on site, not at the workstation.
| Stage | What happens | Covered on |
|---|---|---|
| 1. Capture | Scan the site with the technique the job requires: a single continuous scene, multiple segments for fusion, or a ground scan paired with a drone flight. | Modules 3 to 7; and 9.2 to 9.4 for the field steps each mode needs. |
| 2. Create and import | Connect the device or point LCC at the capture folder, choose the reconstruction mode, and let LCC read the data and detect the device type. | 9.2 to 9.4, and 9.8 for mode selection. |
| 3. Reconstruct | Set the parameters, preview the point cloud, and run the reconstruction. This is the long, compute-heavy step. | 9.8 for settings; 9.2 to 9.5 per mode. |
| 4. Edit | Clean, crop, color grade, measure, and annotate. Scene-level work and point-level work use two different editors. | 9.9 Scene Editor Tools and 9.10 Model Editor Tools. |
| 5. Export and publish | Export to the delivery format the downstream tool needs, publish a shareable web link, or hand over the underlying Creator Data assets. | 9.7 Export Formats. |
The field stage sets the ceiling. Georeferencing, floor plans, fusion connections, and the aerial link all depend on data captured in the field. None can be added at the workstation. If a job needs any of them, plan the capture for it, because a scan that missed them has to be re-flown or re-walked.
The Five Reconstruction Modes
You pick the mode when you create the project. It is fixed at reconstruction and cannot be changed without reprocessing.
| Mode | Input | Use it for |
|---|---|---|
| Single Model | One continuous scan from one device. | The standard starting point. Any single scene that fits one session. |
| Map Fusion | Up to 10 scan segments from the same device type, 200 minutes combined. | Sites too large for one battery or one session. |
| Aerial-Ground Map Fusion | Ground scan segments plus a drone image set, linked by RTK or by control points depending on the device. | Sites that need roof and aerial context joined to ground and interior detail. |
| Aerial Reconstruction | Drone photos alone, no ground scan. | Ultra-large sites captured from the air only. |
| Video Reconstruction | One MP4, MOV, or WEBM video. | Building a model from video footage, with frames extracted automatically. |
Which Device Drives Which Mode
The scanner-driven modes support the L2 Pro, K2, and PortalCam. Aerial Reconstruction and Video Reconstruction are driven by their input, a drone image set or a video file, not by a specific scanner.
| Mode | L2 Pro | K2 | PortalCam |
|---|---|---|---|
| Single Model | Yes | Yes | Yes |
| Map Fusion | Yes | Yes, up to 10 sub-scenes | Yes |
| Aerial-Ground Map Fusion | Yes, ground data | Yes, ground data | Yes, ground data |
Feature support by device. HD Enhancement and Spatial Recognition are available on the L2 Pro and PortalCam. On the L2 Pro, the 16-line and 32-line versions cannot be fused together.
What every aerial-ground job needs
Aerial-Ground Map Fusion always takes three things captured in the field: the ground scan, the drone mission photos, and the vertical bridge photos flown at each marked takeoff and landing point. That holds on every scanner. The bridge photos are what tie the aerial data to the ground data, and none of the three can be added afterward. The full capture procedure is on 9.4 Aerial-Ground Fusion.
What does vary by device is how the drone's positioning data pairs with the ground scan, and how the photos are organized for upload.
| Ground data | Drone data | Result |
|---|---|---|
| L2 Pro with RTK | Drone with RTK | Recommended. The highest-accuracy pairing. |
| PortalCam with fusion control points | Drone with RTK | Recommended. |
| PortalCam with fusion control points | Drone without RTK | Recommended. Lowers the drone requirement and allows a consumer drone. |
| L2 Pro with RTK | Drone without RTK | Avoid. Reconstruction may fail, or the model may separate into layers. |
An RTK ground scan does not rescue a non-RTK drone. Pairing an L2 Pro RTK scan with a drone that has no RTK is the one combination XGRIDS advises against, and the failure often appears as layer separation rather than an outright error, so it can survive a quick review and reach the client. The supported route for a drone without RTK pairs PortalCam ground data with its fusion control points.
Drone RTK toggle. A switch in the reconstruction parameters controls whether the drone's own RTK data is used, and it is on by default. Leave it on when the drone recorded RTK. Turn it off when it did not, which is what makes the consumer-drone path above work. This is separate from the bridge photos, which are captured on every job regardless.
Settings That Recur Across Modes
A few settings appear in more than one mode and shape the result. The full settings reference is on 9.8 LCC Studio Tools; these are the ones worth knowing before you start.
- Reconstruction Quality. Fast, Standard, or Slow. Standard for client work, Slow for final detail-critical models, Fast for quick previews. Slow raises detail and GPU memory use and takes the longest.
- Coordinate System Conversion. Georeferences the model when the scan carries RTK data. It decides automatically: with valid RTK it converts to the target system, and without it the model still builds in local coordinates. Leave it on None indoors.
- Maximum Gaussian Points. In Single Model it caps the model's total points; keep it within GPU memory, usually 25 million or below. In the fusion modes it applies per block and the system auto-adjusts, so higher values have little effect.
- Drone RTK data. In the aerial modes, controls whether the drone's RTK is used in reconstruction. Match it to what the drone actually recorded, and see the combinations table in Section 5 before choosing.
- Layering Optimization. Corrects loop-closure drift that shows as misalignment or a doubled wall on a large loop path. It applies to PortalCam captures and is on by default. Turn it off for repetitive layouts such as identical floors or cubicles, where it can misread one place as another.
- LCC2 output. The current format is smaller, loads faster, and is the native editable format of the Model Editor. Legacy LCC and PLY upgrade to LCC2 automatically on import. See 9.7 Export Formats.
Hardware, and What It Buys You
LCC reconstruction is compute-heavy and roughly split between the GPU and the CPU. GPU memory and system RAM set the ceiling on how large a job you can process. Match the tier to the work.
| Tier | Use for | Specification |
|---|---|---|
| Basic | Single models, standard scale | 8-core CPU (Core i7 10700K or Ryzen 7 3700X), 64 GB RAM, RTX 3060 12 GB or Quadro RTX A4000 16 GB, 120 GB free. |
| Recommended | Single models, larger and higher quality | 16-core CPU (Ryzen 9 7950X), 64 GB or more, RTX 4090 24 GB (or RTX 6000 Ada 48 GB / L40 48 GB), 240 GB free on SSD. |
| Fusion | Map Fusion and Aerial-Ground | 16-core or higher CPU (Ryzen 9 9950X), 64 GB DDR5 minimum, or 128 GB for large data, RTX 3090 minimum, RTX 4090 or 4090D preferred. |
RAM sets the size of job a workstation can hold. Map Fusion needs 64 GB as a minimum, and a job at or above 150 minutes of combined capture run at high quality may need 128 GB. If the machine is short on RAM, drop Reconstruction Quality to Standard, which lowers the memory demand. Single-scene length is governed by its own rule, on 9.2 Single Scene. HD Enhancement and Spatial Recognition additionally need more than 8 GB of GPU memory to run at all.
Reserve at least twice the capture size in free space. Keep it on an SSD separate from the LCC install directory. A reconstruction can fail partway for lack of room, and the higher-end GPU mainly buys speed, not the ability to hold a larger job. Memory is what holds the job.
©2026 Alpine Reality Capture LLC • XGRIDS Pro Guide™ • Site Disclaimer