Aligned E57 geometry
The automated service accepts E57. If the file combines several scanner positions, they should share one coordinate system and have manageable noise and overlap.
Technical guide
LiDAR to Gaussian Splat Conversion
LiDAR gives measured spatial samples. Gaussian Splats are optimized rendering primitives. Converting between them is possible when geometry is paired with enough usable appearance and viewpoint information—but not every LiDAR scan can become a photorealistic scene.
Data model
LiDAR points and Gaussian primitives describe a scene differently
| Characteristic | LiDAR / point cloud | Gaussian Splat |
|---|---|---|
| Primary purpose | Measured geometry and attributes | Real-time appearance rendering |
| Core elements | XYZ samples, intensity, class and optional RGB | Position, scale, rotation, opacity and appearance coefficients |
| View dependence | Usually not an appearance model | Can encode view-dependent appearance |
| Typical strength | Spatial measurement and scan alignment | Fast novel-view rendering |
Inputs
What a strong source package contains
Usable color observations
Per-point RGB helps, while source photos or panoramas can provide richer appearance if they are accessible and compatible.
Camera relationship
Known or recoverable camera viewpoints connect appearance observations to geometry. Incorrect alignment produces blur, floaters or double surfaces.
Use cases
Where a renderable result can help
A Gaussian Splat can complement—but does not replace—the authoritative LiDAR source.
Remote visual review
Share an accessible browser view with stakeholders who do not need the complete measurement dataset.
Documentation and context
Preserve the visual context around an as-built, construction or heritage capture when the source includes suitable imagery.
Creative and spatial workflows
Use an inspected result in visualization, virtual production or interactive experiences that support the chosen output format.
Limitations
Why output quality varies
LiDAR accuracy does not automatically translate into photorealistic rendering quality.
- Uncolored LiDAR lacks photographic appearance, even if its measurements are precise.
- Static-scene methods struggle with people, vehicles, vegetation movement and capture changes.
- Glass, mirrors, water and highly reflective finishes can create inconsistent observations.
- Missing viewpoints and occlusions remain missing; the result must be checked from the angles that matter.
- Gaussian output is not a certified survey deliverable and should not replace the source measurement data.
Outputs
Keep both the source and the renderable asset
Gaussian PLY
A tool-oriented container when the target software understands the particular Gaussian attributes written by the pipeline.
SPLAT
A compact delivery representation for compatible viewers. It is useful for playback but may not preserve every editable attribute.
Have a Matterport Pro3 E57?
Follow the dedicated Matterport export workflow from upload through browser preview.
Matterport Pro3 E57 to Gaussian SplatFAQ
LiDAR conversion questions
Is LiDAR data the same as a Gaussian Splat?
No. LiDAR measures spatial samples, often with intensity or color. A Gaussian Splat represents a scene with oriented Gaussian primitives whose appearance is optimized for rendering.
Can LiDAR alone look photorealistic?
Not reliably. Geometry can provide a strong initialization, but photographic appearance normally depends on usable RGB observations, viewpoint coverage and camera calibration or poses.
Which LiDAR format is accepted?
The automated e57.gs workflow accepts E57 files. Other LiDAR containers are not claimed as direct automated inputs on this site.
What scenes are difficult?
Reflective or transparent materials, moving objects, weak image coverage, unregistered scans, large occlusions and noisy or sparse data can reduce quality or prevent a usable result.
Have an E57 file?
Use the automated service for a supported E57 and inspect the completed result against the original capture.
Explore E57 to Gaussian Splat