SPZ to KSPLAT Converter Online
Convert SPZ compressed Gaussian Splatting to KSPLAT format for Three.js progressive streaming.
Last updated Mar 2026
- 1Upload
- 2Convert
- 3Download
Drag SPZ file here, or click to upload
What You Should Know
When to Convert SPZ to KSPLAT
If you have a scene in SPZ format and are building a Three.js web application using the GaussianSplats3D library, converting to KSPLAT enables progressive streaming. SPZ is excellent for storage and delivery, but the GaussianSplats3D library is optimized to load KSPLAT files in sorted chunks, providing immediate visual feedback while the rest of the scene downloads. This conversion is most useful for developers embedding large Gaussian Splatting scenes in interactive Three.js applications.
Data Changes in SPZ to KSPLAT Conversion
SPZ preserves SH in its compressed form. KSPLAT supports optional SH, but whether a KSPLAT file contains SH depends on the settings used when creating it — specifically the spherical harmonics level parameter (default is off). When you convert SPZ to KSPLAT, SH from the SPZ source can be preserved if the KSPLAT creation settings include SH. When SH is included, it can be stored at 16-bit or 8-bit precision depending on the compression level.
File Size Comparison
KSPLAT is larger than SPZ but smaller than SPLAT. A 50 MB SPZ file will typically produce a KSPLAT file of 130–180 MB, compared to 200–300 MB for an equivalent SPLAT file. If file size is the priority, keep SPZ. If the target runtime is Three.js GaussianSplats3D and you want progressive loading, KSPLAT is the right output format.
SPZ vs KSPLAT
| Feature | SPZ | KSPLAT |
|---|---|---|
| File Size | Very small — ~10% of equivalent PLY | Small — ~35% of equivalent PLY |
| Spherical Harmonics | Yes — preserved through gzip compression | Optional — included only if enabled when creating the file (default is no SH) |
| Web Compatibility | Excellent — Khronos/Niantic standard | Good — Three.js GaussianSplats3D ecosystem |
| Compression | Excellent — quantization + gzip (~90% vs PLY) | Good — quantized with optional chunks |
| Progressive Loading | No | Yes — streaming-optimized |
| Typical Use | Web delivery, archiving, Scaniverse, long-term storage | Three.js integration, large scene streaming |