27 formats, and what each one actually carries.
A format list is easy to pad and hard to trust, so this one says what survives the trip. Every entry is a real codec in the application, and the note beside it is the same sentence the export dialog shows you before you commit.
Meshes
Read by first-party streaming codecs — a memory-mapped OBJ reader, cgltf for glTF, ufbx for FBX. A 9 GB OBJ reads in about 1 GB of memory, because the parser's working set is a budget rather than the file.
| Format | Direction | What it carries |
|---|---|---|
| OBJ | Read + write | Wavefront OBJ + MTL. Widest compatibility; textures written as sidecar images. |
| FBX | Read | Autodesk FBX, via ufbx. |
| glTF | Read + write | Full PBR materials with textures embedded in the buffer. |
| GLB | Read + write | glTF binary — one self-contained file. The best interchange choice. |
| PLY | Read + write | Stanford PLY, binary or ASCII. Positions, normals, UVs and vertex colours. |
| STL | Read + write | Binary or ASCII. Geometry only — for 3D printing and CAD hand-off. |
| 3MF | Read + write | The modern replacement for STL: indexed watertight geometry, units and per-object colour in one package. |
| OBMESH | Read + write | OmniBlick's cluster-LOD cache — a cooked, mmap-able DAG for multi-gigabyte meshes. |
Point clouds
Rasterized by a compute shader rather than drawn as geometry, with continuous density LOD — so the frame cost is a budget in points-per-pixel, not a function of how many points you loaded.
| Format | Direction | What it carries |
|---|---|---|
| LAS | Read + write | Survey standard, point format 2 — XYZ, intensity, RGB, classification, georeferenced. |
| LAZ | Read + write | The same content, LASzip-compressed — typically 5–10× smaller. Best for archiving. |
| E57 | Read + write | The scanning industry's interchange standard: position, intensity and colour in one scan. |
| PTX | Read + write | Leica Cyclone PTX — multi-scan, with the registration matrix applied. |
| PCD | Read + write | PCL's native format — ASCII, binary or LZF-compressed. Read by PCL, ROS and Open3D. |
| PLY | Read + write | Binary or ASCII, header-driven — any scalar type, streamed as fixed-stride records. |
| XYZ / TXT / PTS / ASC | Read + write | Plain text, one point per line. Universally readable. |
| HDF5 | Read + write | One compound dataset (position, intensity, colour) for scientific tooling. |
| PCB | Read + write | OmniBlick's own format: smallest and fastest to reload, and the only one that keeps the CRS and origin exactly. |
Gaussian splats
A first-class asset domain beside meshes and point clouds, with its own render pass, its own VRAM budget and stereo that sorts once from the cyclops eye — so the two eyes cannot disagree about depth order.
| Format | Direction | What it carries |
|---|---|---|
| PLY (3DGS) | Read + write | The INRIA interchange format every splat tool reads: full float precision, position, rotation, scale, opacity and spherical harmonics. |
| PLY (compressed) | Read + write | PlayCanvas / SuperSplat compressed 3DGS PLY — chunked and quantised. |
| SPZ | Read + write | Niantic's compressed delivery format — about 10× smaller than a 3DGS PLY with virtually no perceptible loss. |
| SPLAT | Read + write | The legacy antimatter15 viewer format. |
Photogrammetry
A capture is a first-class scene object: solved camera poses you can see as frusta in the viewport and click to fly into a photograph's exact viewpoint.
| Format | Direction | What it carries |
|---|---|---|
| COLMAP (binary) | Read + write | cameras.bin / images.bin / points3D.bin — the format every SfM tool and 3DGS trainer reads. |
| COLMAP (text) | Read + write | The same model, human-readable — what you want when diffing a solve against COLMAP's own output. |
| transforms.json | Read + write | nerfstudio / instant-ngp: intrinsics at the top and a camera-to-world matrix per frame. |
| EXIF from photos | Read | A folder of photographs or a video, read straight in with its focal metadata. |
Scene layers & GIS
Esri I3S / SLPK read and written first-party — the node tree, materials, per-feature attribute columns, Draco geometry, KTX2 textures and LEPCC points, with WGS84 ↔ ECEF ↔ ENU georeferencing.
| Format | Direction | What it carries |
|---|---|---|
| SLPK | Read + write | Esri scene layer package — 3D objects, integrated meshes and point-cloud (PCSL) layers, I3S 1.6 and 1.7+. |
| I3S | Read | The layer format inside an SLPK, including per-node feature attributes and statistics. |
Convert without opening a scene
The Converter panel takes a pile of files and turns them into another format in one pass — and the same paths are available from the command line, so a nightly batch job runs the exact code the interface does rather than a second implementation that drifts.
OmniBlick.exe --convert scan.e57 scan.lazMissing one?
Adding a format is one self-contained piece of code that touches nothing else in the application — no edits to the interface, the scene or the renderer. That is why a requested format is usually a matter of weeks, and why we quote them rather than declining them.
Ask about a formatPoint it at the file you were told to convert.
The free Viewer reads every format on this page. If it opens, the switch is already most of the way done.
Free Viewer, permanently · 14-day Studio trial, no card · Windows 10/11