Aras Pranckevičius has released v1.1.2 of his Unity gaussian splatting plugin, which makes orthographic cameras render splats correctly.
The release is the first since v1.1.1 in April 2025 and rolls up 20 commits, including six community contributions. The README still carries Aras's December 2023 note that he is not planning significant further development, and the plugin has moved forward on outside contributions since, as when a community pull request added VR support in late 2024.
Orthographic cameras had rendered splats incorrectly since at least October 2023, when the problem was first reported. The fix, from wells2kb, has CalcCovariance2D detect an orthographic projection matrix and build its Jacobian from a constant focal length, without the divide by view depth that perspective projection needs.
The importer also now accepts lower order spherical harmonics. PLY files below SH degree 3 load through a contribution from WindyYam that Aras landed himself, and a change from winnie1994 does the same for SPZ, unpacking only the coefficients a file contains. The SPZ reader still accepts only version 2 files, so the ZSTD based containers of SPZ 4.0, which Niantic Spatial released in May, do not load.
Two PLY import failures are fixed. Files under 2GB whose expanded splat data passes 2GB crashed after Morton reordering, because the NativeArray copy constructor overflows a signed 32 bit size, and the importer now allocates and copies that buffer manually. The PLY header parser now ignores properties outside the vertex element, which had broken files produced by Apple SHARP. A memory leak during PLY reads is also fixed.
Compositing changes from AdrianTheFrog and mwfarb improve how splats blend with skyboxes and other scene objects, with the composite shader now un-premultiplying color by alpha and dropping a 1.5x alpha boost. The splat editing selection rectangle, which was upside down, is corrected.
The package, org.nesnausk.gaussian-splatting, targets Unity 2022.3. Unity Gaussian Splatting v1.1.2 is available now on GitHub under the MIT license.
