-
FBX Viewer 0.2.1
StableAll checks were successfulrelease / release (push) Successful in 1m0sreleased this
2026-08-08 10:48:03 +00:00 | 1 commits to main since this releaseAssigning a texture to a palette-atlas model produced a black or unchanged model.
Three separate causes, all found withtest/fixtures/hum_stand_a.fbx.Fixed
- Textures were applied upside down. Images were decoded with
createImageBitmap, for whichTexture.flipYis not honoured — three's own loader
documents it as ignored. On an atlas that is not a subtle difference: every UV read
a different cell, and for this kit all of them landed in the unused, black half of
the palette. Decoding now goes through an image element, the same path as every
other texture in the pipeline.test/paletteAtlas.test.tsasserts both directions. - Emissive, roughness and metalness maps did nothing. three multiplies each by its
scalar, and this kit exportsEmissiveFactor 0while the Phong conversion picks
metalness 0— so the assigned image was multiplied away to nothing. The scalar is
now set alongside the map, the previous value is remembered, and clearing restores it. - Palette atlases are point sampled. Meshes whose UVs collapse onto a handful of
texels are reading a grid of flat colours; filtering or mipmapping such an image
blends the neighbouring swatches in. Detected from the geometry, never the image. - Ambient occlusion maps are pointed at the first UV set when a model has no second
one, instead of sampling nothing. - A base colour map neutralises a non-white tint so the image shows its own colours,
and restores the authored tint when cleared. - Loose
.fbxand image files in the repository root are no longer packaged into the
.vsix.
The Materials tab now says what was adjusted and why, rather than changing things
quietly.Downloads
-
Source code (ZIP)
0 downloads
-
Source code (TAR.GZ)
0 downloads
- Textures were applied upside down. Images were decoded with