Anisotropy fix #154
Anisotropy fix #154
Conversation
Removed KHR_texture_transform rotation for the normal map, rotated the pixels of the normal map, and reset the anisotropyRotation to zero. Adjusted anisotropyStrength so the normal map will be visible when the camera is perpendicular to the surface.
|
@echadwick-artist Can you pull the latest copy of main and merge it down into this branch? Thanks! |
|
The rotated normal map here doesn't look right to me. The grooves are most pronounced in the green (vertical) channel, not in the red (horizontal) channel. The map was presumably correct when these were horizontal grooves with long faces pointing +Y and -Y, but now that the image rotated, the long faces need to point +X and -X instead. I think the correct answer is to swap the red and green channels of the image, and then invert one of them (maybe the green one after the swap). Also the normal map apparently is itself not normalized (just guessing from the blue channel), and may need the same type of fix as #114. |
|
Oh gosh, you got me haha. I rotated the bitmap, not the normals! How could I forget this is a bunch of vectors, not a just pixels? |
1. Recreated and normalized the normal maps. 2. Increased resolution for the spiral normal map because it was a bit pixely. 3. Set anisotropyStrength 0.5 so the normalTexture is visible when camera is perpendicular to surface. 4. Set all normalTexture Scales to 0.5 so they all match each other.
Removed KHR_texture_transform rotation for the normal map, rotated the pixels of the normal map, and reset the anisotropyRotation to zero. Adjusted anisotropyStrength so the normal map will be visible when the camera is perpendicular to the surface.
Per the reported issue here: #153
Screenshots from various glTF renderers:
Note that Dassault Enterprise PBR does not seem to support anisotropyTexture, so the spiral grooves are not rendering the same as the webGL rasterizers.