Per-pixel quantization now supports rendering of 8-bit alpha channels. This has been implemented because I’ve figured out a method with zero impact on performance, piggybacking stippled patterns into the per-pixel quantization data. The tradeoff is that it only supports three levels of transparency, but at least it helps a bit to smooth out the edges of complex alphamasked textures, making those edges look less jaggy and more grainy.
Screenshots:
And support for external luma/glow textures with 8-bit alpha channels is also in. This also allowed me to generate glow maps from the fullbright texels of the internal 8-bit color textures, use them to generate 32-bit color mipmaps from the internal textures, and compile those for per-pixel quantization.
Screenshots:
