It would be super nice to be able to generate mipmaps for normal maps consisting of only 2 channels, and still have correct normalisation. This would likely also require the ability to specify which channels the 2 normal map channels are stored in, and mipmap the others as normal, which would allow normals to also be packed alongside existing data
I would envision adding channel parameters to -normalmap, so something like -mipmap -normalmap 1,3 (mipmap, treating channels 1 and 2 as the [2 channel] normalmap)
It would assume the source normals were initially normalised, and maybe calculate what the extra channel would be, then normalise the 2 remainder and the third in memory when generating each reduced mipmap, so that lengths were correctly maintained (or maybe some other method if there's a more sound mathematical convention).
Being able to specify channels for parameters like -normalmap would also provide knowledge about channel packing and may allow other abilities later on, when packing data together.
It would be super nice to be able to generate mipmaps for normal maps consisting of only 2 channels, and still have correct normalisation. This would likely also require the ability to specify which channels the 2 normal map channels are stored in, and mipmap the others as normal, which would allow normals to also be packed alongside existing data
I would envision adding channel parameters to
-normalmap
, so something like-mipmap -normalmap 1,3
(mipmap, treating channels 1 and 2 as the [2 channel] normalmap)It would assume the source normals were initially normalised, and maybe calculate what the extra channel would be, then normalise the 2 remainder and the third in memory when generating each reduced mipmap, so that lengths were correctly maintained (or maybe some other method if there's a more sound mathematical convention).
Being able to specify channels for parameters like
-normalmap
would also provide knowledge about channel packing and may allow other abilities later on, when packing data together.