What is a correct way to compress normalmaps?
From reading readme it looks like something like "-normalizexyz -errormetric normalxyz -format RGB8" should work better than "-error metric rgba". But with this errormetric I get some artefacts: https://ibb.co/dyL4dk (left is a source image, center is "-normalizexyz -errormetric normalxyz -effort 50 -format RGB8" and right is "-errormetric rgba -effort 50 -format RGB8").
Source image for tests: https://ibb.co/gkTqJk
What is a correct way to compress normalmaps? From reading readme it looks like something like "-normalizexyz -errormetric normalxyz -format RGB8" should work better than "-error metric rgba". But with this errormetric I get some artefacts: https://ibb.co/dyL4dk (left is a source image, center is "-normalizexyz -errormetric normalxyz -effort 50 -format RGB8" and right is "-errormetric rgba -effort 50 -format RGB8"). Source image for tests: https://ibb.co/gkTqJk