Skip to content

Reconstruct normal Z *properly*

Alexei Kotov requested to merge Capostrophic/openmw:actuallynormalnormalz into master

I keep running into embarrassing math failures these days.
screenshot776
screenshot775

The former might look "more detailed" in this specific case, but note the odd-looking edge areas.
The sanity check is to realize the normal map's 0.5, 0.5 values in X and Y channels should result in a Z of 1.0, not 0.707 like what my actually second bad math attempt results in.

Merge request reports