Oh! So when the sampler node is set to “normal” mode, that implies a “derive normal z”? Makes sense.
…it’s still kind of garbage though.
This is with normal and ao separate: https://i.imgur.com/ZvUii1I.png
And this is with normal and ao combined: https://i.imgur.com/YO7xHPG.png
The material: https://i.imgur.com/VzQOeGQ.png
Having the ao in the B channel seems to introduce artifacts in the other channels… really strange.
