> Presumably some human has mastered the art of “accurately representing black or mixed hair without it turning into a caricature” There absolutely are good ex

> Presumably some human has mastered the art of “accurately representing black or mixed hair without it turning into a caricature” There absolutely are good examples, yes, and I think you're right that the failure of capturing it is largely down to volume in the extant datasets, and particularly in datasets that provide a direct match for the type of line-art colouring in style, which is in itself fairly dated. I agree with all of your recommendations. I think it's definitely a problem that will be solved. The main thing is for people to get used to looking for it.