For this, found 5 pictures online, 2 were good, rest really low so spent time detailing them out converting to more animated as they were realistic. Made the first model, came out okayish, then just rinse and repeat until I got what I wanted haha. Working on a next version, I think I can make this better.

@_1_  Oh 100% most of my models, especially concepts, the first version is always usually quite unstable, because they're usually really low niche datasets, so I experiment and try to get some reasonable gens. So like V1 of this was 6 pictures, 1.2 was 12 (6 original and 6 gens), then 1.3 was 35 pictures and I used all gens for that one, I tried to change the styles of each and keep them varied but as long as the chair in this case was the same and its came out better each time. Because this concept is quite intricate and 2 characters in it is hard, I just need a bigger dataset for the next one to try and capture the different poses the woman can make.

@_1_ I also use pixel cut to alter some images, so if I get some near perfect, I edit them. In this case, the chair kept seeming like a collar, so I blended the collar into the chair to make it part of it again used that.