Testing something different, sculptures and neon aesthetic.
(Pinned images used the dataset training tags)
Faces would benefit from ADetailer (which I did not use on the sample images except for the halo angel)
