What I aimed for and achieved:
Improved facial details, along with a matte finish and gentle gradients.
What might have been lost:
The cel-shaded coloring style.
(Please use this one alongside v1 for different results.)
If the distance goes beyond a portrait and you’re not satisfied with the eyes, try adding another LoRA specifically aimed at improving eye detail (I’ve suggested one).
Since some of the training images were heavily influenced by BLK FUTA, it may sometimes generate results similar to those made with BLK FUTA. I’m experimenting with different prompts, but the model tends to behave a bit unpredictably at times.
