Please note that the lora is bad with hassaku half of the time for some reason, idk why!

Again, on artist request, the model will be deleted.

25/09/25 V1

I trained it on 36 images with x8 pixel size, most of the images are 1024x1024. The base model is PlantMilk, but it's noted that Hades Pixel Art and WAI NSFW works the best.

It is kinda fried and overfitted IMO, but not like there's other way with dataset so small. Fell free to experiment however!