So this was my first attempt to train with MOSS to see if it was any good. And it's certainly different. So, first and foremost, I don't consider this a perfect lora, but it should be good overall for most things. First, despite being trained on tags with a tag, what it really needs to know what you want is something like "comic artwork" despite that not being in the tagged data. Like I said, different. The actual tag I trained it with is 0k0ksty1e, but it doesn't seem to do anything. It is not intended to reproduce content. Only the style. So if you want the content that they make, that's not the purpose of this lora. I didn't train it on any of their tf/inflation stuff, so that's not in the dataset/outputs. One thing it has as a weakness is that it seems to struggle with backgrounds; for some reason the model really likes photo backgrounds, so you may need to put things like 'photo' or the like in your negative prompt.

Files with SHA256: 5fc8cd88eb0fab677d4e31efcde7b5603ee6915050fd25c283b15f7c8180391a