I put my 100+ images of my drawings/sketches, etc. and made this LoRA!
Description
test ver! experiment and find the best outcome!
FAQ
Comments (7)
So that's your own drawing style? Looks cool! And respect for training a LoRA of it
It's more like a collection of drawings putted into mixer and spitted this LoRA out.
I saw your profile and I must say you are pretty skilled in making loras. What articles or guides you used to get that good results?
@chemice Thanks. I mostly just followed @THEANTLERS' guide (https://rentry.org/nv2oi8qg). As for the datasets, I find tagging quality to have a big effect, so I try to mostly scrape from dan-/gelbooru and manually cleaning stuff up later (with some automation involved to convert the tags and such) instead of relying on an autotagger.
@chemice So this new version 2? This one is trained on your art?
@Diamond_Letter both were, but v2 have more pictures and was trained longer
@chemice Thanks, though I meant if it was just yours or if it was still mixed with something else.
Sorry if that was a bit unclear.
