Training data
#7
by
Eddy5646
- opened
These are all quite great. Would you be willing to give an example of what your training data looks like? I'd like to try my hand at training my own, but I'm not sure how best to set it up.
Quality > quantity. It all depends on your goal, if you want to train a artstyle keep it constistent with varied subjects, backgrounds, etc... For subjects, unless you doing multi subject LoRA, don't have anyone else other than your subject in image. Pruning images that include bad tags is also important, things like manga panels and comics. I personally also delete multiple perspectives and chibi.
YoungMasterFromSect
changed discussion status to
closed