This is still in testing. I’m thinking I trained it with too many different poses. So far it doesn’t seem to work well. Currently training a Lora for each individual pose.
Description
FAQ
Comments (2)
Yup, I see you've figured it out. I train a lot of anatomy/poses, and the best quality always comes from keeping the training images as similar as possible. I can get a lora with up to 4 poses, if I have an equal number of images of each pose and the prompt's are different enough, but i've found, it's both easier to train, and to prompt, a lora that only really does one thing, and does it well.
I figured if I had 10ish photos of each pose and tag it correctly, that it would work. Guess not lol. But I just uploaded a new training with one of the poses and it worked out fairly well!