>>2753931No problem man, for me the time-consuming part is how I do the images, - I'll crop them one-by-one to the relevant parts to maintain 1:1 aspect ratio. If it's necessary, for some images I'll do two crops, one that covers more of the body then one that gets the face.
BUT - after rereading one of the guides I used, I don't think you actually need to do this if you're using the kohya-ss powershell script (which I am), only if you're using dreambooth. So shit, I'll try skipping that step next time.
Either way I run them through the tagger after cropping, then through the 'preprocess images' process in the Train tab of the webui (which if it doesn't come standard it's part of the kohya-ss additional networks extension), that resizes them all to 512x512.
Personally I try to have at least 50 images to train on, the two I'm using were both around that many.
Sounds like maybe I've been wasting time with the cropping lol, but I'm pretty happy with how these two came out at least
Here's the guide I mentioned that helped to get it all working, it may have better answers for you:
https://rentry.org/2chAI_LoRA_Dreambooth_guide_english 