「へ～こういうモデルが気に入ったよね。お可愛いこと・・・」
Did it on a whim, only used 8 images to see if I could train an SDXL LoRA locally in a rapid manner. Was expecting it to take 14 hours, but it instead took 3, so the settings I'm using seems to be ideal for me. This model is trained specifically to try and do that pose, so at least on VARXL/SilenteMoney, it won't deviate from a close-up shot of the character's face, no matter how many varied prompts I put in.