Question about training details
Hello, sorry to bother you with my inquiries, but if possible, I would like you to clarify some doubts regarding the creation of characters within the model.
How many images do you use approximately for a particular character?
How many images for fine-tuning/regularization for this same character?
I understand that you are using synthetic images for fine-tuning, so could you give me some guidelines on how to generate them?
Finally (sorry for so many questions), I want to know how you do to prune the tags for the characters, meaning that, for example, for Lumine from Genshin Impact, it's enough to invoke her as "Lumine" without having to list all her additional characteristics (blonde hair, short hair, blue eyes, etc.)
Thank you very much, and again, sorry for so many inquiries.
Thanks for the great work, keep it up! =)