Hi, I have a question about your LoRA training process.

I noticed that the artist's artwork many close-up shots of feet and detailed foot anatomy. I’m curious how you managed to train the LoRA while keeping the generated feet and toes from becoming distorted.

In my own training attempts, the LoRA often produces issues such as deformed toes, fused toes, or broken foot structures, especially in images where the feet are emphasized.

Could you share any advice on dataset preparation, captioning/tagging methods, or training settings？

Thank you very much for sharing your experience. I really appreciate it.