Runa Kagurazaka

Details

Model description

I must be upfront that I faced significant challenges while training this model (it took me two full training cycles).Previously, I encountered a technical issue where sample images showed the error "Some resources used to generate this image are unavailable." I'm unsure of the cause, but I sincerely hope this final version resolves that problem. The main difficulty stemmed from the high diversity of outfits in the dataset and the fact that some original training images lacked complete detail for all costumes. Consequently, the LoRA might not have fully learned every minor detail of every single outfit.

Therefore, achieving the best results requires leveraging the Trigger Word along with detailed Prompting to explicitly define costume elements like color, pattern, and accessories.

I have selected the Epoch with the highest overall flexibility. I sincerely hope this LoRA works well for everyone. If you discover any useful techniques or encounter any issues during generation (such as "bad hands" or inconsistent outfit details), please feel free to comment or send me a message. I welcome all feedback and will strive to improve the model in the future😊

Images made by this model

No Images Found.