V3:
Remade the entire dataset and made sure to make all background white for the training, removed trigger word since it didn't have any use. This version is superior to V1 and V2 and should be preferred over the other two. This is a LoHa so you will need to make sure you can use them with your SD.
Comparisons are included.
Usage:
CFG Scale: 5-9
Weight: Anything you want, I personally prefer 1
Detail your characters well, simple details will not fail you but you may find them lacking
You may need to use the integrated LoRA of WebUI with this extension https://github.com/KohakuBlueleaf/a1111-sd-webui-locon
I'd like to thank everyone that used my LoRA, specially those that reviewed them. It has allowed me to learn a lot about the process of training LoRA. V3 is the final version and will not be improved.
The dataset used can be found here: https://mega.nz/folder/pqRkiQRR#QUT9-e8QsvRcE9fTD56snQ
