Update 2023-06-24:
For example, I feed this pic to roop:
and then I can generate the following pics with SD, without needing LoRA:
Original Description:
This LoRA is based on the American pro wrestler Colby Daniel Lopez, more famously known as his ring name Seth Rollins.
I've been busy lately, so this LoRA is just a quick test. All I did was dump 48 photos and use the dreambooth method to train (instead of finetune method I used before), no captions or reg images at all. The output is surprisingly good compared to the low efforts I put in.
Trained on SD 1.5, tested on Realistic Vision V1.3, SD 1.5, and RPG V4. It performs less well on NAI-based models.