Well, I wanted to try a new technique to create the LoRA and used this dataset I had. Since I went through the time of creating it may as well share it.
Description
FAQ
Comments (2)
Outputs are pretty nice. That new technique with old dataset you mention means that you used small resolution dataset as 512x512? Thanks in advance
SDXL really benefit from proper 1024x1024 pixels in each images. My original dataset was tailored for 512x512... So I had to upscale to 1024x1024.