Yet another ulzzang or Korean beauty lora that aims to avoid photoshopped/edited looks. It is intented to affect overall aesthetics instead of focusing only on faces. Along with the trigger word, this lora overrides <code>korean</code> token because most SDXL checkpoints are generally not good at Korean faces. That makes it possible to try ethnities other than Korean if you want.

KBeauty - Korean aesthetic beauties

# Training<ul><li>Changed training tool</li><li>Changes to dataset</li></ul> # Output<ul><li>More diverse backgrounds and faces</li></ul># Samples<ul><li>No samples of other races/ethnities</li><li>Changes to prompts</li><li>Upgraded Ultraspice</li></ul>

Files with SHA256: ebd3bdc99943a0d8e89bf9a6b7ea6dc19fd545ee60d1ca7d9987172694adfbd1