I find the metallic makeup (eyeshadow and lipstick) on most models lacking. This lora is created to fit my personal tastes and the data set is curated from images generated with Grok Imagine.
The lora has a bias toward the villainess archetype (smiles will tend to look more evil) since that is what the data set is based on. It also favors a more Korean/Asian aesthetic. It is best at gold eyeshadow & lipstick, but can do other variations of metallic/glittery makeup.
Recommended on 0.8~1 strength, DPM++2M or Euler at around 20~25 steps. No activation keywords.