candy charms v2

SD 1.5 Standard LORA

Training dataset was optimized to face likeness.

Model trained on (mainly) instagram images of Candy Charms. You need to boost the weight above 1 for the face likeness and/or use clip skip set at 1.