v1.0: Trained on 59 images of mostly American actress Isabela Merced's face. Gives best results with semi-realistic models.
v1.1: Retrained at a resolution of 1024x1024. Better results for 30-40 steps for higher resolution direct generations. Works best for 896x896 image generations. Keep weight at 0.5-0.7 for direct generations.
Description
Retrained with 1024x1024 resolution. can be used in combination with older version images at lower resolutions. Keep weight at 0.5-0.7 for direct generations.