https://civitai.com/models/2138798/background-inset-zoom-layer-cut-in
V2.0
I re-trained the concept into two separate versions:
V2.eyes and the more general V2.face. See the respective versions for base prompts.
Start with 0.8 strength.
Prompting for conflicting states (eg. "breasts out, covered nipples") works ok with the "face" version
