Trained on realistic images focus on POV.
best output is around 0,8-1 for me
deepthroat_v1.safetensors