v3 is ready to go. It seems to work alot more reliably! It also may have stronger gender preferences, I may update title to reflect this.
Experimenting with lora that make a picture "feel" more like a dominant x submissive scenario by altering the facial expressions of those involved. This one's trained to render one of the subjects in a submissive-feeling eyeless manner (making them look more tense/anxious/overwhelmed), and to give other subjects in the picture a more confident/dominant/predatory look.
The impact of this lora can vary alot depending on the model. I suggest initially leaving the weight at 0.7 and trying to calibrate the impact via prompts instead:
If both subjects still have eyes:
In addition to the main trigger phrase, prompt for "eyeless" (potentially placing both towards the start of the prompt sequence).
Try it with zero prompts describing the sub's mood or facial expression.
If both subjects have no eyes:
Try prompting "blue eyes" or similar, so the AI knows to give at least one subject open eyes with visible pupils.
For the counterpart with sinister eyeless dom and expressive sub: https://civarchive.com/models/532360
Description
FAQ
Comments (3)
Any chance for sdxl illus version?
Could you make one like this, but the dominant characters are male?
My models are trained to have bias for femdom and malesub roles, but bias can be overcome with strong enough prompting- for example, "yaoi, 2boys, faceless male," with a sufficiently high scale and weight.
Which is not to say I won't! Not in the cards anytime soon, though.