When you don't get what you want, there's always pouting to make you feel better.
Good for bringing down the mood where most models make beaming smiles that aren't appropriate for a scene. Also good for mugshots!
Starter prompt:
photo of <subject>, pouting <lora:bitch_breaker_v2:0.8>
Description
FAQ
Comments (16)
Why is it called "B!tch breaker" when I download it? lol.
The first version when uploaded were presented as a sexual LoRA used to make certain expressions during sexual acts. The moderators removed it, which honestly were a shame considering the expressions are applicable to so much more than just NSFW generated artwork. It can be used for a subject in deep emotional pain, perhaps having lost a loved one, bawling their eyes out, as an example. Hopefully this time it gets to remain uploaded.
@natewilliams I agree that v1 had more potential applications than the name "bitch breaker" implied.
Maybe if it was recontextualized as "ugly crying" or an even more general "contorted facial expressions", without the sexual or abusive connotations, it could stay up.
@thegipper Fair enough! Wasn't sure since I didn't have a chance to grab v1
@lamech000 all good, just didn't want anyone to miss it
@thegipper do you have a huggingface?
Man, I had a thing for mugshots years ago, one of my more unusual interests. I liked the ones with the devious smile, like "Yeah I did it, and I liked it, so what!" That might be a good idea.
Also, it would be so hard to prompt the expression of a dirty girl who knows how to get what she wants. It's a look in the eye that says "Hey daddy, what do I have to do for the keys to the Ferrari?"
And there was a crying one that was up for like 3 seconds and I missed it, but I've been hoping to see that again. It wasn't like a sad sniffle, a real ugly cry, probably too extreme for post-neuter Civitai.
But they are some ideas anyway. I would bet there are some less censored model sites out there, the internet used to be a place where a man could damage himself, so boring now :(
Hello friend. I'm glad you remember my other model, it was sadly up for a very brief period and the mods wouldn't respond to my appeal request.
Just for a lark you might try prompting this Lora with its other activation word and some similar words you might have fed to that model and see if you like the outputs. Loras are flexible and pick up all kinds of context, who knows what you might get.
@maxhitman (tagging you for previous comment as well)
@thegipper Do you remember the activation words for the original one? or maybe @MysteryWrecked ?
@Jellai If I recall the image he is referring to came from prompting "broken woman" with absolutely nothing else in the prompt or negatives (literally no other words), which is oddly intense compared to other expression models. It depends a lot on your underlying model since it doesn't get much context from elsewhere other than the expression to map onto whatever the model has already. Still works with LastUnicorn.
I'm honestly not sure why that happens, as the training data is just run of the mill BDSM shoots and, like, stock photo type expressions that are way less intense (and, like, 3d?) than what it outputs. With some models "broken man" as a solo prompt has a really consistent visually intense feel that's closer to that solo-prompt LoRA output than the training data is, so I think there might be some, like, art project in the base model set that's visually close enough that it amplified both.
Nice! I remembered the last model had some messed up make-up for the girl while broken. How do you activate that?
Interestingly none of that's actually prompted in captions (the two activation words listed are the extent of captioning currently). SD already understands the context and the training data seems to just let it express it better, so I'd try natural language guesses on what might bring it out.
There are two things that I've seen result in what you're describing. One is using class words that SD already associates with makeup, so things like "photo of prostitute" instead of "photo of woman."
The other is turning off any gloss keywords. I generally use ICBINP, and most of the time add the trailing keywords (sharp focus, etc) and negatives, but those gloss words (*especially* negatives and universal negative embeddings) really neuter the outcome image. If you want a good visceral emotionally cathartic ugly cry with tears running down her face (say, for representing the emotional gravitas of a funeral), I'd remove all negatives and quality improvements initially.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.