A model attempting to replicate the ear_sex/mind_fuck booru tag (without the gore). For best results the weight will very from prompt to prompt and other TI's/LoRAs. Only tested with anime SD models, no plans for realistic. Sometimes the character's get the fucked_silly face, other times if turns into a blowjob. Set the weight low enough and you get a good ahegao face.
Note: consider adding "animal ears" to the negative prompt as that often shows up. That is from the images used to train.
Description
FAQ
Comments (19)
Huh but also thanks
Random image, to random tag, tried to replicate. 1/10 will not try again.
GOD! it S the frist time I know this way~
Is it always censored?
Depends on the weight, but most of the time yes.
Amazing degen fetish. Would be better as a lora tho
Edit: uploaded lora
Edit2: lora got deleted, lame
I made a lora for this a while back but it needed some more dataset cleaning and stuff and I lost interest and didn't think many people would care about such an obscure concept anyway. Maybe I'll just upload what I have anyway.
sadge, why was the LoRA deleted?
@fakealtaccount no reason given, looking at TOS, maybe something like implied "depictions of human death" heh
@wewewew wonder why this one was left then
I find it funny that under the account settings they have "Visually Disturbing" -> "Corpses" or "Emaciated Bodies" but a little bit of ear smashin is bad.
Idk where to even find the TOS, nor did I see the original post, but i suspect human death wasnt the reason (Unless maybe "Corpses" implies "You're allowed to show dead bodies, just not people actively dying")
@NotTheNax I don't know. Maybe it has something to do with my version being censored?
@Raz No, just here. Lots of things not released, no plans though.
@Raz Would you be able to reupload the lora somewhere else?
@eventbills Here, https://files.catbox.moe/56lpe1.safetensors
Main trigger is earfuck, but you can try others too.
@wewewew It seems like the LORA by itself doesnt work (even when tagging earfuck), and the embedding itself seems finnicky, However, when combining the two you get a result greater that its parts.
I would 100% recommend using the embed and the LORA together
@EricHuanggle1001 It works, but you need more tags. You can check the metadata of the LoRA ("i" button in the thumbnail view on the LoRA tab of A1111)
In everyone's interest, here's an XYZ plot (more like an Y plot) of that LoRA in various checkpoints.
https://files.catbox.moe/4bsviz.jpg
@wewewew Thanks! Checking the tags, I noticed some cute and funny tags that might have triggered the deletion. Maybe omit those tags from the metadata and upload again?
Do one for animal ears
Thank you very much ; P