Recently I noticed that most anime models are bad at generating foot licking with a floating disembodied tongue (probably they better produce those with a whole head but I never tried that kind yet so IDK), and I have an extreme fetish for foot licking in general, also add the fact that in general there aren't that many drawings about disembodied feet licking around the internet AFAIK. So I trained this concept LoRA with IllustriousXL about that kind of foot licking in mind. (Similar to a clothing 1.5 LoRA I trained in the past about single bare shoulder shirts)Please keep in mind that perhaps due to model's fault and from dataset too, it sometimes does not generate a floating tongue at all or if it does, it completely misses the foot or tongue might look bifurcated or not in a proper shape, so I would suggest putting these tags in negative prompt besides the trigger ones in positive prompt alongside having some luck as much as with a gacha game when getting acceptable enough or perfect feet too, that's just how it is sadly:<pre><code>multiple tongues, tongue out, french kiss, licking lips, licking another's face, licking another's neck, licking another's cheek, licking another's lips, holding tongue, pulling tongue</code></pre>For the most part, it does produce some pretty good results at least and does work well with other artstyles and XL LoRAs together, IDK if Pony Diffusion XL can even generate disembodied tongues natively by its own without needing a LoRA such as this...

Disembodied Foot Licking | Concept LoRA | IllustriousXL/NoobAI

I deleted images having multiple tongues and crappy tongues in place for better ones with single tongue in dataset and trained it with my PsygaMix model, despite all of that, it still has issues such as additional tongue on face and tongue missing foot entirely, probably due to NoobAI EPS 1.1 model which I merged with others for my PsygaMix and not this LoRA's fault by itself since I tested out this concept natively with that model which is capable of but not as much without this LoRA.You can try using these tags in negative prompt which can minimize those issue but not neutralize 'em entirely yet so far:<pre><code>multiple tongues, tongue out, french kiss, licking lips, licking another's face, licking another's neck, licking another's cheek, licking another's lips, holding tongue, pulling tongue</code></pre>Overall I think it's an improvement over the previous version but maybe not by much considering the issues with NoobAI, hope that one day these issues can be addressed in whatever way someone comes up with...

Files with SHA256: c679147abdc128bdc9bd202bdde4b61d027f6b886d6c950c11e4c8fcf923c656