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

Update trained with a NoobAI merged model and changing foot licking tag to licking foot, when I tested it I barely noticed any improvement over the previous version and realized that tongue missing foot or licking another body part or having tongue in girl's mouth while mouth is open is all due to NoobAI's fault rather than my LoRA, let me know some workarounds in order to prevent any of those things as much as possible and have more consistent results (besides consistently having more anatomically accurate feet with all 5 toes...)Multiple disembodied tongues trigger tag is kind of useless when I tried it as it doesn't seem to work that much at all and I rarely get on its own without using it (probably due to amount of images I used in dataset for this LoRA due to that kind of foot licking variant to be really somewhat obscure in boorus and other image boards.

Files with SHA256: 9eac324804349c31a526b1bb584d3ff654e19a69935d66418d5d153258e0bfe9