This Lora came out unexpected, I tried to fix this Lora, "cure" it from bad quality artifacts, but understood, that soles are too complicated for SD, and create morbid mutants if left in the dataset. So I thought to cleanse my set from images that containt soles.
Then I thought, that this pose is beautifull, and it woul be a shame to lose it completely. And then I gathered some additional images with soles and feet, and here we are.
Well feet are really complicated (try to draw one yourself, to understand that). So expect only 3 out of 10 decent images, or even less - others will be weird.
Tags (shouldn't be necessary)
Soles, from behind
Parameters suggestions
Should work in txtToImg with sides somewhere between 512-768
Weights between 0.5-1
Clip skip - I didn't use it
Pay attention to the sampler model - my standard Karras 2M wasn't good all the time. For realistic I found out that dpm adaptive is better (at leasts for me and for that exact Lora). So use x/y/z to find the best combination of parametrs.
Enjoy)
Feel free to ask any questions here in comments, or in my discord channel.
Also I am making games with AI arts. They will be free in the future, but if you want to participate in making or have an early access - you can support me on patreon)
Description
FAQ
Comments (10)
accidental doggy style 💀
oddly specific xD
not kink shaming here <_<;
@4claude207 i think you confuse this site with a pornsite, i will kink shame anyone i want xD
@aleksml_84 I understand text can be taken out of context, no worries! I was being light-hearted and pornhub is blocked in my state and I refuse to use a VPN xD
Omg, this works surprisingly well for me and actually produces GOOD FEET! (well most of the time I get decent feet) Super hot. Didn't know I needed this pose in my life.
I user orangemixs model with this lora,Stunning LORA, the generated feet are very complete, even in the case of wrong posture, it can generate correct feet, hope to produce a separate feet LORA.
@cloveriowo Yeah i am gathering dataset for it) If I succesfully made this Lora, I then add it to my model, because feet are mostly disgusting in sd)
This LORA has surprising performance in expressing dynamism, such as motion blur and motion lines. By adding keywords such as {trembling, motion line, motion blur} etc., you can create the dynamic effect of R18 pictures. This can be achieved accurately even in situations that are not pov perspective or in other positions. I recommend you try
Could you give some examples of positions other than POV? 😊