Put some lipstick where nature didn't intend it to go (i.e.: a penis). Allows you to add lipstick rings to penises, consistent with the lipstick color of the girl in the picture.
Results are not always consistent and may sometimes require a little luck, especially with the lite version.
If you like the LoRA, please leave a reaction, review and/or tip. Enjoy!
Usage
Trigger word:
<color> lipstickSome colors work better than others, but all colors are possible
Recommended strength 0.8 - 1.2 (lite version may need a lot of extra strength)
Try also increasing the weight of the trigger word
Use Clip skip=1 or Clip skip=2 (other values may also work, use at own risk)
Use at 1024x1024
Optional: Use in-painting to nudge the LoRA in the right direction
Optional: Add tags that work well with this LoRA:
povafter fellatio, saliva on penislipstick marks("kiss marks", different from the lipstick ring)fellatio(for actual blowjob)oral, licking(for licking penis/testicles)
Example
See preview images (generated with pruned version).
Unpruned vs. Pruned vs. Lite
I offer three variations of the model, which only differ in the file size and quality. Pick the one that best suits your preferences. If you can not decide, simply pick the "pruned" version. Note that the quality is relative, even the lite version usually produces images which closely resemble the intended concept.
unpruned: The originally trained model. Largest file size, best quality. Pick if you want the best possible quality, at the cost of disk space.
pruned: Downsized model. Almost same quality as the unpruned version, with noticably smaller file size.
lite: Drastically downsized model. Smallest file size. Still has good quality, but noticeably worse than unpruned and pruned.
Notes
Suprised there was no Pony version of this yet. I expected the training to be much harder, but it turned out quite decent. The LoRA is not very consistent yet, but that is likely due to the training data being a little lacking. I might revisit this in the future and add some synthetic data.
Using training data
All training data has been sourced from publicly available sources. Feel free to contact me if you want access to the training data.
Description
This is the lite version. It has a minor, but noticeable loss in quality compared to the unpruned version.
Use this version if you want to save some disk space.
WARNING: This can get the concept right, but it will be very random whether the lipstick rings actually show up or not. You will either need very high LoRA strength, some luck, in-painting or a little bit of all three to make it work consistently.
FAQ
Comments (8)
How do you prune a lora? I almost never see xl loras this small.
Strictly speaking, I just resize them to a lower rank using KohyaSS and do not prune them in a traditional sense, but the concept is pretty much the same and I think unpruned/pruned/lite is easier to understand than rank 8/4/1.
Additionally, I also train my LoRAs on rank 8. I did a lot of testing and using a rank higher than 8 never significantly affected the quality of the LoRA. It's a bummer so many people make LoRAs as large as 600mb when it's really unnecessary. But then again, if you don't have a good GPU or are training you LoRAs on civit, then resizing may take quite a while.
I can only recommend people try training their LoRAs on a lower size. While disk space is rather cheap these days, it's not unlimited.
The lite version is marked as v1.5, can you confirm if that's true or a typo? I'd assume it's SDXL same as the other two models.
You're right, that's a mistake on my side, it should be marked as Pony. I was pretty tired when uploading this. Thanks for noticing!
How do you evade kissing marks being slapped on girl's breasts and butt? I always getting it on weird places which make no sense and so far fail to filter it out.
Also thank you for the great lora! It's truly amazing.
I wonder whether one could make a rainbow blowjob with this? Seven different girls, seven different lipsticks, taking turns, leaving the cock with a rainbow on his dick.