Please check out my Ko-Fi page if you are interested in supporting me.
LoRA commissions open! -> Fiverr
This is a LoHa LoRA! You need the extension! Why? They do a lot better with multi-concepts.
Trigger words for V1.20:
<lora:CuplessCrotchless120:0.8>(bra, cupless bra, crotchless panties)
Specifying: "pussy", "nipples", "panties", "bra" helps it with generating the actual keywords for the LoRA. If you get full panties, specify pussy for example. If you get just a "shelf bra" then specifying "bra" helps it draw the upper part...
Don't want both at the same time? Go with just "panties" or just "bra" without "nipples" or "pussy".
You can try "micro bra", "pearl thong", "nursing bra" and "string panties", "string bra". I tagged for those in the dataset BUT HAD NO LUCK with generating them (they COULD work, but not guaranteed)! Maybe for a next update.
Follow the steps in the Model Version notes to reproduce my images.
Sample images generated with Hassaku.
All negative embeds should be easy to find. and Prompts should be included in the sample images! Click the little icon in the bottom right to see it.
It learned pretty well how to do covered nipples! Probably capable of using it for any type of shirt if you just want nipples texture.
Could use same tags in negatives to control how it generates.
It was trained on 185 512x512 images scraped from the internet and 20 epochs using animefull_final_pruned.
Danbooru tags generated by Waifu Diffusion 1.4 Tagger and fine tuned for specific details like nipples texture, see-through, and impossible shirt or form fitting.
Description
Generate using a weight of 0.6-1 like this: <lora:CuplessCrotchlessv125:0.8>.
This version is here mostly for me, as a way to save it. Feel free to try it but I recommend using V1.20, there's literally no improvement from going over 20 epochs.
How I archived my generations:
txt2img: DPM++ SDE Karras with 20 steps, 512x512(can go up to 768 for wide or tall), Restore faces on (using CodeFormer weight 1.0 in Settings), Hires-fix on, to 1024x1024 (double original size) with your preferred upscaler and Denoise 0.4, CFG 7. Then send to img2img!
img2img: DPM++ SED Karras with 20 steps, Restore faces on (CodeFormer weight 1), rescale to 1536x1536, CFG 7, Denoise between 0.3-0.7 (based on how much you want to improve the image), SAME PROMPT! Send to extras when happy with img2img/inpaint results.
extras: Scale to x3 (final will be 4608x4608, I suggest going lower, I did x3 just for the samples here) Upscaler 1 - FatalAnime 4x, Upscaler 2 - SwinIR 4x with 0.10 visibility, GFPGAN 0.10 visibility, CodeFormer 0.10 visibility.
If anyone is curious about my version names: v{version}.{trainingEpochs}
FAQ
Comments (12)
still not very sure on how to get it up and running, though. the extension is installed but this one did not pop up in the lora list
Install the extension and just install as a normal LoRA inside Models -> Loras
@Part_LoRAs Thx!
you can just download the extension and put it on the extension folder after that you will see there is lycoris there after
which vae used?
Orangemix VAE, which is basically the NAI VAE.
can this works with pantyhose?
Probably but I'm not sure. Each model deals differently with LoRAs, you have to test it yourself
what's the difference with v1.25?
For some reason the colors go weird when I use hires, the person going pink and the background going green.
Most likely cus its over fit.
Its not a good idea to use it.
This LoHa is a bit old, my knowledge about creating and training models evolved a lot since.
Only this type of underwear can make me excited！