INSTRUCTIONS!
This LoRA is quite unstable and currently only works with a select few checkpoints.
It works with some but for others not at all, so if you are getting bad results over and over again it might help to switch checkpoints.
I've yet to successfully use hires-fix to upscale any image. Currently it adds cum on both sides once it upscales it.
Ones I tested it on with varying results
If you find a checkpoint that works. Or any finding, I'd love to hear it, maybe I can improve the model.
Description
FAQ
Comments (10)
afaik during hires.fix pass only the base model works on your image, so without the lora guidance using "cum on face" in your prompt will add cum on all faces
You can get better results in hires fix with a more stable upscaler, which is able to upscale the image at a very low denoising strenght. I added a couple of examples to the gallery with hires fix using the upscaler 4x_foolhardi_Remacri at 0.15 denoising. It's essentially the same before upscaling.
@psoft the artifacts in my first upload stem from a higher base res, used just 1.334 upscale in hi.res fix. i uploaded the imges solely for their included (though ofc unintentional) back view ;)
proper tries with 512x768 and 2x swineir or remacri @ 0.2 denoising were fine, even without using one of the suggested checkpoints, even better when combined with regional prompting like in my second upload
Wow you have my respect thank you sir 👏
do you have a discord you use or somewhere else? i was quite upset they deleted your other lora, yours was the only one that actually worked.
Keep on good work. It's not easy at all to train a Lora with complicated NSFW concept and have it working on photorealistic checkpoints.
I think there's some heavy wowifier in your lora that destroys the images. I managed to fix it by using wowifier myself, but setting it to -0.5.
These are AWESOME!! You ever do any customs for $?
Can anyone help me to create some prompt, I tried to make my image by using existing prompt, but it only show me one image. How can I create two of my character at the same time before and after version.
The ones by eurtaku look to use Regional Prompter (well the one I looked at did). I managed to create one similar to wolfofragnarok by using their prompt but a different model. Have you tried downloading one of the images, opening it in PNGInfo (assuming you are using Auto1111) and selecting Send to txt2img? It will then copy in all the prompt and other info you.
Actually with wolfofragnarok I played around with, as soon as I changed the seed it only gave a single image. And the same with eurtaku one. Can only assume it's luck of the draw as to if it produces the 2 images or not.