For regular nose hook: woman with [color] nose hook with [color] [material] straps/strings over head. E.g.,: woman with Silver nose hook with black leather straps over head.
For three-way nose hook: woman with [color] three way nose hook with [material] straps/strings over head and around both sides of nose, 3way style nose hook.
Works with base and edit. Play around with weights to get what you're looking for depending on T2I or I2I for base or edit; usually use 0.4 to 1.4 as needed.
Description
FAQ
Comments (10)
oink oink 🐽
Yay!!
I cannot seem to get my QWEN Lora's to work on Civitai's generation feature...but it works in comfyui. Any suggestions/help? Lora was trained on TA if that helps.
Your v1.0 works really well, but it's a bit complicated to connect it to your Hood Lora :) but it works
Thanks for the feedback. Still learning how to make LoRas, haven’t crossed the threshold of knowing how to keep things strong enough to have it learn “independently”
Did you do anything in particular to make it work? I’m curious :)
I can imagine it's not easy, which is why I haven't started yet :) But I'll post pictures of your Loras from time to time :) Would it be possible to create ZIT versions of your Latex Hood/Armbinder/Nose Hook/Elbow Bondage/Breast Bondage/Harness Ballgag and Ballgag? I actually really like using ZIT, I'm only on Qwen now because I found your Loras and there isn't really anything like that on ZIT yet :) ZIT is much easier for me to use, but I also know that creating ZIT Loras is supposed to be even more complicated :)
@Gorean Unfortunately I have very minimal experience with ZiT/ZiB and I am focusing on models that have edit capabilities too (like F2Klein and Qwen) - waiting for Z Image to drop its edit models for now...thanks for posting pictures, always great to see how others use it and their creativity. Loving your work!
@Gorean google "phr00t" for qwen Rapid AIO version - I use V18 / V19 NSFW - its amazing and lightweight with baked in NSFW and lightning LORA's, it does better than base fp8
@vackyadmiral167 I tested versions V23 and V19. It has incredibly good prompt adherence. But it's not exactly lightweight, using 50GB of RAM :) However, creating ultra-realistic images with it is very complicated. I also haven't yet found the right LoRa for that level of realism. But thanks for the tip.
@Gorean Sorry lightweight wasn't the right word. Simple workflow is what it was. It works fairly great on my 32gb ram/16gb Vram setup. 30 second generation time on 5070ti.