Skipped a few versions, took a lot of work to make the lora more stable...
This LoRA is designed to add canine-style genitalia to furry or animal characters. It supports both normal and spread anatomy with strong detail and consistent structure.
For best results, use the following trigger tags:
Normal anatomy:canine pussy
Spread anatomy:canine pussy, spread pussy, clitoris
(Note: When using spread tags, lower LoRA weight by ~ 0.1 can improve results.)
Negative Prompt:
pussy is recommended for all gens.
spread pussy is recommended if you're getting concept blending while trying to generate a non spread image.
Recommended base model weights:
NovaFurry / NovaAnimal: 0.5 to 0.7
FurryToonMix: 0.5 to 0.7 (can be increased)
WAI / IWonderMix: 0.8 to 1 (may conflict slightly with the LoRA)
Description
FAQ
Comments (11)
The best snacks on civitai
I am absolutely curious as to why it's so large
It's 4 96 dim lora merged. I really wanted to use the tag canine pussy instead of some random tag; So I had to train multiple different concepts as their own lora to reduce blending of concepts and merge them.
I 100% could of trained at 64 dim but went with 96 to be safe.
@Vennin Well after a bit of playtime, it works wonderfully. Results soon. I'm normally training concepts at 10-12 dim to preserve space, with my own anatomically correct LoRA being 14
@yLora 10-12!? does that work well? I always worry about loss of fine detail.
Thanks for the kind words though. <3
@Vennin I've experimented a lot in Kohya over the past 2 years. I've found the sacrifice isn't large for the ability to be able to use the LoRA in tandem with others. Super nice LoRA though, it's working very nicely
The best cookie baker in CivitAI, excited to try those recipes on the prompts.
Would love a spade bulge concept like transparent panties and the cookie is bulging thru👀
Hey! This lora seems to be what you're requesting and works pretty well!
https://civitai.com/models/1610023/canine-bulge
Maybe try mixing with mine? Who knows?
Question, what did you use for tagging, before training?
Kohya trainer with wd1.4. Afterward I went in and cleaned up tags. I tagged three separate data sets and then trained three separate lora. I don't remember the alphas but I had to lower the trained alpha of the spread set. Afterward I merged with the supermerger plugin on SDForge.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.