Celebrating my 20th model upload with some 20-inch rims.
Concept model based on 53 images of large areolae (generally aiming for 50% breast coverage), separated into different breast sizes with an equal representation of each size.
NoobAI version is trained on NoobAI ε-pred 1.0 and is untested on base ILXL.
Usage
Basic
Works best at a strength of 1.0
Base trigger is huge areolae
Breast sizes
See XYZ comparison picture pinned in gallery below.
These follow standard danbooru tagging:
flat chested(has no trained data but still works)small breastsmedium breastslarge breastshuge breastsgigantic breasts
Extras
In the interest of openness and community-helping, I have uploaded the dataset along with the model, this includes: images, captions, and the training json (for kohya_ss).
Description
trained on 53 images on NoobAI ε-pred 1.0
FAQ
Comments (15)
Pure Gold
@DrA_263 I'm testing your Lora. Looks good!
Can you include Perky breasts for the v2 version?
Sadly, short of training on AI-generated images (which I try to avoid), this is all the data that I could find over a week of looking 😢
But simply adding 'perky breasts' to the prompt seems to work for me, though people's definitions of perky vary, so I can't be sure exactly what you may be looking for!
Edit: added example image https://civitai.com/images/44903534
@DrA_263 thanks!
Excellent work!!!
I need pony version.
Could you make a PONY version?
Pleeeeease!!!
I don't really use Pony anymore, but I can see there are already quite a few for it 😮
This one for example looks quite good: https://civitai.com/models/825676/huge-areolae
I can try training it on Pony if you still really want it, but I don't really want to be usurping someone else's work :x
I have that LORA and it doesn't seem to work well for me.
Your work seems better.
I wondering if you could make it ?
@granjohnny And done o/
Have fun 😊
What a generous gentleman!
Thank you very much！
@granjohnny Not a gentleman, but you're welcome 😄
@DrA_263 can you train a Noobai V-Pred 1.0 Version? You can upload with Early Access. V-Pred is a more powerful AI engine than the Epsilon-Pred version. Moreover, it is already complete because it is the latest version.
@nokiasupreme973 Hi hi ^^
I've currently paused contributing to Civit because of recent changes made by the staff (see link on my profile for more), but I figured I better answer anyway:
I am aware of what v-pred is, I did some comparisons of it to ε-pred from 0.5 up to the 1.0 release and the detail always came out worse compared to ε-pred.
On paper, v-pred is indeed mathematically superior to ε-pred, but it seems that the NoobAI attempt at it is not very good, so I ended up sticking with ε-pred.
As such I have neither the experience nor knowledge to work with or train on v-pred.
As for Early Access, I'll never use that because I don't agree with it.
That said, my dataset is included alongside the LoRA, so if anyone else wanted to train a v-pred version they are more than welcome to, so long as they follow the same usage restrictions I have set (i.e. no monetizing the model or the images created using it, including use of early access).
@DrA_263 ok, thanks. I read this article: https://civitai.com/articles/11306
What alternative platform do you value? Tensor.Art I believe was blocking any NSFW content currently.
@nokiasupreme973 I don't have any alternate platform at the moment, because none of them seem particularly viable.
If Civit doesn't clean up its act, I'll simply keep to myself as I did before 🙂
@nokiasupreme973 Have given my response on the article now - thank you for linking it 😊
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.