If you enjoy my content, consider supporting me on Ko-Fi! ☕
Requests are always welcome in my channel in the Furry Diffusion discord! 📝
--v1 NXL-- [LoCon]
Trained & tested on NoobAI EPS v1.1.
Update for Noob/Illustrious with an updated dataset! Like before, the tail is more consistent when used alongside "raised tail".
Recommended positives: mhmitsu, blue eyes, chest tuft*, arm tuft*, leg tuft*, yellow nose**, yellow mouth**
*Optional.
**Also optional, but enforces canon coloration.
--v1 XL-- [LoRA, PonyXL]
Trained on PonyXL v6, tested on AutismMix Pony.
Everyone's favorite bubble-fox-lizard, now in XL format. :)
Recommended positives: mizutsune, blue eyes, chest tuft*, arm tuft*, yellow nose**, yellow mouth**
*Optional.
**Also optional, but enforces canon coloration.
--v7--
Trained on fluffyrock-1088-zsnr-vpred-e257-minsnr-e31-ema, tested on EasyFluff v11.2.
This was trained with V-Prediction & may produce less consistent results when used with older models that lack VPred support.
I'm quite happy with this update! The fins and tails have been improved significantly.
General positives: mizutsune, blue eyes, raised tail*, chest tuft**
*Helps generate better & more stable tails.
**I recommend a lower (~0.5) weight if you want just the chest, as it seems to like to overflow onto the belly in instances where it would be exposed on anthros. You don't need to worry about it with clothing, though.
General negatives: stylized tail*
*Not required, but pushes towards proper tails.
--v6--
Trained on fluffyrock-576-704-832-960-1088-lion-low-lr-e83-offset-noise-e68 and tested on yiffymix 3.1.
Tested at weight of 1.
Recommended positives: mizutsune, blue eyes, red pupils*, chest tuft*, tail
*if desired. Chest tuft works better with higher weight.Description
FAQ
Comments (6)
What to do to avoid deformation during sex scene generate
The features of the noob version are much more stable,nice work!
I'm curious how you upgraded the dataset?
@lse14 Nothing too special, I upped the image count to 62 while pruning out some of the less accurate images while adding more accurate ones. Also included some Violet Mizu in there, but the coloration didn't end up very stable since there was only a handful.
@Valstrix Can I take a look at the new version of the training set?
My biggest problem with Lora currently is that the training set is too large, which leads to unstable features. I want to learn how to optimize it.
@lse14 Sure, in a little while I can zip it up and send you a link like last time :)





