santa hat
deerdeer nosedeer glow
Sign In

NJfurry

815
6.8k
90
Updated: Oct 5, 2024
styleanimenijikemonoart stylefurry
Verified:
SafeTensor
Type
LoRA
Stats
465
Reviews
Published
Jul 6, 2023
Base Model
SD 1.5
Usage Tips
Clip Skip: 1
Trigger Words
nj5furry
Hash
AutoV2
136E9D9F57
default creator card background decoration
Philip's Avatar
Philip

Please add AI tag if you post images on media!

About NJ5furry-v1,v2

1500 images trained on Fluffyrock-NAI mixed model.

I used "fluffyrock-576-704-832-960-1088-lion-low-lr-e59-terminal-snr-e32.safetensors" to generate images, you can choose other versions or mix model like Indigo furry mix v45.

Use "nj5furry" tag to get Niji style images without using any artist tag! (but you can add some)

Recommended Clip Skip : 1

Recommended Weight : 0.6-1

It may be likely to generate muscular, chubby or whatever, just pull them in negative prompts.

Because of the training data, it may have difficult to generate good eyes (inpaint it XD) and complex positions or kinks.

nj5furry-v1 (1 epoch) : More NSFW but weak style?

nj5furry-v2 (7 epoch) : Strong style but less NSFW?

I'm not sure, I suggest that you could try both.

About NJ5furry_old:

460 images generated by Niji5 trained on NAI.

Works well on both anime model and furry model. (I might redo it to make it more compatible with the furry model)

Good at producing Niji5-style images with sharp lines.(more sharp lines on anime models like anything)

This lora has been merged into Indigo Furry mix v25, it's more convenient to use.

Because there are too many pose pics on niji5, so you may often generate a picture of a raised hand with a fist XD

About NJ4furry_old:

440 images generated by Niji4 trained on anything-4.5

Use it to generate furry at high weight or as a style at low weight. It depends.

Use inpainting with only mask in 0.4-0.7 denoising to fix face.

Because of the training set, it tends to generate dragons, armor and hoodie. Fill them in the negative prompt to avoid overfitting if you don't want.

The tagger words are not necessary, but you can try it.