Sign In

FurryArt

559
6.0k
1.2m
132
Verified:
SafeTensor
Type
Checkpoint Merge
Stats
1,222
72,316
Reviews
Published
Feb 9, 2024
Base Model
SD 1.5
Usage Tips
Clip Skip: 1
Hash
AutoV2
EA622367BF
default creator card background decoration
Luxe
A_B's Avatar
A_B

"My main motivation to creating these models, was to make more artistic or stylised furry art."

FurryArt_R ver. 3 - inprovement of the model "R ver.2"

I made this models with yiff-e18, zack3D(v1), BB95 Furry Mix, Fluffrock and many others.

Note: This is a Male NSFW/SFW furry focused model

however you can still produce females

* please, read all text to choose the best model 😉 *

About the models:

  • FurryArt_noise-offset: This is my noise-offset model. This allows to produce high dynamic range images, you can use optionally CFG_Rescale_webui extensions.

  • FurryArt_B: Similar to the previous model, but this is a no noise-offest model. Is more accurate for furry artists. Work very well with very low CFG and very high CFG scale, for this use dynamic-thresholding

    -> FurryArt_Bv2:

Features:

  1. moderate photorealism

  2. sharpness

  3. well-balanced images between photorealism and digital illustration

  4. accepts very low CFG, good for painting effect

  5. better anatomy in general

  • FurryArt_C: renders fur and hair better

    ->F. Art_C ver.2 It is an improvement of the previous model (merged with FluffyRock), more photorealistic

    EDITED: tends to have more ornamentations.

  • FurryArt_FR: This was first FluffyRock model merged. The purpose was to create a model more easy and flexible to write prompts.

    ->F. Art_FR ver.2 It is an improvement of previous model, and is more photorealistic.

  • FurryArt_D: This is a derivate of "FurryArt_B". I believe the other model is a finished project. So, I decided to create it from scratch.

    Model for soft/painting and semi photorealistic images

    • NOTE: This is not a "noise offset" model, but it seems to behave like a "noise offset". So, you can use CFG_Rescale_webui extension if you want or use this great model nLoRA to control dinamic light.

    • EDITED: Works great with very very low cfg, you can use SAG additionally

  • FurryArt_R: Model designed to be semi photorealistic, realistic/3d, but it is possible to generate other types of art, works good with "hires.fix" or using "Refiner (you need chosse any SD1.5 model)". Can be unstable with LoRas, also a bit unstable with hands.

    -> EDITED: When use C.N. Open Pose, for some reason, it seems inaccurate, so, use high weigh in negative prompt, for example: (human:1.4) and high weight positive prompt: (wolf:1.2), or use just use NegPiP extension

    • FurryArt_R ver. 2: express correction and inprovements for the model "R"

      • Anatomy stability

      • face stability

      • Lora stability

      • ControlNet stability

    • FurryArt_R ver. 3 inprovement of the model "R ver.2"

  • FurryArt_GU SPECIAL EDITION

    Model derivate from FurryArt_B and Redwater

    This is a very Artistic model, you can make smooth, painting images or very sharp, photorealistic images.

    NOTE:

    "GU" is a honour for the Redwater developer that have a nickname "Gungie"

    This project only was possible because of your help.

    https://civitai.com/models/160293?modelVersionId=180346

Recommended VAE: vae-ft-mse-840000

Edited: You can use any VAE as you want

Usefull extensions:

dynamic-thresholding to use in very high cfg scales

"Alternate Init Noise" , CD-TUNER and diffusion cg

to control colors, contrast and lighting expousure

CFG_Rescale_webui extensions to control dynamic range lighting (in offset noise models)

SAG to improve sharpness -> updated

nLoRA to control dinamic light.

NegPip a very poweful "negative prompt in prompt"

Others:

for improve the compositions:

https://github.com/Haoming02/sd-webui-mosaic-outpaint

https://github.com/v0xie/sd-webui-cads

https://github.com/dvruette/sd-webui-fabric

https://github.com/wcde/sd-webui-kohya-hiresfix.git