Pseudo Ink Wash for Furry (NoobAI-XL)

LORA
Reprint


Updated:

An experimental sloppy style lora that introduces furry into traditional ink wash with wavy ink outlines and diluted shading.

It was trained on vpred model NoobAIXL_vPred10 and has better color/style resemblance than the one I trained on eps version. It's still very usable on the eps model NoobAIXL_epsilonPred11.

Trigger Word

traditional ink wash, or traditional ink wash, style-d for extremity. Down weight or up weight depending on the base model's bias.

Inference

  • CFG: 5 for normal sampler, 2 for CFG++ sampler

  • Schedule: For noob's vpred models, use beta/uniform/simple, avoid karras/exponential.

Negative Prompt

  • It's not necessary to use negative quality tags on base Noob for the most of the time. worst quality, normal quality will make the strokes drier.

  • Noob has strong style bias for some copyrighted characters. For example, generic style tags like toony in negatives might help for cartoon character.

  • When applied to merged models with a strong default style, the ink wash style might be diluted while the consistency will improve. For Indigo XL Vpred 4, I figured it helps to use less colors and put glistening clothing, soft shading etc in negatives.

Auxiliary Styles for Training

Since traditional paintings are more focused on feral animals and human, other auxiliary digital styles are trained along to help introduce anthropomorphic features. They are not the targets of this lora but are partially present in it:

  • digital ink wash, style-a: craggy brush stroke with cel shading

  • digital ink wash, style-b: smooth brush stroke with flat shading

  • digital ink wash, style-c: inconsistent brushlike outlines with smooth shading

Lora Author: https://civitai.com/user/VahlokAlunmid.

Version Detail

SDXL 1.0
1520
10
An experimental sloppy style lora that introduces furry into traditional ink wash with wavy ink outlines and diluted shading. It was trained on vpred model NoobAIXL_vPred10 and has better color/style resemblance than the one I trained on eps version. It's still very usable on the eps model NoobAIXL_epsilonPred11. Trigger Word traditional ink wash, or traditional ink wash, style-d for extremity. Down weight or up weight depending on the base model's bias. Inference CFG: 5 for normal sampler, 2 for CFG++ sampler Schedule: For noob's vpred models, use beta/uniform/simple, avoid karras/exponential. Negative Prompt It's not necessary to use negative quality tags on base Noob for the most of the time. worst quality, normal quality will make the strokes drier. Noob has strong style bias for some copyrighted characters. For example, generic style tags like toony in negatives might help for cartoon character. When applied to merged models with a strong default style, the ink wash style might be diluted while the consistency will improve. For Indigo XL Vpred 4, I figured it helps to use less colors and put glistening clothing, soft shading etc in negatives. Auxiliary Styles for Training Since traditional paintings are more focused on feral animals and human, other auxiliary digital styles are trained along to help introduce anthropomorphic features. They are not the targets of this lora but are partially present in it: digital ink wash, style-a: craggy brush stroke with cel shading digital ink wash, style-b: smooth brush stroke with flat shading digital ink wash, style-c: inconsistent brushlike outlines with smooth shading

Project Permissions

Model reprinted from : https://civitai.com/models/1511289/pseudo-ink-wash-for-furry-noobai-xl?modelVersionId=1709552

Reprinted models are for communication and learning purposes only, not for commercial use. Original authors can contact us to transfer the models through our Discord channel --- #claim-models.

Related Posts