Cleora

shumaiishumaii
48
479
base model
v1WIP
model-image-0
model-image-1
model-image-2
model-image-3
model-image-4
model-image-5
model-image-6
model-image-7
model-image-8
model-image-9
model-image-10
model-image-11
model-image-12

Details

Download Files (1)

Model description

Redistribution is prohibited.

If you like my work, consider giving me more support, likes, buzz, feedback and reactions.

ABOUT THE MODEL

20260615 v1

Trained on Anima Base v1

Features

  • Modern Cel-Shading like / Semi-Impasto game illustrations.

This version leans more toward semi-Impasto. Further versions may consider distinguishing more styles/modules, such as a more dedicated cel shading-like style, semi-Impasto, and others.

  • Clear and clean visuals, with smooth, distortion-free contour lines and balanced brushstroke strength, while reducing JPEG artifacts.

  • No random, inexplicable noise/white dots that make the image look obviously AI-generated.

  • No AI-looking plastic appearance.

As shown in the image below, the left side shows the default effect of the base model (without adding any stylization tags), while the right side shows Cleora. When the style is Impasto-like, these JPEG artifacts are not particularly noticeable. However, when the style is cel-shading-like or has more prominent contour lines, such JPEG artifacts are especially likely to appear in common models: the contour lines may have overly uneven sharpening and chaotic details. This phenomenon is one of the major reasons why images look obviously AI-generated.

  • Retains most of the prior knowledge from Anima, while strengthening/adding some character knowledge that I personally prefer. More knowledge will be added in the future.

  • Most of the data and captions in the dataset have been manually edited and processed by me, including both boorutags and natural language.

  • More modules will be added gradually.

Issues

  • The current model is not yet designed for diverse outputs.

Its style is relatively fixed, and the visual impact still needs further improvement. The model has some forgetting issues with tags such as 3D and anime screenshot, so the response to these prompts may not be particularly strong.

  • The current dataset is not large, so there may be some degree of forgetting or overfitting.

  • When the output becomes blurry, consider using MultiDiffusion to fix it

  • When the face becomes blurry, consider using ADetailer to fix it

*A unique tag has been baked into the model as a special response trigger.

*If this model is used for a merge, please disclose the merge materials.

Tag

The quality tags below are sufficient; there is no need to add too many additional quality tags.

Positive Prompt (Must Include)

masterpiece,best quality,amazing quality,

Negative Prompt

bad quality,worst quality,worst detail,
bad anatomy,bad hands,extra digits,
watermark,

Requirements

Checkpoints: Anima

Recommended Tools: ForgeNeo, ComfyUI

Recommended Step: 20~30

Recommended size: 768*1344, 832*1216, 896*1156, 1024*1024, 1344*768, 1216*8

32, 1156*896

ARCHIVES

20260529 WIP

Trained on Anima Base v1

Features

  • Aesthetic fine-tuning, WIP

  • The goal is to produce clean, highly controllable modern cel-shaded like game illustrations

  • more modules will be added gradually

Images made by this model