Kou (Tamagosan1001) style | Illustrious版本v1.0 (ID: 1097542)

Please support the original creator.

This model was not created to harm or otherwise inconvenience the original creator nor Civitai or its users. It can be removed upon request.

Model based on the amazing works of Kou /Tamagosan1001 on Twitter. They are in my top 5 favourite artists and I would appreciate it greatly if you would support them on their active platforms.

Noob EPS 1.1 update

Trained on NoobAIXL EPS 1.1. Focuses more on character faces and details. Noob does a great job of translating natural language, so please make use of this feature.

Please read;

This model is a work in progress.

I will be working on it on and off until I find it satisfying enough. Its current version has a few quirks and wrinkles which need ironing. Though it works well currently, please understand that you may get unexpected results with this model.

I'm posting this model as a way to test out its limitations: if you end up using it, please leave a review or comment specifying what you could and could not do using it, what you did or did not enjoy. Use different models, parameters, combinations; basically have fun !

The training data for this model includes a balanced mixed of complex and simple images. Close-up images will mostly not suffer from lack of detail, but generating more complex images will require inpainting.

Tag quirks:

• To generate oni: use "oversized oni, oversized human" for larger creatures; "small oni" for smaller creatures.

• To generate animals: use "oversized animal" followed by which animal you'd like to generate. There are three animals which can be rendered in the original style: tiger, wolf and dragon. Other animals can be generated as well, as the style will be applied to them instead.

• Specifying facial features may make generate better face details. Even simple things such as eye color can make a difference. You can also use "face focus" to improve these details.

Known issues:

• Small details such as hands and eyes will obviously suffer in the grand scheme of things since these images have a lot of details

• Oversized oni come in different flavours, but often in the same open hand pose. This comes from an oversight I had when tagging the images; should be an easy fix

• Fingers.

Recommended settings (external gen - comfyui):

• lora weight: 1.0

• sampler: euler a/euler a cfg pp

• scheduler: sgm_uniform/normal

• steps: 40-50

• CFG: 6.0 -7.0 1.0 /1.0-2.5 1.1

• hires fix

Please enjoy, and thank you for downloading this model.

描述:

Trained on WAI-SDXL-V7. Leverages useful tags which complement the artist's style.

训练词语: