Introduction

Cat Collar is a text-to-image model fine-tuned from Chroma1-HD.
This model was trained with the goal of enhancing anime style.
No learning was conducted regarding the addition of characters.

Model Components

• Diffusion model: This model

• Text Encoder: T5 XXL Text Encoder or FLAN-T5-XXL

• AutoEncoder: Flux.1 AE

The model uploaded to Civitai is bf16 precision.

If you want to use Quantized model, please download them from my Hugging Face page.

• Q8_0

• Q5_K_M

How to Get Started with the Model

Please refer to the Chroma1-HD 's model card.

You need to use the webui that supports Chroma.

• ComfyUI

• Forge Neo

• Forge

Recommended settings

• Sampler: euler

• Scheduler: beta

• Steps: >=30

• CFG (guidance): 3 – 7

• Resolution: 1M pixel (1024 × 1024, 768 × 1532, 968 × 1322, 1152 × 832, and so on)

Prompt

You can use Danbooru Tags and Natural Language Prompts (English).

If you use Natural Language Prompts, I recommend the following prompt structure.

1. Begins with phrases like "Anime-style illustration of," "Anime illustration of," or "Digital illustration of"

2. Mention pivotal elements like people, objects, scenery

3. Mention concrete details like color, shape, texture, and spatial relationships

Negative Prompt

low quality, ugly, unfinished, out of focus, deformed, disfigure, blurry, smudged, restricted palette

Training Information

Please refer my Hugging Face page

Acknowledgments

• lodestones: Thanks to lodestones for sharing awesome model