|
|
--- |
|
|
license: openrail++ |
|
|
language: |
|
|
- en |
|
|
library_name: diffusers |
|
|
pipeline_tag: text-to-image |
|
|
tags: |
|
|
- text-to-image |
|
|
base_model: |
|
|
- suzushi/miso-diffusion-xl-1.0 |
|
|
--- |
|
|
<div style="display: flex; justify-content: center; gap: 20px; margin-bottom: 20px;"> |
|
|
<img src="demo/demo1.png" width="400" /> |
|
|
<img src="demo/demo2.png" width="400" /> |
|
|
</div> |
|
|
|
|
|
# Anime SDXL Model |
|
|
|
|
|
A Stable Diffusion XL model fine-tuned for generating high-quality anime-style images. |
|
|
|
|
|
## Version History |
|
|
|
|
|
| Version | Base Training | Aesthetic Training | Total Epochs | |
|
|
|---------|--------------|-------------------|--------------| |
|
|
| 1.0 | 160K images | 10K images | 5 | |
|
|
| 1.1 | 200K images | 12K images | 5 | |
|
|
| 1.2 | - | 23K images | 9 | |
|
|
|
|
|
## Training Methodology |
|
|
|
|
|
The model underwent a multi-stage training process: |
|
|
|
|
|
1. **Base Pre-training** |
|
|
- Initial training on a diverse dataset of anime-style images |
|
|
- Focus on learning fundamental anime art styles and characteristics |
|
|
|
|
|
2. **Aesthetic Fine-tuning** |
|
|
- Secondary training phase focusing on artistic quality and consistency |
|
|
- Curated dataset of high-quality anime artwork |
|
|
- Progressive improvements across versions |
|
|
|
