---
license: apache-2.0
library_name: diffusers
tags:
- text-to-image
- stable-diffusion
- diffusers
widget:
- text: A samurai warrior in traditional armor holding a katana.
  output:
    url: images/000.png
- text: A vase of camellias
  output:
    url: images/001.png
- text: The geysers of Enceladus
  output:
    url: images/002.png
- text: a serene temple by a lake
  output:
    url: images/003.png
- text: A fierce lion roaring in the savanna.
  output:
    url: images/004.png
- text: a quaint lakeside cabin.
  output:
    url: images/005.png
instance_prompt: A fierce lion roaring in the savanna.
datasets:
- Chan-Y/Mangaset
pipeline_tag: text-to-image
---

# MangaDF

<Gallery />

## Model Description

**Base Model:** [stabilityai/stable-diffusion-3-medium](https://huggingface.co./stabilityai/stable-diffusion-3-medium)  
**Finetuned On:** Images created by [alvdansen/BandW-Manga](https://huggingface.co./alvdansen/BandW-Manga)  

## Overview

MangaDF is a finetuned version of the ['stabilityai/stable-diffusion-3-medium'](https://huggingface.co./stabilityai/stable-diffusion-3-medium) model. It has been specifically trained to generate bold line portrait illustrations, inspired by manga-style art. The finetuning dataset was created using images generated by the ['alvdansen/BandW-Manga'](https://huggingface.co./alvdansen/BandW-Manga) stable diffusion model, ensuring high-quality and stylistically consistent outputs.


## Intended Use

This model is designed for artists, illustrators, and enthusiasts who want to create manga-style bold line portraits. It can be used for various purposes, including concept art, character design, and graphic novels.

## How to Use

To use this model, you need to load it with the diffusers library in Python. Here's a quick start guide:

```python
from diffusers import DiffusionPipeline
import torch
pipeline = DiffusionPipeline.from_pretrained(
    "Chan-Y/MangaDF",
    torch_dtype=torch.float16
).to("cuda")

# Generate an image
prompt = "A bold line portrait of a fantasy character"
image = pipeline(prompt).images[0]
image
```

## Training Data

The finetuning dataset was created using images generated by the `alvdansen/BandW-Manga` stable diffusion model, which specializes in generating black and white manga-style illustrations. This dataset ensures that the MangaDF model produces high-quality and consistent manga-style bold line portraits.

## Contact

For any questions or issues, please contact [me](https://www.linkedin.com/in/chanyalcin/)