README.md

---
library_name: diffusers
license: mit
tags:
- art
---

### Model Description

This model combines the capabilities of the stable diffusion medium model with a Civit AI text-to-image model fine-tuned on a custom dataset of high-quality images. 
It aims to generate realistic and detailed images based on textual prompts.

- **Developed by:** [M.Cihan Yalçın](https://www.linkedin.com/in/chanyalcin/)
- **Model type:** Stable Diffusion
- **License:** MIT
- **Finetuned from models:**
  - [stabilityai/stable-diffusion-3-medium-diffusers](https://huggingface.co./stabilityai/stable-diffusion-3-medium)
  - [CyberRealistic](https://civitai.com/models/15003/cyberrealistic)

![Collage](imgs/photo-collage.png)

## Uses


### Direct Use

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

prompt = "A bowl of ramen shaped like a cute kawaii bear, by Feng Zikai"
negative = ""
image = pipeline(prompt,
                negative_prompt=negative).images[0]
image
```

## Bias, Risks, and Limitations

- The model may not always perfectly capture highly complex or abstract concepts.
- The quality of the output can be influenced by the specificity and clarity of the prompt.
- Ethical considerations should be taken into account when generating images to avoid misuse.


## Finetuning Details

### Finetuning Data

- Model is finetuned with sentetic high quality images collected from high performance Text-to-Image models.