/Anime-SD

Anime style SD

Primary LanguagePython

Anime-style Stable Diffusion!

Ever wanted to generate images of a particular style with the current state of the art stable diffusion models? With this repo, you can generate all types of styles.

Image 1
Girl with bunny ears 		Image 2
A man wearing a mask 		Image 2
A powerful man with fire around his body. 		Image 2
An old portrait of a woman
Image 1
A girl playing with a dog 		Image 1
Two boys celebrating a soccer goal 		Image 2
A portrait of a blonde man 		Image 2
A powerful female warrior

Run the model locally:

  • Install dependencies
git clone https://github.com/RSDP101/Anime-SD.git
pip install virtualenv
python -m venv venv
source venv/bin/activate
cd Anime-SD/examples/text_to_image/
pip install git+https://github.com/huggingface/diffusers.git
pip install -U -r requirements.txt
accelerate config default
  • Run training on a SD model and dataset. The default ones are 'CompVis/stable-diffusion-v1-4' and rod101/Anime1K.
# Change model and dataset configuration at the training.py script.
python3 training.py

Training will save model's weight to sd-anime-model.

  • Run inference on your fine-tuned model.
python3 inference.py

You can change model and dataset configuration at the training.py script. Pass the prompt you want to generate at the inference.py script.

Enjoy!