Stable Diffusion Implementation, our method is built on Prompt-to-Prompt and Null-Text

Project Page   Paper

Setup

This code was tested with Python 3.8, Pytorch 1.11 using pre-trained models through huggingface / diffusers. Specifically, we implemented our method over Latent Diffusion and Stable Diffusion. Additional required packages are listed in the requirements file. The code was tested on a RTX3090 24GB but should work on other cards with at least 23GB VRAM.

You can set up the conda environment (with python3.8+CUDA11.1) using the command below:

pip install -r requirements.txt

QuickStart

To quickly generate results with different styles and different extents, you can use the following command:

bash bash/exp_styles.sh

To quickly generate results with editable style transfer, you can use the following command:

bash bash/exp_modify.sh

Alternatively, you can use the following command to generate results with specific parameters:

python cartoon_styles.py --description "A handsome man in short hair" --path example_images/portrait5.jpg --blend man --self_ratio 0.4

The --description parameter allows you to provide a description of the image. The --path parameter allows you to specify the path of the image. The --blend parameter allows you to specify the part of the image that you want to modify. The --self_ratio parameter allows you to control the strength of the style, with a range of