An adaption to stable diffusion XL to allow for gradual diffsuion of an image allowing for modifications to be made
https://colab.research.google.com/drive/1S8NkV8NM0s_rDg2bsN-9hRM63fIwwXlY?usp=sharing
you can use anaconda or just install packages yourself with pip
Please use anaconda3 to install dependencies
from the base directory run the following command dependent on your operating system
conda create --name Diffuse-diffusion --file requirements_linux.txt
conda create --name Diffuse-diffusion --file requirements_windows.txt
If installing yourself please install the following packages using pip
pip install transformers
pip install accelerate
pip install safetensors
pip install opencv-python
pip install tqdm
pip install diffusers
pip install xformers
Run the following command from the base directory to test the software
make sure to include the whole file path in "SET_SAVE_FOLDER_HERE"
python Main.py --output_folder_name example_tree_burning --save_location SET_SAVE_FOLDER_HERE --seed_file Forest_Clean.jpg --positive_prompt "A Forest on fire, burning ravaged by wildfire" --negative_prompt "green, lush, trees"
--output_folder_name - Name of output folder
--save_location - Full path to stoare all generations
--seed_file - Full path to image to guide the generation
--positive_prompt - A prompt to gudie the diffusion process
--seed - automatically randomly generated. Set if you want repeatable generations
--negative_prompt - A prompt to gudie the diffusion process
--num_of_images - total number of images to generate (more takes longer)
--fps - total number of images per second (more makes shorter videos)
--minstrength - the amount of pixels to change at img 0
--maxstrength - the amount of pixels to change at img n where n = num_of_images
--guidance_scale - How much to deviate from the original image lower = less deviation higher = more"
--med_VRAM MED_VRAM Set if GPU VRAM is lower than 12GB
--low_VRAM LOW_VRAM Set if GPU VRAM is lower than 8GB