Stable diffusion with upscaler

Environment

Supported on Python 3.8 to 3.11. Expected to run on a GPU.

How to install

1. Clone this repo

git clone https://github.com/y26805/stable_diffusion_with_upscaler.git

2. Install dependencies

For poetry users, run

cd stable_diffusion_with_upscaler/
poetry install

For pip users, run

cd stable_diffusion_with_upscaler/
pip install -e .

3. Clone related directories

git clone https://github.com/CompVis/stable-diffusion
git clone https://github.com/CompVis/taming-transformers
git clone https://github.com/CompVis/latent-diffusion

How to run

Basic example

(Optional) for poetry users, run

poetry shell

Running the follow will save a file in the outputs/ folder.

python3 stable_diffusion_with_upscaler/run.py --prompt "panda in space"

Output images have a resolution of 1024x1024 pixels.

More detailed example

python3 stable_diffusion_with_upscaler/run.py --prompt "panda in space" --steps 50 --seed 123 --outdir myfolder

This generates an image with the text promot panda in space, with 50 training steps and seed set as 123. The image is saved to myfolder/.