Setup

Make sure you have the latest version of docker installed

TLDR:

docker compose --profile download up --build
# wait until its done, then:
docker compose --profile [ui] up --build
# where [ui] is one of: auto | auto-cpu | hlky | lstein

More details:

Automatic setup

Download this repo and run:

docker compose --profile download up --build

This will download all of the required models / files, and validate their integrity. You only have to download the data once (regardless of the UI). There are roughly 8GB of data to be downloaded.

Next, choose which UI you want to run (you can easily change later):

• hlky: The most popular fork, Repo
• auto: Another great fork, many features with neat UI, Repo
• auto-cpu: for users without a GPU.
• lstein: One of the earliest forks, the UI is simple but has a lot of potential, Repo

After the download is done, you can run the UI using:

docker compose --profile [ui] up --build
# for example:
# docker compose --profile auto up --build
# or
# docker compose --profile hlky up --build

Will start the app on http://localhost:7860/. Feel free to try out the different UIs.

Note: the first start will take some time since other models will be downloaded, these will be cached in the cache folder, so next runs are faster. First time setup might take between 15 minutes and 1 hour depending on your internet connection, other times are much faster, roughly 20 seconds.

If you face any problems, check the FAQ page, or create a new issue.

Manual Download

Download the models:

• Stable Diffusion v1.4 (4GB), rename to model.ckpt
• GFPGANv1.3.pth (333MB).
• RealESRGAN_x4plus.pth (64MB) and RealESRGAN_x4plus_anime_6B.pth (18MB).
• LDSR (2GB) and its configuration, rename to LDSR.ckpt and LDSR.yaml respectively.

Put all of the downloaded files in the cache/models folder (create the folder if not there), it should look something like this:

cache/models/
├── model.ckpt
├── GFPGANv1.3.pth
├── RealESRGAN_x4plus.pth
├── RealESRGAN_x4plus_anime_6B.pth
├── LDSR.ckpt
└── LDSR.yaml