Fastcasso is a web server and UI for diffusion based image generation Fastcasso makes running stable diffusion on a web server easy and fast. Fastcasso will also support other types of image generation / fine tune / asset generation in the future.
- huggingface-cli login
- If running from CLI instead of docker,
python initdb.py
- Set .env values
- POSTGRES_PWD
- DEVICE_TYPE (generally cuda or mps)
- Create appropriate conda environment depending for either windows or mac.
uvicorn main:app --reload
navigate to 127.0.0.1:8000/docs
- Can also run
python bootstrap.py
docker build -t fastcasso .
docker run -dp 8000:8000 fastcasso
- Default settings / pipelines are tested against CUDA device types on windows.
- Mac also supported, simply swap out for MPS device type.
- Some Hardcoded presets include stable diffusion 1.4, 1.5, 2.0 and 2.1 with certain fine tunes. 512 and 768 resolution on certain diffuser pipelines.
- Supports search on generated images.
- Supports custom presets. (used for specifying parameters, keywords, negative keywords.)
- UI for testing / exploration.
- Transform UI for performing Pix2pix operations on existing images.
- Fastcasso will support the following features in the future (view issues for exact list):
- Textual embeddings.
- Custom Model Training.
- Queued Jobs.
- In and Out painting.
- Advanced Image Editing.