simple-diffusion
Stable Diffusion Web UI without Web UI
Stars
2
Ecosystems:
Python
The implementation of the stable diffusion Web UI is more efficient than diffusers. I like its rich features, but the library mixes UI, API, and other logic, making it very complex. For learning purposes, I want to extract its core generation logic and separate it from the UI and API components.
I didn't use the Web UI directly; instead, I used Web UI Forge. This is because I wanted to understand its patcher methodology (maybe from ComfyUI), which appears to provide better extensibility for the models.
For test:
python ./simplediffusion/__init__.py
Coverage:
coverage run .\simplediffusion\__init__.py; coverage xml
Roadmap:
- Stable Diffusion XL Support
- Basic text to image
- Basic image to image
- Reference only
- ControlNet
- Other forge built-in extensions
- Stable Diffusion 3 Support