Local Image Generation API for MacOS

This FastAPI Python project enables local image generation using Stable Diffusion/Flux on an M-based Mac. It leverages Diffusion Kit to utilize onboard GPUs.

Image Quality Settings:

Default image quality settings:

• Low: Uses Sd3-medium model, 256x256 resolution, 30 inference steps.
• Medium: Uses Sd3-medium model, 512x512 resolution, 50 inference steps.
• High: Uses Flux-schnell model, 1024x1024 resolution, 4 inference steps.
• Extreme: Uses Flux-Schnell model, 1280x1280 resolution, 6 inference steps.

Note: The Flux model requires 26 GB of RAM. Ensure adequate RAM before using high or extreme settings.

The image quality settings can be adjusted in the config.ini file.

Running the API

This project uses Poetry for dependency management.

1. Create a new virtual environment.
2. Install dependencies with poetry install.
3. Run the server with make runDev.
4. Visit http://127.0.0.1:8000/docs to view and test the API.

Alternatively, you can use curl to access the API:

curl -X 'POST' \
  'http://127.0.0.1:8000/image' \
  -H 'accept: application/json' \
  -H 'x-token: secret-token' \
  -H 'Content-Type: application/json' \
  --output 'image.png' \
  -d '{
  "prompt": "a photograph of an astronaut walking on the moon",
  "image_quality": "medium"
}'

Image generation times:

• Low quality: ~30-40 seconds
• Medium quality: ~1.5 minutes
• High quality: ~1.2 minutes
• Extreme quality: ~3 minutes

The generation times were obtained by running the API on a M3 Mac

Environment Variables

• LOCAL_IMAGE_GEN_API_KEY: The API key for the local image generation API. Set this to any value and use the same value for the x-token parameter when calling the API.

By default, running the application with make runDev sets the API key to secret-token.