Skip to content

Latest commit

 

History

History
307 lines (256 loc) · 8.09 KB

README.md

File metadata and controls

307 lines (256 loc) · 8.09 KB
Raw

Stable Diffusion REST API npm Version stability experimental build

Run Stable Diffusion locally via a REST API on an M1/M2 MacBook

Pre-requisites

Initial setup

Adapted from Run Stable Diffusion on your M1 Mac’s GPU by Ben Firshman

Update Homebrew and upgrade all existing Homebrew packages:

brew update
brew upgrade

Install Homebrew dependencies:

brew install cmake mkcert protobuf rust

Clone the Stable Diffusion fork:

git clone -b apple-silicon-mps-support https://github.com/bfirsh/stable-diffusion.git

Set up a virtualenv and install dependencies:

cd stable-diffusion
python3 -m pip install virtualenv
python3 -m virtualenv venv
pip install -r requirements.txt
cd ..

Download the text-to-image and inpaint model checkpoints:

mkdir -p models
curl --output models/text-to-image.ckpt https://www.googleapis.com/storage/v1/b/aai-blog-files/o/sd-v1-4.ckpt?alt=media
curl --output models/inpaint-image.ckpt https://ommer-lab.com/files/latent-diffusion/inpainting_big.zip

Generate a HTTPS certificate and key:

mkcert -install
mkcert -cert-file cert.pem -key-file key.pem 0.0.0.0

Usage


  Usage
    $ stable-diffusion-rest-api [options]

  Options
    --cert                   Path to the SSL certicate  (default ./cert.pem)
    --concurrency            Number of concurrent image generation tasks  (default 1)
    --cors                   Whether to enable CORS  (default true)
    --delete-incomplete      Delete all incomplete image generation tasks before starting the server  (default false)
    --inpaint-image-model    Path to the inpaint image model checkpoint  (default ./models/inpaint-image.ckpt)
    --key                    Path to the SSL certicate key  (default ./key.pem)
    --output                 Directory to output generated images  (default ./output)
    --port                   Port to serve the REST API  (default 8888)
    --repository             Path to the Stable Diffusion repository  (default ./stable-diffusion)
    --text-to-image-model    Path to the text-to-image model checkpoint  (default ./models/text-to-image.ckpt)
    -v, --version            Displays current version
    -h, --help               Displays this message

Start the API server:

npx --yes -- stable-diffusion-rest-api \
  --text-to-image-model ./models/text-to-image.ckpt \
  --inpaint-image-model ./models/inpaint-image.ckpt \
  --concurrency 1 \
  --output ./output \
  --cert ./cert.pem \
  --key ./key.pem \
  --port 8888

API response

All REST API endpoints return JSON with one of the following shapes, depending on the status of the image generation task:

{
  status: 'QUEUED'
  resultUrl: string
}
{
  status: 'IN_PROGRESS'
  resultUrl: string
  progress: {
    currentImageIndex: number
    currentImageProgress: number
    totalImages: number
  }
  imageUrls: Array<string>
}
{
  status: 'COMPLETE'
  resultUrl: string
  imageUrls: Array<string>
}
  • status is one of QUEUED, IN_PROGRESS or COMPLETE
  • resultUrl is the URL to access the results of the image generation task
  • progress contains details about the progress of the image generation task:
    • currentImageIndex is the index of the image currently being generated
    • currentImageProgress is a value between 0 and 1 representing the progress of generating the current image
    • totalImages is the total number of images to be generated
  • imageUrls is the list of URLs of the generated images

Text to Image

POST /text-to-image

curl https://0.0.0.0:8888/text-to-image \
  --form prompt="A digital illustration of a beautiful mountain landscape, detailed, thom tenerys, epic composition, 4k, trending on artstation, fantasy vivid colors" \
  --form iterations="3" \
  --form steps="8" \
  --form seed="42" \
  --header "Content-Type: multipart/form-data" \
  --location

Sample response

{
  "status": "QUEUED",
  "resultUrl": "/text-to-image/61f957e4462ea8eff36d9e7a7b650994"
}

GET /text-to-image/<ID>

curl https://0.0.0.0:8888/text-to-image/61f957e4462ea8eff36d9e7a7b650994

Sample response

{
  "status": "IN_PROGRESS",
  "resultUrl": "/text-to-image/61f957e4462ea8eff36d9e7a7b650994",
  "progress": {
    "totalImages": 3,
    "currentImageIndex": 3,
    "currentImageProgress": 0.5
  },
  "imageUrls": [
    "/text-to-image/61f957e4462ea8eff36d9e7a7b650994/1.png",
    "/text-to-image/61f957e4462ea8eff36d9e7a7b650994/2.png"
  ]
}
{
  "status": "COMPLETE",
  "resultUrl": "/text-to-image/61f957e4462ea8eff36d9e7a7b650994",
  "imageUrls": [
    "/text-to-image/61f957e4462ea8eff36d9e7a7b650994/1.png",
    "/text-to-image/61f957e4462ea8eff36d9e7a7b650994/2.png",
    "/text-to-image/61f957e4462ea8eff36d9e7a7b650994/3.png"
  ]
}

Image to Image

POST /image-to-image

curl https://0.0.0.0:8888/image-to-image \
  --form prompt="A digital illustration of a beautiful mountain landscape, detailed, thom tenery, epic composition, 4k, trending on artstation, fantasy vivid colors" \
  --form image=@./image.png \
  --form iterations="3" \
  --form steps="24" \
  --form seed="42" \
  --header "Content-Type: multipart/form-data" \
  --location

Sample response

{
  "status": "QUEUED",
  "resultUrl": "/image-to-image/ab1104f3b55fbab7779cdbdc73ed276e"
}

GET /image-to-image/<ID>

curl https://0.0.0.0:8888/image-to-image/ab1104f3b55fbab7779cdbdc73ed276e

Sample response

{
  "status": "IN_PROGRESS",
  "resultUrl": "/image-to-image/ab1104f3b55fbab7779cdbdc73ed276e",
  "progress": {
    "totalImages": 3,
    "currentImageIndex": 3,
    "currentImageProgress": 0.5
  },
  "imageUrls": [
    "/image-to-image/ab1104f3b55fbab7779cdbdc73ed276e/1.png",
    "/image-to-image/ab1104f3b55fbab7779cdbdc73ed276e/2.png"
  ]
}
{
  "status": "COMPLETE",
  "resultUrl": "/image-to-image/ab1104f3b55fbab7779cdbdc73ed276e",
  "imageUrls": [
    "/image-to-image/ab1104f3b55fbab7779cdbdc73ed276e/1.png",
    "/image-to-image/ab1104f3b55fbab7779cdbdc73ed276e/2.png"
    "/image-to-image/ab1104f3b55fbab7779cdbdc73ed276e/3.png"
  ]
}

Inpaint Image

POST /inpaint-image

curl https://0.0.0.0:8888/inpaint-image \
  --form image=@./image.png \
  --form mask=@./image-mask.png \
  --form steps="32" \
  --form seed="42" \
  --header "Content-Type: multipart/form-data" \
  --location

Sample response

{
  "status": "QUEUED",
  "resultUrl": "/inpaint-image/59a89dfc9f075942ce9afc08312b8296"
}

GET /inpaint-image/<ID>

curl https://0.0.0.0:8888/inpaint-image/59a89dfc9f075942ce9afc08312b8296

Sample response

{
  "status": "IN_PROGRESS",
  "resultUrl": "/inpaint-image/59a89dfc9f075942ce9afc08312b8296",
  "progress": {
    "totalImages": 1,
    "currentImageIndex": 1,
    "currentImageProgress": 0.5
  },
  "imageUrls": []
}
{
  "status": "COMPLETE",
  "resultUrl": "/inpaint-image/59a89dfc9f075942ce9afc08312b8296",
  "imageUrls": [
    "/inpaint-image/59a89dfc9f075942ce9afc08312b8296/1.png",
  ]
}