diffuzers

A web ui and deployable API for diffusers.

< under development, request features using issues, prs not accepted atm >

If something doesnt work as expected, or if you need some features which are not available, then create request using github issues

Features available in the app:

• text to image
• image to image
• instruct pix2pix
• textual inversion
• inpainting
• outpainting (coming soon)
• image info
• stable diffusion upscaler
• gfpgan
• clip interrogator
• more coming soon!

Features available in the api:

• text to image
• image to image
• instruct pix2pix
• textual inversion
• inpainting
• outpainting (via inpainting)
• more coming soon!

Installation

To install bleeding edge version of diffuzers, clone the repo and install it using pip.

git clone https://github.com/abhishekkrthakur/diffuzers
cd diffuzers
pip install -e .

Installation using pip:

pip install diffuzers

Usage

Web App

To run the web app, run the following command:

diffuzers app

API

To run the api, run the following command:

diffuzers api

Starting the API requires the following environment variables:

export X2IMG_MODEL=stabilityai/stable-diffusion-2-1
export DEVICE=cuda

If you want to use inpainting:

export INPAINTING_MODEL=stabilityai/stable-diffusion-2-inpainting

To use long prompt weighting, use:

export PIPELINE=lpw_stable_diffusion

If you have OUTPUT_PATH in environment variables, all generations will be saved in OUTPUT_PATH. You can also use other (or private) huggingface models. To use private models, you must login using huggingface-cli login.

API docs are available at host:port/docs. For example, with default settings, you can access docs at: 127.0.0.1:10000/docs.

All CLI Options for running the app:

 diffuzers app --help
usage: diffuzers <command> [<args>] app [-h] [--output OUTPUT] [--share] [--port PORT] [--host HOST]
                                        [--device DEVICE] [--ngrok_key NGROK_KEY]

 Run diffuzers app

optional arguments:
  -h, --help            show this help message and exit
  --output OUTPUT       Output path is optional, but if provided, all generations will automatically be saved to this
                        path.
  --share               Share the app
  --port PORT           Port to run the app on
  --host HOST           Host to run the app on
  --device DEVICE       Device to use, e.g. cpu, cuda, cuda:0, mps (for m1 mac) etc.
  --ngrok_key NGROK_KEY
                        Ngrok key to use for sharing the app. Only required if you want to share the app

All CLI Options for running the api:

 diffuzers api --help
usage: diffuzers <command> [<args>] api [-h] [--output OUTPUT] [--port PORT] [--host HOST] [--device DEVICE]
                                        [--workers WORKERS]

 Run diffuzers api

optional arguments:
  -h, --help         show this help message and exit
  --output OUTPUT    Output path is optional, but if provided, all generations will automatically be saved to this
                     path.
  --port PORT        Port to run the app on
  --host HOST        Host to run the app on
  --device DEVICE    Device to use, e.g. cpu, cuda, cuda:0, mps (for m1 mac) etc.
  --workers WORKERS  Number of workers to use

Using private models from huggingface hub

If you want to use private models from huggingface hub, then you need to login using huggingface-cli login command.

Note: You can also save your generations directly to huggingface hub if your output path points to a huggingface hub dataset repo and you have access to push to that repository. Thus, you will end up saving a lot of disk space.