napari-stable-diffusion

A demo of stable diffusion in napari.

This napari plugin was generated with Cookiecutter using @napari's cookiecutter-napari-plugin template.

Installation

You can install napari-stable-diffusion via pip:

pip install napari-stable-diffusion

To install latest development version :

pip install git+https://github.com/kephale/napari-stable-diffusion.git

You will also need to sign up with HuggingFace and generate an access token to get access to the Stable Diffusion model we use.

When you have generated your access token you can either permanently set the HF_TOKEN_SD environment variable in your .bashrc or whichever file your OS uses, or you can include it on the command line

HF_TOKEN_SD="hf_aaaAaaaasdadsadsaoaoaoasoidijo" napari

For more information on the Stable Diffusion model itself, please see https://huggingface.co/CompVis/stable-diffusion-v1-4.

Apple M1 specific instructions

To utilize the M1 GPU, the nightly version of PyTorch needs to be installed. Consider using conda or mamba like this:

mamba create -c pytorch-nightly -n napari-stable-diffusion python=3.9 pip pyqt pytorch torchvision
pip install git+https://github.com/kephale/napari-stable-diffusion.git

Next steps

• Image 2 Image support
• Inpainting support

Contributing

Contributions are very welcome. Tests can be run with tox, please ensure the coverage at least stays the same before you submit a pull request.

License

Distributed under the terms of the BSD-3 license, "napari-stable-diffusion" is free and open source software

Issues

If you encounter any problems, please file an issue along with a detailed description.