Domc.Dev
Published on

Three easy ways to run stable diffusion

Authors

If you're like me, I'm sure you've spent the last few months messing about with the lates AI trend - text-to-image ML. The most popular and arguably the easiest way to start playing about with this tech is DALLE 2 from OpenAI. While this is an easy, open (sort of), and free (sort of) way to get started, this post is about the alternatives to DALLE and how you can set them up as simply as possible.

Stable Diffusion on Google Colab

You can spin up and access some GPU power for free using Google Colab. Once you've set up an account, open up this notebook to get started.

Take a look at the repo here for instructions if you need them.

Stable Diffusion UI wrapper arround Colab

For a nicer user experience, you can plug your details into this handy web UI made by Altryne that gives you a nice gui to play with the model. Check out the repo here.

UI Wrapper

It also syncs all your generated images into Google Drive automatically. Handy!.

Note: If you want to use image-to-image, make sure that your input image is 512 x 512 or else it will throw an error.

Stable Diffusion running on Apple Sillicon

If your lucky enough to have a shiny new mac with apple silicon, why not try running Diffusion Bee, the new M1/M2 GUI of stable diffusion from divamgupta. It's a nice one click install.

stableimage

While it's missing a few features, it's a great way to get started.

Stable Diffusion running on Windows w/ Nvidia

Does not work with AMD

You can download this one click install GUI called NMKD.

windowsstable

I can't test this out myself as I'm not currentlty running a Windows machine, but I have heard this is a great option if you have an Nvidia GPU with at least 10GB.