M

Try Stable Diffusion!

If you've been watching the recent trend of computer software that can turn text descriptions into actual images and wondering if you should get into it, my answer would be yes!

It's becoming easier to find and use these systems, and now we have Stable Diffusion, a system that you can run with great results on your local computer.

You'll need either one of the more recent M1 Macs or a Windows computer with an Nvidia GPU (used to run the machine learning model) but if you do, and many people do have these systems, then you can create artwork with some creative description sentences.

I spent yesterday trying to generate new movie posters for famous films with the additional descriptor "Polish movie poster", in the hopes of reproducing some of the amazing artwork seen in movie posters from Poland. While I didn't get close, the results are fascinating, as this Imgur gallery will show. The "Get Out" and "Poltergeist" posters are my personal favorites.

This system is fairly easy to set up, if you're not afraid of using a command-line interface. You may have to try a couple of different versions of Stable Diffusion before you encounter one that works on your particular system. I ended up using this form of the software](https://github.com/hlky/stable-diffusion) that reduces the amount of memory used, and also provides a nice browser-based UI to minimize the amount of command-line typing you need to do.

If you don't have the wherewithal to set it up locally, you can always look around for various online implementations (and I suspect it's just a matter of time before these systems show up as Docker containers or AWS images that you can set up with a simple click of a button). For now, Stable Diffusion appears to be the most accessible system right now, and if you have any interest at all in creating art, I encourage you to test it out.

(Sidebar image is the result of the text prompt "Renaissance painting of Nicolas Cage")

This article is my 92nd oldest. It is 336 words long