Playing with Local AI Image Generation

Isn’t he cute?! This little guy was generated by Fooocus, which feels like a perfectly blended mashup of Stable Diffusion and Midjourney.

For those that haven’t heard of any of these, they are all AI image generation tools that allow you to create images from text. Both Stable Diffusion and Midjourney can produce fascinating results in just a few seconds, but the ways they go about achieving those results is very different for users. Stable Diffusion (not to be confused with SD Online) is meant to be run on your machine and has a complex setup process. It’s a great tool for those that are willing to tinker and learn about the process of AI image generation. On the other hand, Midjourney is a server / bot within Discord that you can send messages to and it will generate images for you. It’s a great tool for those that want to play with AI image generation without having to worry about a cumbersome setup process or those that don’t have a powerful enough machine to run Stable Diffusion.

Fooocus seems to take the best parts of both tools and combines them into a free, open source, and local tool that’s pretty easy to setup, navigate, and use. The initial setup can look a little daunting, but it’s pretty straightforward and the GitHub repo has a good walkthrough. The best part of the setup is the tool will automatically download any models you will need. Once you’re up and running, you’ll see a very simple browser interface with a prompt box and a generate button. That’s it.

If you want more, you can have that too. Clicking the Advanced button gives you the type of control you would expect from Stable Diffusion. You can change the models, the number of steps, temperature, and much more.

All this to say if you’re interested in playing with AI image generation, Fooocus is a great way to dive in.

I’ll leave you with a few of my favorite images Fooocus has generated so far.