Stable Diffusion for Apple Silicon (M1/M2 Macs)

If you want to run Stable Diffusion on M1/M2 Macs, it’s actually very easy. I don’t recommend installing any of the GUI implementations like DiffusionBeeopen in new window, because while they are more user-friendly and easy to pick up as a beginner, soon you will find yourself being extremely limited to the options that you have access to, and what makes Stable Diffusion truly powerful.

Install the real deal. Metaphor: learn to use Photoshop even if it’s hard to pick up. Don’t use some random app that can only do 1% of what Photoshop can do but easier. You will be limited by it later on.

If you only have an Intel Mac, your options are extremely limited, and I don’t recommend trying to do anything on this page. If you want to try playing with Stable Diffusion on an Intel Mac, then you can try DiffusionBeeopen in new window. It will run it, even if slowly.

If you don’t have a powerful computer, you can consider using Stable Diffusion as a service, e.g. Mage.Spaceopen in new window. It will be very easy to use, but the settings might not be immediately be obvious to you. You can get very good results if you use the non-standard models, which requires a pro subscription. At $15/month for unlimited renders, it’s a steal.

Another alternative is to rent a GPU, but I don’t recommend this unless you have already used Automatic1111 before. Even if you have a slow computer, you should try to install it to familiarize yourself with the interface, before renting GPUs. Learning an interface while paying for it hourly is not very efficient. I use Vast.AI for rendering final works and batching huge jobs at high resolution. You can read my guide here.


I use a Mac. Currently, I use a maxed out M2 Max with 96GB Ram.

I didn’t buy it to run Stable Diffusion. I needed a powerful machine because I run resource-hungry programs to do my design work, which includes 2D designs, video production, 2D and 3D animations, and more. I use every single program in the Adobe Creative Cloud daily. Often, I need to open all these programs simultaneously, which are all very memory hungry.

Additionally, technology is an integral part of my design process, and I often write custom programs to assist in realizing my ideas. Example: I would write custom Python programs to drive particle animations in 3D. I also write production code for apps, and developing with multiple virtual servers for different types of services, and that means being able to run multiple docker images simultaneously, in addition to all the design apps that need to be opened.

With such a powerful machine, I was curious if I could run Stable Diffusion on it. I was skeptical, but to my surprise, it actually runs Stable Diffusion at decent speeds, around 3 IT/s for most flows. Sure, it’s a far cry from a 4090 but it had better do something if I paid 6500 for it. It‘s fast enough to do quick experiments and prompt dance inside Stable Diffusion.

To create my final renders, I would use a cloud service like Vast.AI and rent an RTX 4090 instead. You can see my instructions on how to do that on the Vast.AI page here. Since you will be using the Automatic1111 on Vast, being familiar with how to use it would be important. So even if you have a slower Mac that won’t gen at 3 IT/s, you should still try to install and play with the A1111 WebUI, so that you can run it efficiently on Vast.AI later.


Stable Diffusion is an open source library that can be controlled by programming languages, but the most natural way to use it would be through a GUI (graphical user interface). There are primarily three main implementations that are used by everyone. The popularity of these libraries mean that besides having excellent support by the core team, many developers develop plugins and extensions for them, thus greatly expanding their capabilities.

Last Updated: