Watch the video:
0:00 - Explanation, and details
0:35 - Install Homebrew
1:28 - Downloading AUTOMATIC1111's Stable Diffusion WebUI
1:54 - Downloading Stable Diffusion models
3:28 - Starting A1's SDUI on Mac
5:00 - Using Stable Diffusion on a Mac
5:40 - Drawbacks of SD on Mac
5:50 - Launch arguments & Less VRAM
7:35 - Opening SDUI in the future
8:20 - Is Mac good for SDUI? Not really...Stable Diffusion on a Mac?
Well, it’ll be slow… But with a more powerful mac setup, this is something you’d want to consider doing, especially if you don’t have access to a PC, or even one as powerful as the mac or Macbook you have.
By the end you should hopefully be generating images happily!
Notes
Some features like the CLIP interrogator and training/Dreambooth may not work, or work properly. This is experimental, somewhat.
This article is based off the Discussion on the AUTOMATIC1111’s Stable Diffusion Discussions page - In this thread, posted by @brkirch.
Installing
1. Brew
You will need the package manager Brew (https://brew.sh/) to get started. Download and install it by running the command:
| |
2. Required packages
Install the required packages by running the following command (all of these should be run in the same terminal window).
The second command will download the Stable Diffusion GitHub repo.
The third command takes us into the correct directory, and the final one opens the folder in Finder. We’ll need this in the next step.
| |
3. Downloading some Stable Diffusion models
Now, something for Stable Diffusion to work with.
We can download and place any Stable Diffusion model in the ./models folder. So, go ahead and do that now. For example, the Stable Diffusion 1.5 model from HuggingFace. Click the Files and versions tab, and then next to v1-5-pruned.ckpt click the small arrow (should be just beside the file size in the same row).
Then, just drag the ckpt or safetensors file into the stable-diffusion-webui/models folder. Your finder should be open from the previous step.
Other downloads include (but are not limited to):
- Stable DIffusion 1.4 (sd-v1-4.ckpt)
- Stable Diffusion 1.5 (v1-5-pruned-emaonly.ckpt)
- Stable Diffusion 1.5 Inpainting (sd-v1-5-inpainting.ckpt)
You can also dive into SD 2.0+. Just make sure to use a width and height of 768px and above, and you will also need to download a configuration file from the links too!
4. Installing AUTOMATIC1111’s Stable Diffusion
Simply run the command ./webui.sh.
If you closed the terminal, open the same folder (assuming you didn’t navigate anywhere before starting this guide) by running cd ~/stable-diffusion-webui first.
Otherwise, for future reference, just remember to run ~/stable-diffusion-webui/webui.sh to open it from a terminal anywhere.
This will then download and set up the required packages. Get some coffee ready as you may be waiting a while. It’s usually not so bad.
5. Updating AUTOMATIC1111’s Stable Diffusion
To update, run git pull in the ~/stable-diffusion-webui folder.
The repo doesn’t automatically update, but there are tons of commits (updates) pushed every day (usually). Just make sure to update it once in a while for better performance and the latest features.
Sometimes it may require things to update on the next open, so do keep in mind that things may require a download after pulling a new update.
6. Better performance and customization
Your system may not be cut out to run SDUI to it’s best ability… That’s alright. We can tweak some options.
Just open ~/stable-diffusion-webui/webui-user.sh in a text editor. Then uncomment the line that starts with #COMMANDLINE_ARGS= by removing the hash, so it appears as COMMANDLINE_ARGS= instead.
Then, in the quotes just after, enter command line arguments to help tell the program what to do. For example:
--medvram or --lowvram.
Other options include: --opt-split-attention-v1, --skip-torch-cuda-test, --no-half, --use-cpu all. There are many options, each with their tradeoffs (usually time generating an image).
For further troubleshooting and updates to the guide, please see the Discussion this guide was based off.
