ComfyUI Stable Diffusion Installation for Apple Silicon (M1/M2/M3)

Installation Guide: Setting Up ComfyUI on Apple Silicon for Seamless Local Image Generation on Your Computer

4 min readDec 28, 2023

This article walks through all the steps to get ComfyUI installed on Apple Silicon & guides you all the way to loading in models and generating images.

I tested this process on an M1 Mac (32GB).

Thanks to:

Thanks to AIAnimation and Chukwubuikem for inspiring this post with their great YouTube tutorials.

Requirements

Mac with Apple Silicon (M1 / M2 / M3)
Homebrew installed (run brew doctor to check if you have it installed)
Python 3.11
You should be comfortable with the command line

Installation

Go to Pytorch and copy the Command you get for this configuration:

Copy the command you get when you go to PyTorch: https://pytorch.org/get-started/locally/

2. Create an comfy-ui folder in a convenient directory: mkdir comfy-ui

3. Move into the folder: cd comfy-ui

4. Git clone: git clone https://github.com/comfyanonymous/ComfyUI.git

5. Move into the folder you cloned: cd ComfyUI

6. Install requirements: pip install -r requirements.txt

7. To Start the ComfyUI, simply run python3.11 main.py from within the ComfyUI directory

To see the GUI go to: http://127.0.0.1:8188

Great! Now you have your Graphical User Interface (GUI) running when you go to: http://127.0.0.1:8188
The next step is to install some weights / models / checkpoints so you can start generating images.

Install Models

Download a model from HuggingFace — for this article, we’ll use v1–5-pruned-emaonly.safetensors

Download v1–5-pruned-emaonly.safetensors

2. Install the model in this directory: /Users/<path/to>/comfy-ui/ComfyUI/models/checkpoints/

3. Reload the UI and then select the new model in the Load Checkpoint node (thanks Goeran!)

4. Add a prompt and click “Queue Prompt”. After ~30 seconds you will see your image.

Prompt: “photo of young Caucasian woman highlight hair, sitting outside restaurant, wearing dress, rim lighting, studio lighting, looking at the camera, dsir, ultra quality, sharp focus, tack sharp, dof, film grain, Fujifilm XT3, crystal clear, 8K UHD, highly detailed glossy eyes, high detailed skin, skin pores”

Image Closeup from Model: v1–5-pruned-emaonly.safetensors

Trying a Different Model

I also tried the same prompt as above, but swapped out the model for DreamShaper_8. The file I downloaded was “dreamshaper_8.safetensors” and I placed it in the directory: /Users/<path/to>/comfy-ui/ComfyUI/models/checkpoints/
Here are my results:

The legs and the fingers are a little odd, but the image itself looks quite nice. | Closeup of Model: DreamShaper_8

Next, I wanted to see the response to a different sort of prompt:
Prompt: “analog film photo Three Kittens in teacups in space, riding a spaceship. In the style of buck rogers comics. . faded film, desaturated, 35mm photo, grainy, vignette, vintage, Kodachrome, Lomography, stained, highly detailed, found footage”

Cats in Space Teacups | Closeup of Model: DreamShaper_8

And another —
Prompt: “workmen at a job site rebuilding a road”

Road Work | Closeup of Model: DreamShaper_8

And a last one —
Prompt: “three beautiful kittens playing with string”

As you can see, the model does some interpretation!

Optional: Install ComfyUI Manager

This manager enables easy downloads of new models

Move into the directory: cd /Users/<path/to>/comfy-ui/ComfyUI/custom_nodes
Git Clone the Managers repo: git clone https://github.com/ltdrdata/ComfyUI-Manager.git
Restart ComfyUI from the ComfyUI directory: python3.11 main.py

To dive deeper into ComfyUI, I recommend checking out this detailed video: