Skip to content
Tony

Type to search…

↑↓ to navigate
Esc

Deploying ComfyUI on macOS

A guide to installing and running ComfyUI on Apple Silicon Mac, including Python setup, PyTorch, dependencies, and troubleshooting.

Tech , Large Language Models 1 min read

ComfyUI is a powerful and modular Stable Diffusion graphical user interface (GUI) and backend tool. It provides a graph/nodes/flowchart-based interface that lets you design and execute complex Stable Diffusion workflows without writing code. Here are its features and capabilities:

  1. Node/graph/flowchart interface: Experiment with and create complex Stable Diffusion workflows without writing any code.
  2. Comprehensive support: ComfyUI supports SD1.x, SD2.x, SDXL, Stable Video Diffusion, and Stable Cascade.
  3. Asynchronous queue system: The optimized queue system only re-executes the parts of the workflow that have changed.
  4. Low VRAM support: With the --lowvram option, it can run even on GPUs with less than 3GB of VRAM (automatically enabled on low-VRAM GPUs).
  5. Offline operation: ComfyUI works entirely offline and does not download anything.
  6. Model support: Can load ckpt, safetensors, and diffusers models/checkpoints, as well as standalone VAE and CLIP models.
  7. Workflow save/load: You can save workflows as JSON files and load complete workflows (including seeds) from generated PNG files.

We can use ComfyUI to easily generate images.

Local environment:

  • 16-inch MacBook Pro (Apple Silicon M1 Pro)
  • 16GB unified memory
  • 512GB SSD
  • macOS Sonoma 14.4 (23E214)

macOS comes with Python 3 pre-installed. To avoid having to manually type python3 every time, let’s set up some aliases.

Terminal window
$ vim ~/.zshrc
// Add the following three lines
export PATH="/Users/tony/Library/Python/3.9/bin:$PATH"
alias python='python3'
alias pip='pip3'
$ source ~/.zshrc

Terminal window
$ pip install --pre torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/nightly/cpu

Repository URL: https://github.com/comfyanonymous/ComfyUI.git

Place models in the models/checkpoints directory.

Switch to the ComfyUI directory and install dependencies:

Terminal window
$ pip install -r requirements.txt

During installation, you may encounter the following error:

/Users/tony/Library/Python/3.9/lib/python/site-packages/urllib3/__init__.py:35:
NotOpenSSLWarning: urllib3 v2 only supports OpenSSL 1.1.1+,
currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'.
See: https://github.com/urllib3/urllib3/issues/3020

This indicates an OpenSSL compilation issue — you’ll need to downgrade urllib3:

Terminal window
$ openssl version
LibreSSL 3.3.6
$ pip install urllib3==1.26.6

Terminal window
$ python main.py --force-fp16
Total VRAM 16384 MB, total RAM 16384 MB
Forcing FP16.
Set vram state to: SHARED
Device: mps
VAE dtype: torch.float32
Using sub quadratic optimization for cross attention, if you have memory or speed issues try using: --use-split-cross-attention
Starting server


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

ComfyUI

  1. comfyanonymous/ComfyUI
  2. Accelerated PyTorch training on Mac
  3. ImportError: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the ‘ssl’ module is compiled with LibreSSL 2.8.3
  4. ComfyUI outputs Rothko-esque solid color images