Fix ComfyUI Install On MacOS ARM64

Dec 21, 2025 by Alex Johnson 35 views

Troubleshooting ComfyUI Installation on macOS ARM64

The Challenge: ComfyUI Install Fails on macOS ARM64

It can be incredibly frustrating when new software, especially something as exciting as ComfyUI, doesn't install smoothly on your macOS machine. Many users have encountered a roadblock specifically when trying to set up ComfyUI on ARM64 architecture, which is common in newer Macs. This article dives deep into the common causes behind this installation failure and provides clear, actionable steps to get you up and running. We understand the importance of a seamless setup, especially for users diving into the world of AI image generation, and aim to demystify the process. The error messages often point to dependency issues, particularly with PyTorch, and sometimes involve the package installer, uv, not finding compatible wheels for your specific system. We'll break down what these errors mean and how to overcome them, ensuring you can harness the full potential of ComfyUI without the installation hassle. This guide is tailored for those who might be new to command-line installations or troubleshooting dependency conflicts, offering a friendly and comprehensive approach.

Understanding the Core Issue: Dependency Hell

At the heart of the ComfyUI installation failure on macOS ARM64 often lies a classic software development headache: dependency hell. This term refers to the complex and sometimes intractable problems that arise when a piece of software requires specific versions of other software (its dependencies), and those dependencies, in turn, have their own requirements. It's like a cascading effect where a conflict at one level can bring down the entire installation. In the case of ComfyUI on ARM64 Macs, the primary culprit is often PyTorch, a fundamental library for deep learning. The log snippet you provided clearly indicates a problem: torch==2.8.0+cu129 has no wheels with a matching platform tag (e.g., macosx_26_0_arm64). Let's unpack this. PyTorch, by default, often tries to install versions optimized for specific hardware and operating systems. The +cu129 part usually signifies CUDA support, which is primarily for NVIDIA GPUs and not relevant or available on Apple's ARM-based silicon (like M1, M2, M3 chips). Furthermore, the macosx_26_0_arm64 tag is the specific identifier for a Python package built to run on macOS with an ARM64 processor. When the installer, in this case, uv (a fast Python package installer), cannot find a compatible pre-compiled version (a "wheel") for your architecture, it throws an error. This means the version of PyTorch that ComfyUI is trying to install simply isn't available in a format that your Mac can directly use. The system requirements and dependencies aren't met because the correct version of the dependency isn't accessible. It's not that PyTorch can't run on ARM64 Macs, but rather that the specific version being requested or the way it's being requested is incompatible. We need to find a way to tell the installer to fetch a version of PyTorch that is compatible, or to install it in a way that works.

Step-by-Step Solution: Installing the Right PyTorch

Now that we understand the root cause, let's get down to fixing the ComfyUI installation on macOS ARM64. The key is to manually install a compatible version of PyTorch before or during the ComfyUI setup. We need to bypass the automatic installation of a version that doesn't support your hardware. Here’s how you can typically approach this:

1. Prepare Your Environment

Before anything else, ensure you have a clean environment. If you've attempted installations before, it's often best to remove any partial installations to avoid further conflicts. This might involve deleting the ComfyUI folder and any related virtual environment folders (like .venv).

2. Install PyTorch for macOS ARM64

This is the most crucial step. You'll need to use your terminal. The official PyTorch website is the best resource for finding the correct installation commands. Navigate to pytorch.org and use their installation guide. Select the following options:

PyTorch Build: Stable (or the version recommended by ComfyUI if specified)
Your OS: macOS
Package: Pip
Compute Platform: None (or CPU, as CUDA is not applicable here)

This should generate a command similar to:

pip3 install torch torchvision torchaudio

However, sometimes the default pip or uv might still try to grab an incompatible version. A more robust way, especially if the above fails, is to explicitly specify a version that is known to work or to use a command that ensures compatibility. Often, specifying just torch without any CUDA variants is enough. If you encounter issues, you might need to find a specific wheel file, but let's try the simpler approach first.

For example, if the automatic uv installation failed, try this within your project directory (where you intend to install ComfyUI):

# Ensure you are in your project directory (e.g., /Users/sam/Dione/apps/ComfyUI/)
# If you have a .venv, activate it first:
# source .venv/bin/activate

# Install a compatible PyTorch version
pip install torch torchvision torchaudio

Important Note: If pip install torch torchvision torchaudio still fetches an incompatible version, you might need to consult the PyTorch website for specific commands for Apple Silicon, or even try installing an older, known-stable version. Sometimes, removing torchvision and torchaudio temporarily, installing just torch, and then adding the others back can resolve conflicts.

3. Re-run the ComfyUI Installer

Once you've successfully installed a compatible PyTorch, you can re-run the Dione installer script for ComfyUI. The installer should now recognize that the core dependency is met and proceed to the next steps, which typically involve cloning the ComfyUI repository and installing its specific Python dependencies.

If the installer script from dioneapp/official-scripts is still failing, you might need to manually clone ComfyUI and set up its environment. After installing PyTorch as described above, you would navigate to your desired installation directory and run:

# Clone ComfyUI (if not already done by installer)
git clone https://github.com/comfyanonymous/ComfyUI.git
cd ComfyUI

# Create and activate a virtual environment (if not using one already)
python -m venv .venv
source .venv/bin/activate

# Install ComfyUI's requirements (after PyTorch is installed!)
pip install -r requirements.txt

This manual approach bypasses the installer script entirely and gives you more control over the installation process, especially when dealing with tricky dependencies like PyTorch on specific architectures.

Advanced Troubleshooting and Common Pitfalls

Sometimes, even with the correct PyTorch version installed, the ComfyUI installation on macOS ARM64 can hit snags. Let's explore some advanced troubleshooting steps and common pitfalls to watch out for. Understanding these can save you a lot of time and frustration.

Virtual Environment Issues

Virtual environments (venv or conda) are crucial for isolating project dependencies. However, they can sometimes cause confusion. Ensure that your virtual environment is correctly activated before you attempt to install PyTorch or run the ComfyUI installer. The log shows uv is being used, which is a newer, faster package installer. While efficient, uv might have its own quirks or specific ways it interacts with virtual environments. If you're using uv, make sure it's up-to-date (uv --version to check, and uv --upgrade if possible). If you suspect issues with uv, you might try switching back to standard pip for the PyTorch installation step.

To ensure you're using the correct environment, always check your terminal prompt. It should indicate the active virtual environment (e.g., (.venv) your-mac:your-project sam$ ). If it doesn't, activate it using source .venv/bin/activate (or the appropriate command for your environment manager).

Python Version Compatibility

While the log indicates Python 3.10.18 is being used, compatibility issues can sometimes arise if ComfyUI or its dependencies have specific Python version requirements that aren't perfectly met. ComfyUI generally works well with Python 3.9+, but extremely new or old versions can sometimes cause problems. Ensure the Python version you're using to create the virtual environment is stable and widely supported. If you have multiple Python versions installed, be explicit about which one you are using when creating the virtual environment (e.g., python3.10 -m venv .venv).

Network and Firewall Issues

Less common, but still possible, are network-related problems. Firewalls or proxy settings can sometimes block access to package repositories like PyPI (Python Package Index), preventing the download of necessary libraries. If you're on a corporate network or have strict security settings, check if they might be interfering. Trying to install from a different network (like your home Wi-Fi if you're on a work network) can help diagnose this.

Outdated Installer Scripts

If you're using a script provided by a third party (like the dioneapp/official-scripts mentioned), it's possible the script itself is outdated and points to or attempts to install older versions of dependencies that are no longer compatible. In such cases, manually performing the installation steps as outlined in the previous section (installing PyTorch, cloning ComfyUI, running pip install -r requirements.txt) is often the most reliable workaround. You can check the dioneapp/official-scripts repository on GitHub for any recent updates or issues related to ComfyUI.

Hardware-Specific Configurations

For macOS ARM64, the lack of CUDA support is the primary factor. However, other hardware-specific nuances can emerge. For instance, ensuring your macOS is up-to-date is generally a good practice, as OS updates can sometimes include updated libraries or drivers that improve compatibility. Always refer to the official ComfyUI documentation or community forums for any specific hardware recommendations or known issues related to Apple Silicon.

By systematically addressing these potential issues, you significantly increase your chances of a successful ComfyUI installation.

Conclusion: Enjoying ComfyUI on Your Mac

Successfully installing ComfyUI on macOS ARM64 might seem like a technical hurdle, but by understanding the underlying dependency issues, particularly with PyTorch compatibility, you can navigate the process effectively. The key takeaway is ensuring that the version of PyTorch you install is specifically built for your Apple Silicon architecture and doesn't rely on CUDA. By following the step-by-step guide, focusing on manually installing the correct PyTorch version, and being mindful of virtual environment configurations, you should be able to overcome the common errors that prevent installation. Remember to consult official documentation and community resources when you encounter specific problems. With these steps, you'll be well on your way to exploring the creative possibilities of ComfyUI. For more in-depth information on managing Python environments and troubleshooting package installations, you can refer to resources like The Hitchhiker's Guide to Python or the official PyTorch installation guide.