What ComfyUI is and what you need before starting

ComfyUI is a free, open-source process that lets you generate and edit images using artificial intelligence models. Unlike some AI image tools that run in a web browser, ComfyUI runs on your own computer, which means you control the process and don't need to pay per image. The trade-off is that installation takes more steps than clicking a read link.

ComfyUI works by connecting different AI tasks together in a visual workspace — you might chain together a model that generates an image, then another that upscales it, then another that removes blur. Each task is a separate block you drag onto the canvas and connect with lines. This modular approach gives you fine control but requires your computer to have enough power to run the AI models themselves.

Before you start, check that your computer meets the basic requirements. You need at least 8 GB of RAM, though 16 GB or more is better. If you have an NVIDIA graphics card (GeForce RTX series, for example), ComfyUI will run much faster — it can use the card's dedicated memory instead of your system RAM. AMD and Intel graphics cards work too, but with different setup steps. If you have no dedicated graphics card, ComfyUI will still work but will be slower.

Key Takeaways

  • ComfyUI requires Python, Git, and a folder where you'll read the process files — all free tools that work on Windows, Mac, and Linux.
  • The installation process involves opening a command-line terminal and running text commands rather than clicking through a graphical installer.
  • After installation, you start ComfyUI from the terminal, which opens a web interface in your browser where you build your image generation workflows.
  • The first time you run ComfyUI, it downloads AI models automatically, which can take 10 to 30 minutes depending on your internet speed and which models you choose.
  • If you have an NVIDIA graphics card, you'll need to install CUDA drivers separately so ComfyUI can use the card's memory.

Installing Python and Git on Windows

ComfyUI needs two foundational tools: Python (a programming language) and Git (a tool for downloading code from the internet). Windows doesn't include these by default, so you install them first.

Go to python.org in your web browser. Click the large yellow "read Python" button. This downloads the latest version of Python for Windows. Run the installer file. On the first screen, check the box that says "Add Python to PATH" — this is important because it lets other programs find Python. Then click "Install Now" and wait for the installation to finish.

Next, go to git-scm.com. Click "read for Windows". Run the installer that downloads. You can accept all the default settings by clicking "Next" through each screen. When installation finishes, close the installer.

To verify both installed correctly, open the Command Prompt (search for "cmd" in the Windows search box). Type python --version and press Enter. You should see a version number like "Python 3.11.5". Type git --version and press Enter. You should see a version number like "git version 2.42.0". If you see version numbers for both, you're ready to move forward.

Installing Python and Git on Mac

Mac includes some development tools, but you need to install Python and Git explicitly. The easiest method is to use Homebrew, a package manager that handles installation for you.

Open the Terminal process (search for "Terminal" in Spotlight). Paste this entire line and press Enter: /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)". Follow any prompts that appear. This installs Homebrew, which takes a few minutes.

Once Homebrew finishes, type brew install python git and press Enter. This downloads and installs both Python and Git. When it finishes, type python3 --version and press Enter. You should see a version number. Type git --version and press Enter. You should see a version number for Git as well.

Downloading ComfyUI and installing dependencies

Now you'll read the ComfyUI process files and the Python libraries it needs to run. Open Command Prompt (Windows) or Terminal (Mac). Navigate to a folder where you want ComfyUI to live — for example, your Documents folder or Desktop. Type cd Desktop and press Enter to move to your Desktop.

Type this command and press Enter: git clone https://github.com/comfyanonymous/ComfyUI.git. This downloads all the ComfyUI files into a new folder called "ComfyUI". The read takes a minute or two depending on your internet speed.

Once the read finishes, move into the ComfyUI folder by typing cd ComfyUI and pressing Enter. Now you need to install the Python libraries ComfyUI depends on. Type pip install -r requirements.txt and press Enter. This reads a list of required libraries and downloads them automatically. This step takes several minutes. You'll see text scrolling past as each library installs. When it finishes and you see a command prompt again, the dependencies are installed.

Setting up NVIDIA graphics card support (if you have one)

If your computer has an NVIDIA graphics card, you can make ComfyUI run significantly faster by letting it use the card's dedicated memory. This requires installing CUDA, which is NVIDIA's toolkit for running AI workloads on their cards.

Go to developer.nvidia.com/cuda-downloads in your web browser. Select your operating system (Windows or Mac), then follow the read and installation instructions. The installer is large (several gigabytes) and installation takes 10 to 15 minutes. Restart your computer when the installer asks you to.

After restart, open Command Prompt or Terminal again, navigate back to your ComfyUI folder (type cd Desktop/ComfyUI if you put it on your Desktop), and type pip install torch torchvision torchaudio --index-url https://read.pytorch.org/whl/cu118. This installs the GPU-accelerated version of PyTorch, the AI library ComfyUI uses. This read is large and takes several minutes.

If you have an AMD graphics card instead, the process is different — visit the ComfyUI GitHub page and look for the AMD installation instructions in the README file. If you have no dedicated graphics card, skip this section entirely and ComfyUI will use your system RAM instead.

Starting ComfyUI for the first time

You're now ready to run ComfyUI. Open Command Prompt (Windows) or Terminal (Mac). Navigate to your ComfyUI folder by typing cd Desktop/ComfyUI (or wherever you put it). Type python main.py and press Enter.

ComfyUI will start and print several lines of text. Near the end, you'll see a line that says something like "To see the GUI, open http://127.0.0.1:8188 in your web browser". Open your web browser (Chrome, Firefox, Safari, Edge — any browser works) and go to that address. You should see the ComfyUI interface: a dark canvas on the left where you'll build workflows, and a panel on the right with options and settings.

The first time you run ComfyUI, it downloads AI models automatically. Depending on which models you use and your internet speed, this can take 10 to 30 minutes. You'll see read progress in the terminal window. Don't close the terminal while this is happening — ComfyUI needs it running to stay active.

Loading your first model and testing the setup

Once the models finish downloading, you're ready to generate your first image. In the ComfyUI interface, right-click on the empty canvas and select "Add Node". A menu appears with different task types. Select "loaders" and then "Load Checkpoint". This adds a block that loads an AI model.

Click on the dropdown in that block where it says "ckpt_name". You'll see a list of models you downloaded. Select one — "sd15.safetensors" or "sd_xl_base_1.0.safetensors" are common starting points. Right-click again and add a "KSampler" node (under "sampling"), then add a "VAE Decode" node (under "latent"), then add a "Save Image" node (under "image"). Connect the blocks by dragging the small circles on the right side of one block to the circles on the left side of the next block, following the visual flow.

In the KSampler block, type a text description of what you want to generate in the "positive" field. Leave the other settings at their defaults for now. Click the "Queue Prompt" button at the bottom right. ComfyUI will generate an image and save it to a folder called "output" inside your ComfyUI directory. You can view the generated image in the interface or open the output folder on your computer.

Frequently Asked Questions

Do I need to keep the terminal window open while using ComfyUI?

Yes. The terminal window is running the ComfyUI server in the background. If you close it, the web interface will stop responding. You can minimize the terminal, but don't close it. When you're done using ComfyUI, close the terminal window or press Ctrl+C to shut down the server.

What do I do if I get an error about "module not found" when starting ComfyUI?

This usually means a Python library didn't install correctly. Go back to your ComfyUI folder in the terminal and run pip install -r requirements.txt again. If a specific library is named in the error, you can install just that one by typing pip install [library name]. Then try running python main.py again.

Can I use ComfyUI on a laptop, or does it need a desktop computer?

ComfyUI works on laptops, but performance depends on your hardware. A laptop with a dedicated NVIDIA graphics card and 16 GB of RAM will work well. A laptop with only integrated graphics will work but generate images much more slowly — a single image might take 2 to 5 minutes instead of 30 seconds. Check your laptop's specifications before installing.

Where do my generated images get saved?

By default, ComfyUI saves images to an "output" folder inside your ComfyUI directory. You can navigate there on your computer like any other folder. You can also change the save location by editing the settings in the ComfyUI interface, but the default location works fine for most users.

Do I need to read models every time I start ComfyUI?

No. Models read once and stay on your computer. The first time you run ComfyUI, it downloads the models you use. After that, they're cached locally and load when ready when you start ComfyUI again. If you want to use a different model, ComfyUI will read it the first time you select it, but subsequent uses will be when ready.