Installing ComfyUI and Generating Your First Image

Comfy Desktop is the official desktop application for ComfyUI. It installs ComfyUI together with the required runtime environment, so you do not need to configure Python, CUDA, or pip separately. You can also keep multiple installations on the same computer. Each installation remains independent, making it suitable for trying different versions or configurations.

Before You Start

First, confirm that your computer meets the following requirements:

• Windows 10 or later, with x64 or ARM64 support. A dedicated NVIDIA or AMD graphics card is recommended, but installation is still possible without one.
• macOS 13 (Ventura) or later, with an Apple silicon chip required (M1 or later).
• Linux. Installation packages are available as .deb and AppImage files. Ubuntu 22.04 or later on x64 is recommended.

Each ComfyUI installation requires approximately 4.85 GB of disk space. Prepare at least 8 GB of memory; 16 GB provides a better experience. Image models downloaded later will require additional disk space.

macOS

Open the dmg

• Drag it into Applications
• For the first launch, go to Privacy & Security
• Click Open Anyway

Windows

Run the exe

• Windows 10 or later
• x64 or ARM64
• Create the installation from the welcome page

Linux

deb or AppImage

• Ubuntu 22.04+ recommended
• x64
• If launch fails, review the sandbox instructions

Install and Launch for the First Time

Go to the Comfy Desktop download page to get the installation package. The page automatically detects the system you are using.

macOS

1. Open the downloaded .dmg.
2. Drag Comfy Desktop into the Applications folder.
3. Open Comfy Desktop from Launchpad or Spotlight Search.

When you launch it for the first time, macOS may display a security warning. Open Privacy & Security in System Settings and click Open Anyway. This button allows the system to launch the application that was just blocked.

Windows

Run the downloaded .exe (NSIS installer), follow the on-screen instructions to complete the installation, and then open Comfy Desktop from the Start menu or desktop shortcut.

The welcome screen appears on first launch. You can create your first ComfyUI installation there.

Linux

Linux uses a general download link. Debian or Ubuntu users should open the download directory containing the installation package and run:

sudo apt install ./*.deb

When using the AppImage, run:

chmod +x ./*.AppImage
./*.AppImage

If the application fails to launch, your Linux distribution may be blocking the Electron sandbox. Ubuntu 24.04 and later restrict unprivileged user namespaces through AppArmor. Add --no-sandbox and try again only after confirming that the sandbox is causing the launch failure. This option disables the Electron sandbox for each process and is not recommended in other situations.

After opening Comfy Desktop, the Chooser page lists existing installation instances. Click New Instance to create a new, independent ComfyUI installation, then click its instance card to launch it.

The desktop version tracks the stable release by default. If you want to follow every latest commit on GitHub, use the portable version, install manually with git, or change the instance’s update channel to Latest on GitHub.

Get Connection Details from OmniProxy

There is no universal address, key, or model ID to enter here. Installing Comfy Desktop and completing the basic text-to-image workflow do not require OmniProxy.

Do not enter another tool’s Base URL into nodes from an unclear source. Wait until explicit instructions for a general integration method are available before configuring it separately.

Enter the Tool Configuration

There is no OmniProxy configuration form here. To generate your first image, you must first prepare a workflow and an image model.

A workflow can be understood as a sequence of image-generation steps. Nodes are the operational modules within it that load models, accept text, or save images. After ComfyUI starts, it usually loads the default text-to-image workflow automatically.

To open another workflow, select Workflows → Open from the menu. This option loads an image containing workflow information or a JSON workflow file.

Most installations do not include a base model by default. After loading the workflow, if v1-5-pruned-emaonly-fp16.safetensors is not on your computer, the interface will report that the model is missing. Open the warning to view the missing model and its download link.

To view the model directory used by the current installation, click the capsule menu at the top and open Storage. This section shows where models and other files are actually stored. The top capsule menu is available only in the desktop application. Click the folder icon next to a directory to open it directly in your system’s file manager.

There are two ways to download the model:

• Click Download in the interface. This button starts downloading the missing model. The desktop version saves the model to ComfyUI/models/checkpoints in the current installation. You can view the download progress in the model panel in the sidebar. Automatic downloads usually come from Hugging Face, so your network must be able to access Hugging Face.
• If there is no download progress for a long time, copy the download link from the missing model’s details, download the file manually, and save it to the same checkpoints directory. Then refresh or restart ComfyUI.

After placing the model in the directory, press R on your keyboard so the interface reloads the model lists in the nodes.

App

Comfy Desktop

• It manages the environment installation
• Use the welcome page or create a new instance
• Click the card to launch

Workflow

Text to image

• Usually loads automatically
• You can also open it from the menu
• Missing models trigger a warning

Model

checkpoints

• Download from the link in the warning
• The desktop version can save it automatically
• Refresh the list with R

Verify a Real Task

After preparing the model, follow these steps to try generating your first image:

1. In the Load Checkpoint node, select v1-5-pruned-emaonly-fp16.safetensors. This node loads the model used for the current image-generation task.
2. Click Run, or press Ctrl + Enter. Run starts executing the current image-generation task.
3. The generated result appears in the Save Image node. This node previews and saves the final image. You can right-click the image to save it locally.

If no model appears in Load Checkpoint, or if the drop-down list displays null, check the model directory first, then refresh or restart ComfyUI.

As of 2026-10-09, Comfy Desktop 1.1.6 obtained from the official dmg was confirmed to launch on macOS. Because an installation still needed to be created in the application and a checkpoint model downloaded, the first-image workflow was not completed. Therefore, this guide does not claim that an image was successfully generated, and OmniProxy was not integrated with ComfyUI.

Model

The dropdown shows a name

• It is not null
• Select a checkpoint
• Refresh with R

Generate

Click Run

• Or press Ctrl+Enter
• Wait for the nodes to finish
• The image is in Save Image

Scope

Do not connect the gateway yet

• There is no universal configuration field
• Do not install nodes from unknown sources
• Successful image generation is enough

Common Errors

• macOS cannot open the application: Go to System Settings → Privacy & Security and click Open Anyway. The installation package supports only Apple silicon and macOS 13 or later.
• The application exits immediately on Linux: First confirm that you are using the .deb or AppImage. Add --no-sandbox only after confirming that the sandbox is causing the launch failure.
• A missing model warning appears: Download the checkpoint model through the link in the warning, save it to the current installation’s models/checkpoints directory, and then press R to refresh.
• The model drop-down list displays null: Use Storage to check the model directory, confirm that the file is in the correct location, and then refresh or restart.
• You want to enter OmniProxy details here: There is currently no general integration field. Do not install additional nodes that are not explicitly documented as a general integration method.

Official References

For an installation overview, see Comfy Desktop. For platform-specific installation steps, see macOS, Windows, and Linux. For instructions on using the desktop interface, see Using Comfy Desktop. For the first-image generation steps, see Getting Started with AI Image Generation.

Replace every placeholder with values from your own OmniProxy configuration. Never publish a real key.

A saved result or HTTP 200 alone does not prove full Agent compatibility. Check the source and model in OmniProxy's call record.

Official sources · ComfyUI ↗