How to Install ComfyUI Locally on Windows?

ComfyUI is a powerful and flexible node-based interface for running AI models locally, giving you full control over complex workflows. While widely used for image generation, its modular design supports a range of AI tasks, allowing you to build, customize, and experiment with pipelines directly on your machine. Running locally also means improved performance, privacy, and creative freedom without relying on the cloud.

Here's how to get ComfyUI up and running on Windows with NVIDIA GPU support:

Prerequisites for Installing ComfyUI

• A Windows PC with an NVIDIA GPU

• Basic familiarity with file paths and command prompts

ComfyUI Installation Steps

1. Download ComfyUI
   Visit the official ComfyUI GitHub releases page, then scroll down to the "Assets" section to find the available download file.

   Download the file:
   ComfyUI_windows_portable_nvidia.7z

2. Extract the Files
   Use a tool like 7-Zip to extract the archive to an easy-to-remember location, such as:
   C:\comfyui

3. Verify the Folder
   Your extracted folder should contain a file named run_nvidia_gpu.

   You'll use this file to launch the app later, but before double clicking on run_nvidia_gpu.bat, we'll first install the ComfyUI Node Manager.

Custom nodes significantly expand ComfyUI’s functionality, allowing you to add new features, models, and capabilities. The Node Manager is a powerful extension that simplifies this process — making it easy to install, remove, enable, or disable custom nodes directly from within ComfyUI. It also provides a central hub to access tools, updates, and community resources, helping streamline your workflow.

How to Install the Node Manager

1. Open the Custom Nodes Folder

   Navigate to: C:\comfyui\ComfyUI_windows_portable\ComfyUI\custom_nodes

   

2. Open a Terminal Window
   In the File Explorer address bar, type cmd and press Enter to open a terminal in that folder.

   

3. Clone the Node Manager Repository
   After typing cmd in the File Explorer address bar and pressing Enter, a terminal window will open directly in the current folder. Next, type the following command and press enter:

   ts
   1git clone https://github.com/Comfy-Org/ComfyUI-Manager.git

   When you press Enter, this command will clone the GitHub repository directly into the ComfyUI folder structure.

   Once the cloning process is complete, you can safely close the terminal window. With the Node Manager installed, you’re now ready to manage a wide variety of custom nodes — unlocking new tools and possibilities within ComfyUI.

4. Launch ComfyUI
   Go back to: C:\comfyui\ComfyUI_windows_portable and double-click run_nvidia_gpu to start the ComfyUI application.

   

• Go to the top-left corner of the ComfyUI interface.

• Click on “Workflow” and then select “Browse Templates.”

• In the left sidebar, click on “Flux.”

• Select “Flux Dev” to load the workflow into the interface.

• ⚠️ Don’t click “Run” yet!
  You'll get an error because the Flux Dev model file hasn’t been downloaded yet.

Let’s take care of that error—how about we download the model together?

Flux Dev is a specialized image generation model that produces high-quality visual outputs. Here’s how to load it in ComfyUI:

Steps to Load Flux Dev Model/Checkpoint

1. Download the Checkpoint
   Visit the Flux Dev Checkpoint page and download:
   flux1-dev-fp8.safetensors

2. Move to Checkpoints Folder
   Place the file here:
   C:\comfyui\ComfyUI_windows_portable\ComfyUI\models\checkpoints

   

3. Refresh Node Definitions

   • In ComfyUI, click Edit > Refresh Node Definitions

   • Or simply press R on your keyboard

Now that everything is set up, we're ready to create our first image using the Flux Dev model. Let's walk through the process together—ready to see it in action?

You’re all set! Let’s create your first AI-generated image using ComfyUI and the Flux Dev model.

How to Generate an Image

1. Select the Checkpoint Node
   In the ComfyUI interface, locate the load checkpoint node and select:
   flux1-dev-fp8.safetensors

2. Craft Your Prompt
   Write a descriptive prompt that guides the model.
   Example prompt:

   ts
   1cute anime girl, pink hair, fantasy outfit, magical background, soft lighting

3. Adjust the Output Dimensions (Width × Height)

1. Run the Workflow\

Click the Run button at the bottom of the interface and watch as the model processes your prompt and generates a unique image.

All examples below were generated locally using ComfyUI with the Flux Dev model (flux1-dev-fp8.safetensors) on a 12GB VRAM RTX 3060. Each prompt includes the render time, so you can get a sense of performance on similar hardware.

These examples showcase different styles you can achieve—feel free to use them as inspiration or modify the prompts to suit your vision.

Example Prompts & Outputs

Example 1: Anime

• Prompt: “Cute anime girl, pink hair, sakura background, soft lighting, detailed eyes”

• Render Time: ~90 seconds

Example 2: Realistic

• Prompt: "Ultra-photorealistic portrait of a Latina woman with long, dark brown hair and warm hazel eyes, lying sensually on a bed wearing elegant black lingerie. Soft studio lighting highlights her features, with a shallow depth of field creating a dreamy bokeh effect. She has natural makeup, subtle freckles across her cheeks and nose, warm skin tones with detailed texture, and a gentle, inviting expression."

• Render Time: ~99 seconds

Example 3: Environmental Shot

• Prompt: “Cinematic mountain landscape at sunset, dramatic lighting, ultra-detailed clouds, foggy atmosphere”

• Render Time: ~95 seconds

Example 4: Sci-Fi Concept Art

• Prompt: “Futuristic city skyline at night, glowing neon lights, futuristic cars, cyberpunk aesthetic”

• Render Time: ~89 seconds

Example 4: Line art

• Prompt: “Anime-style black and white line art of a sensual, female character. Soft expression with slightly parted lips, seductive gaze, and stylized long hair flowing around her shoulders. Wearing a revealing, elegant outfit with lace or straps. Clean, crisp linework with varied line thickness and expressive detailing in the eyes, hair, and fabric folds. Blank or minimal background."

• Render Time: ~97 seconds

Note: All of your generated images will be saved automatically in the following folder: "C:\comfyui\ComfyUI_windows_portable\ComfyUI\output"

Running ComfyUI locally gives you a powerful and flexible platform for exploring AI-generated art right from your own machine. In this guide, you successfully installed ComfyUI on Windows, added the Node Manager to unlock advanced features, loaded the Flux Dev model, and created your very first image.

Looking to run Flux Dev on a laptop or low-VRAM machine? We’ve published a follow-up guide that shows you how to run the model in GGUF format using CPU-friendly tools and quantized models:

👉 How to Run Flux Dev GGUF Models in ComfyUI (Low VRAM Guide)

Now that you're set up, there’s a lot more to discover — try experimenting with different models and checkpoints, exploring more complex workflows using custom nodes, or engaging with the vibrant ComfyUI community on GitHub and Discord.