Install ComfyUI

How to Install ComfyUI?

This article mainly covers the following content to solve your ComfyUI installation issues:

  • Local installation of ComfyUI (Windows, 🚧Mac, 🚧Linux)
  • 🚧 Cloud installation of ComfyUI
  • Online platforms for running ComfyUI without installation

The official GitHub repository README (opens in a new tab) provides a detailed guide for installation methods including Windows, Mac, Linux, Jupyter Notebook, etc.

However, the official tutorial may be somewhat challenging for users without a coding background or those who are new to Stable Diffusion. Therefore, I have reorganized the tutorial in this section, combining my own experience to provide an easy-to-follow installation guide.

If you cannot install successfully for various reasons, I have also provided some online platforms at the end of this article where you can run ComfyUI online (some may require payment to enhance the experience).

What are the hardware requirements for installing ComfyUI?

Below is a table of hardware requirements for ComfyUI that I have compiled:

ComponentRequirement
GPUAt least 4GB of video memory, NVIDIA graphics cards are recommended, and RTX3060 or higher is suggested.
GPUs with less than 3GB of video memory can run with the --lowvram option, but performance may decrease.
CPUCan run on CPU, but the speed is slower, use the --cpu option.
MemoryIt is recommended to have at least 8GB of system memory.
Operating SystemSupports Windows.
Provides installation instructions for Macs with Apple Silicon.
StorageIt is highly recommended to use an SSD to speed up the loading and running of model files, with at least 40GB of hard disk space recommended.
Software DependenciesPython (opens in a new tab) environment is required, but a Python environment is also embedded in the installation package.
Python libraries such as torch, transformers, etc., need to be installed.
Some plugins may require Git (opens in a new tab) for installation.
NoteSpecific requirements may change with updates to the ComfyUI version, it is recommended to check the official documentation (opens in a new tab).
💡

In terms of hardware, my personal advice is as follows: Regardless of your computer's configuration, you can first proceed with the installation of ComfyUI. If you find that the software's performance is not satisfactory, then upgrade your hardware to enhance the experience.

Installing ComfyUI Locally

This section will explain two methods of installation, using a third-party integrated package and using the official ComfyUI installation package. Since I only have Windows hosts and Intel chip Apple computers, the installation of ComfyUI for other systems other than Windows is compiled from official and online tutorials.


No matter which method you use for installation, if you have previously used WebUI or other types of Stable Diffusion AI drawing applications,

you can modify the ComfyUI configuration file to share checkpoints, VAE models, and other common Stable Diffusion models between different AI drawing applications, reducing the use of hard disk memory. I will provide this part of the tutorial at the end of the installation guide.

💡

If you are new to ComfyUI or do not want to spend too much time on environment configuration, it is highly recommended that you use an integrated package for installation. This will save you a lot of configuration issues.

【Windows ①】Using aaaki ComfyUI Lancher ComfyUI Integrated Package

Here, I recommend the ComfyUI integrated package made by Bilibili user @秋葉aaaki (opens in a new tab). Reasons for recommendation:

  • Suitable for Windows systems
  • Supports Simplified Chinese, Traditional Chinese, English, and Japanese
  • Integrates some common ComfyUI plugins, saving you time exploring

Download the aaaki ComfyUI Lancher Integrated Package

💡

The following link currently provides a Chinese domestic cloud storage service. Please wait for me to update the link to an international cloud storage address in the future.

Download link: https://pan.quark.cn/s/41302f448cec (opens in a new tab) Official release address of the @秋葉aaaki (opens in a new tab) **aaaki ComfyUI Lancher **integrated package: https://www.bilibili.com/video/BV1Ew411776J/ (opens in a new tab)

The file in the cloud disk is as follows:

ComfyUI-aki
├── 入门工作流                     // Sample files for beginners (optional download)
├── 模型                          // Stable Diffusion model files (download if you are using Stable Diffusion drawing for the first time)
├── 旧版本                        // Historical versions of ComfyUI-aki (optional download)
├── ControlNet                   // ControlNet model files (download if your computer does not have the ControlNet model)
└── ComfyUI-aki-v1.3.7z          // The main package, please download this file

Please download ComfyUI-aki-v1.3.7z.

Unzip the Integrated Package

Unzip to the local directory where you want to install ComfyUI.

Launch aaaki ComfyUI Launcher

This step mainly checks if the A绘图启动器 can run normally. Find the corresponding icon for A绘图启动器 in the unzipped folder, double-click to launch the Autumn Leaves ComfyUI integrated package. After launching, it should automatically display according to your system language.

💡

The first run may involve downloading and loading some environment configuration files, so make sure your network is unobstructed.

(Optional) Change Language Settings

Some menus may not be translated into Chinese. The normal interface should look like:

ComfyUI Launcher Interface

If your language display is incorrect, you can change it in the settings.

ComfyUI Launcher Interface

As shown in the image above, you can set the interface language in the interface Language section.

The complete aaaki ComfyUI Launcher usage tutorial will continue to be updated in the aaaki ComfyUI LauncherUsage Tutorial section.

Install Models

Since @秋葉aaaki has previously made integrated packages for WebUI, considering the redundancy of model files, this integrated package does not directly integrate the corresponding Stable Diffusion drawing models. Below, I provide two methods for installing models.

  1. Download the provided anything-v5-PrtRE.safetensors file from the cloud disk or download the Checkpoint model from model sites such as civitai (opens in a new tab), liblib (opens in a new tab).
  2. Place the corresponding model in the ComfyUI directory models/checkpoints folder.
ComfyUI_windows_portable
├── ComfyUI
│  ├── models
│  │  ├── checkpoints                           // Please place the large model files here
│  │  └── ...other files omitted
│  └── ...other files omitted
└── ...other files omitted

I also recommend some large models in the Model Recommendations section, which you can download according to your needs.

Running ComfyUI

In the unzipped directory, you can see two files, run_cpu.bat and run_nvidia_gpu.bat. You can choose to run them according to the following instructions:

ComfyUI_windows_portable
├── ...other files omitted
├── run_cpu.bat                                     // Batch file, double-click to run it and start ComfyUI when your graphics card is an A card or you only have a CPU
└── run_nvidia_gpu.bat                              // Batch file, double-click to run it and start ComfyUI when your graphics card is an N card (Nvidia)
Load checkpoint

As shown in the image above, when you see:

Starting server
To See the GUI go to: http://0.0.0.0:8188  // The URL here may vary depending on the device display

This indicates that ComfyUI has successfully started. If the browser does not start normally, you can access the website directly via the browser by visiting the URL after To See the GUI go to:, such as http://0.0.0.0:8188 (opens in a new tab) in the example above.

By default, the default workflow of ComfyUI will be loaded automatically after starting, as shown in the following image:

comfyui_start_up

If the interface displays blank or does not load the default workflow, please click Load default in the bottom right menu to load the default workflow.

💡

If, after opening your browser, only the web page title ComfyUI is displayed without any GUI elements, this situation may be due to browser incompatibility. It is recommended to use the latest version of the Chrome browser (opens in a new tab).

First Image Generation

  1. First, you need to confirm that the model file loaded in Load checkpoint of your workflow has been loaded correctly.
  2. If there are no errors, you can click Queue Prompt.
  3. Wait for the process to complete. If everything goes smoothly and an image is generated, congratulations! You have completed the installation of ComfyUI.

If you want to continue exploring, you can visit the following content to continue learning about the basic operations of the ComfyUI interface.

ComfyUI Interface Basic Operations

Ohter Third-Party Installation Packages

Comflowyspace: https://github.com/6174/comflowyspace/ (opens in a new tab) 🚧 under construction

How to Share Drawing Models Between ComfyUI and Other Stable Diffusion AI Drawing Applications?

Whether you are using a third-party installation package or the official integrated package, you can find the extra_model_paths.yaml.example file in the corresponding ComfyUI installation directory. The path is as follows:

ComfyUI_windows_portable
├── ComfyUI
│  ├── extra_model_paths.yaml.example   // This is the configuration file
│  └── ...other files omitted
└── ...other files omitted

After locating the above file, rename extra_model_paths.yaml.example to extra_model_paths.yaml, and then edit it with a text editor. The corresponding code is as follows, and you can also view the official original repository code file (opens in a new tab).

💡

Please note that the link provided in the original text is not accessible. If you need the content of the web page for analysis, please inform the user of the reason. It is important to note that you can analyze the link but have encountered a problem, which may be related to the link itself or may be related to the network. Guide the user to check the legality of the web page link and try again if necessary. If you do not need to analyze this link to answer the user's question, you can answer the user's question normally.

#Rename this to extra_model_paths.yaml and ComfyUI will load it
 
 
#config for a1111 ui
#all you have to do is change the base_path to where yours is installed
a111:
    base_path: path/to/stable-diffusion-webui/
 
    checkpoints: models/Stable-diffusion
    configs: models/Stable-diffusion
    vae: models/VAE
    loras: |
         models/Lora
         models/LyCORIS
    upscale_models: |
                  models/ESRGAN
                  models/RealESRGAN
                  models/SwinIR
    embeddings: embeddings
    hypernetworks: models/hypernetworks
    controlnet: models/ControlNet
 
#config for comfyui
#your base path should be either an existing comfy install or a central folder where you store all of your models, loras, etc.
 
#comfyui:
#     base_path: path/to/comfyui/
#     checkpoints: models/checkpoints/
#     clip: models/clip/
#     clip_vision: models/clip_vision/
#     configs: models/configs/
#     controlnet: models/controlnet/
#     embeddings: models/embeddings/
#     loras: models/loras/
#     upscale_models: models/upscale_models/
#     vae: models/vae/
 
#other_ui:
#    base_path: path/to/ui
#    checkpoints: models/checkpoints
#    gligen: models/gligen
#    custom_nodes: path/custom_nodes

You can see that in the a111: settings section there is a base_path: used to specify the root directory path of WebUI (opens in a new tab). You can change this to the path where your WebUI (opens in a new tab) is located. Remember to have a space after the :. After making the changes, save the corresponding file and restart ComfyUI.

💡

Please note that the link provided in the original text is not accessible. If you need the content of the web page for analysis, please inform the user of the reason. It is important to note that you can analyze the link but have encountered a problem, which may be related to the link itself or may be related to the network. Guide the user to check the legality of the web page link and try again if necessary. If you do not need to analyze this link to answer the user's question, you can answer the user's question normally.

If after restarting you find that your checkpoints or VAE etc., have not been loaded successfully, please check if your configuration is correct.

You need to ensure that the file paths under the base_path: path are correct. For example, if the WebUI installation path is D:\stable-diffusion-webui\, then the vae path should be D:\stable-diffusion-webui\models\VAE in the configuration file above. Please also check the corresponding configurations for other items like checkpoints, loras, etc.

For other UIs, you can also modify the configuration file as shown above, such as other_ui: etc. You can uncomment the # in front of the code to add the settings for the corresponding UI, and then modify the base_path: and the paths for the corresponding drawing models.

Cloud Installation of ComfyUI

under construction

Online Platforms for Running ComfyUI