ComfyUI Tutorial: Installation, Setup, and First Image Generation

컴피유는 단계별 가이드를 통해 안정적인 확산 이미지 생성을 간소화합니다.

### ComfyUI 완벽 가이드 1부: 설치 및 체크포인트 설정
* 👨‍💻 본 강의는 Stable Diffusion을 이용한 이미지 생성을 쉽게 배우고자 하는 사용자를 위한 ComfyUI 설치 및 사용법 가이드입니다.
* 💡 ComfyUI의 복잡한 노드 구성에 어려움을 느끼는 사용자들을 위해 기본부터 단계별로 설명합니다.
* ⚙️ Git 프로그램 설치, ComfyUI 설치, 체크포인트 모델 설정 순서로 진행되며,  **각 단계별 상세한 설명과 이미지를 포함**합니다.
* 💻  Git 및 7-zip 프로그램 설치 후, ComfyUI 공식 GitHub에서 Windows용 압축 파일을 다운로드하여 설치합니다.  NVIDIA 그래픽 카드 사용자는 GPU 버전, 그렇지 않은 경우 CPU 버전을 실행합니다.
* 🖱️ ComfyUI 인터페이스 기본 조작법(화면 이동, 확대/축소, 노드 이동 및 크기 조정, 다중 노드 선택 및 이동)을 설명합니다.
* 🖼️ Civitai 웹사이트에서 원하는 체크포인트 모델(예: DreamShaper)을 다운로드하여 ComfyUI의 `checkpoints` 폴더에 넣고, ComfyUI에서 새로고침하여 모델을 불러옵니다.
* ⚙️  **`KSampler` 노드 설정 (Steps: 30, Sampler: DPM++ 2M Karras)** 과 기본 워크플로우에 사용되는 노드 (Load Checkpoint, CLIP Text Encode, Empty Latent Image, KSampler, VAE Decode, Save Image)의 역할을 설명합니다.
* 📁 기존 Web UI에서 사용하던 모델들을 ComfyUI에서도 사용할 수 있도록 `Extra Model Path` 파일을 수정하는 방법을 제시합니다. "**기존 webui 경로를 복사하여 붙여넣기**" 하여 모델들을 공유하고 관리하는 방법을 알려줍니다.

Hello, this is Jopidi. Many subscribers have been requesting a Comfy UI lecture. If you want to create images

easily using Stable Diffusion, please watch the previous Web UI lecture. If you want to go into more depth about Stable Diffusion,

I recommend this lecture. I think many people thought Comfy UI was too difficult and gave up after seeing the complex nodes on the first screen.

But don't worry. From today, I will provide a systematic guide to help you understand the difficulties you are thinking of

by explaining them step by step from the basics. This Comfy UI lecture will also be a long-term project that will be produced as a series.

We will prepare it so thoroughly that anyone can use it like an expert if they follow it carefully without missing anything.

Shall we start right away? Here is the table of contents for today's lecture. We will proceed in the order

of installing the Git program, installing Comfy UI, and setting up the checkpoint model. Let's start with the installation first.

If you do not have the Git program installed on your computer, you must first install the Git program.

The site address is in the 'More' section of the description. Download the program that matches your computer's OS from the download page.

Just run the downloaded program and follow the installation process. Now let's install Comfy UI. Go to the official Comfy UI GitHub address in the description box.

Scroll down a bit and click on the Windows Download link under 'Install' to download the compressed file. The compressed file is in 7-zip format.

If you do not have the 7-zip program installed on your PC, you can install the 7-zip program through this link.

Unzip the downloaded compressed file and copy and paste it to the path where you want to install it.

I copied and pasted it to the D drive. The file we just received is portable, so Python is included without the need for a separate installation.

There are two executable files. If you have an NVIDIA graphics card, run the NVIDIA GPU batch file. If not, run the CPU batch file.

You can run ComfyUI without a graphics card using the CPU batch file, but the CPU version will probably be very slow.

I will run the NVIDIA GPU version. It runs really fast. Those who have used the Web UI may not be familiar with this fast execution speed,

but this speed is correct. It is proof that ComfyUI is that light. Let's take a quick look at the ComfyUI screen.

You can pan the screen by clicking and dragging the left mouse button on an empty space on the screen,

and you can zoom in and out by rolling the mouse wheel up and down. Nodes can also be arranged by clicking and dragging with the mouse

to change their position. You can resize a node by clicking and dragging its border. When selecting multiple nodes, hold down the Control

key on your keyboard and drag the mouse to select all nodes within the selected area. When moving multiple nodes at the same time,

you can move multiple nodes simultaneously by holding down Shift on your keyboard and dragging the mouse.

The most important element in ComfyUI is the node. In ComfyUI, nodes represent different units of work

used in the image generation process. The left side of the node receives input and the right side produces output.

You can generate an image by clicking the "Queue Prompt" button in the menu panel. But when I click on it,

the load checkpoint node turns red and I get an error. If you check the error message, it says that the currently set checkpoint cannot be loaded.

Currently, I only have ComfyUI installed and there is no checkpoint model. Go to the Civitai site to download the checkpoint model.

Select a model from the menu and select a checkpoint from the filter. You can download the desired model from the checkpoint list.

I will download the DreamShaper model. Select the version of the model at the top of the model page

and click the Download button to download it. The downloaded model is located in the ComfyUI Windows Portable folder.

Then copy and paste it into the checkpoints folder inside the models folder. Go back to the ComfyUI screen and click the Refresh button in the menu panel.

Once the screen refreshes, you can click on the Load Checkpoint node to select the downloaded model.

Again, you can click on the example image in the checkpoint model screen of Civitai to see the detailed options for the image.

You can see that Steps is set to 30 and the sampler is set to DPM++ 2M Karras.

First, let's set Steps to 30 in KSampler in ComfyUI. When setting options, you can edit them by

clicking the left and right arrows, or you can click and enter the desired options directly. Select the Sampler Name option as DPM++ 2M and select Schedule as Karras.

Taking a look at the nodes in this basic workflow, we select a model from the Load Checkpoint

node and input positive and negative prompts through two CLIP Text Encode nodes, respectively. The Empty Latent Image node sets the resolution

size and batch size of the image to be created. The KSampler node plays an important role in converting the prompts

you set into something that the AI can understand and removing noise. And the VAE Decode node is responsible for restoring the code

existing in the latent space into an image. The Save Image node is a node that saves the image. You can set the name of the image saved here.

If you set the name to ComfyUI, the name of the image will be determined in the format of Comfy

UI followed by a number. Here's a tip: If you enter letters or numbers separated by a '/' symbol in front of this name,

The part before the "/" will become the folder name. If you create a lot of images, it's easy to manage them later by creating folders.

like this and saving the images. Click the Queue Prompt button again to generate the image. This time you can see a green border being created on the node.

This is a feedback presentation to intuitively express the progress of ComfyUI in a visually easy-to-understand manner.

The image has been created. When you create an image, if you click the Queue Prompt button once, you can see that the Queue Size becomes 1.

At this time, if you click the Queue Prompt button multiple times, the Queue Size value will increase according to the number of times you clicked.

Images will be created in order, as many as the value set in Queue Size. The generated images are saved in the Output folder within the ComfyUI folder.

You can see that the folder has been newly added and the images have been created as set in the Save Image node.

Most users who use ComfyUI probably have prior experience with the Web UI. Then, it would be very convenient to share

and use the LoRA model, upscale model, etc., including the checkpoint model used in the existing Web UI.

ComfyUI has a system prepared in advance for such cases. There is a file called `Extra Model Path` inside the ComfyUI folder.

If you delete the extension example at the end of this file, the extension will be YAML. When you change the extension,

a warning pop-up will appear; click the Yes button. Open the file with the changed extension in Notepad.

You need to edit the base path part in Notepad. Copy and paste the existing webui path here.

Save this file and then re-run ComfyUI. If you check the load checkpoint model node, you will see that all previously used models are loaded and displayed.

With this setup, you can store and share models in one place without having to store large models in both the Web

UI and ComfyUI, allowing you to use them together. That's all for today's lecture. Today we learned about installing ComfyUI

and setting up a checkpoint model. In fact, there are many different ways to install ComfyUI, but I can't cover everything in this video,

so I just showed you the easiest way. I have published a book called *ComfyUI Complete Guide*,

which covers all the installation methods, including how to install on Google Colab for those with insufficient computer specifications,

how to install on macOS, and how to install manually on Windows. For those who want to learn ComfyUI in more depth,

it would be better to study by watching videos along with the e-book. Well, everyone, have fun

ComfyUI 완벽 가이드 Part.1 설치부터 시작

쉽고 재미있는 AI 기술, 빠르게 발전하는 기술들을 쉽게 설명하기 위한 연구소.

------------------------------------------------------------------
✉️ 강연/협업  문의: jopdlab@gmail.com
------------------------------------------------------------------


조피디 연구소 JoPD LAB

ComfyUI 강의 시작 안내 및 Stable Diffusion 관련 이전 강의 소개

ComfyUI의 어려움에 대한 우려 해소 및 단계별 설명 강조

강의 목차 소개: Git 설치, ComfyUI 설치, 체크포인트 모델 설정

Git 프로그램 설치 방법 안내 (설치 파일 링크는 설명란에)

ComfyUI 설치 방법 안내 (GitHub 링크 및 압축 파일 다운로드)

압축 해제 및 설치 경로 설정, Python 설치 불필요 및 실행 파일 선택 방법 설명

ComfyUI 기본 인터페이스 및 기능 사용법

화면 이동, 확대/축소, 노드 이동 및 크기 조절 방법 설명

노드의 개념 및 입력/출력 설명

이미지 생성 시 발생하는 오류 (체크포인트 모델 부재) 설명

체크포인트 모델 다운로드 및 설정

Civitai 웹사이트에서 체크포인트 모델 다운로드 방법 설명

다운로드한 모델 파일 위치 및 ComfyUI checkpoints 폴더로 이동 방법

ComfyUI에서 모델 로드 및 Civitai에서 이미지 옵션 확인 방법

KSampler 설정 (Steps, Sampler Name, Schedule)

ComfyUI 기본 워크플로우 및 이미지 생성

주요 노드 (Load Checkpoint, CLIP Text Encode, Empty Latent Image, KSampler, VAE Decode, Save Image) 설명

이미지 저장 이름 설정 및 폴더 생성 팁

이미지 생성 과정 및 Queue Prompt 버튼 기능 설명

Web UI 모델과의 호환성 및 추가 모델 경로 설정

Web UI 모델과 ComfyUI 모델 공유 방법 설명

`Extra Model Path` 파일 수정을 통한 Web UI 경로 설정 방법

모델 공유 및 효율적인 관리 방법 강조

강의 요약 및 추가 설치 방법에 대한 책 소개

책 내용 소개 (Google Colab, macOS, Windows 수동 설치 등)