Introduction and installation

0:00 Each chapter builds on the previous one. We start simple, repeat the important ideas, and only add complexity when it actually makes sense. You do not need any coding knowledge. You do not need to be technical. If you can think visually, you can learn Comfy UI. If you want to understand how AI image generation really works locally, and how to use Comfy UI without feeling lost, this course is for you. My name is Pixel Roma, and on this channel I focus on creating and teaching Comfy UI workflows in a simple and practical way.

0:46 I am a graphic designer, not a programmer, and that is actually a good thing. Developers are great at writing code, but they often explain things in a very technical way. This course is designed from a visual thinker's perspective. My goal is to explain Comfy UI logically and visually, without needing any coding knowledge. Even if Comfy UI looks confusing right now, that is completely normal. We will start from the absolute basics and build up step by step.

1:09 But before we talk about Comfy UI itself, we first need to understand what AI image generation actually is. Today, AI is not just one thing. There are many different AI models that can run locally on your own computer, such as Stable Diffusion, Flux, Kwen, and many others. It is also important to understand that Comfy UI is not limited to image generation. Comfy UI is a general interface for running many different types of AI models locally. While it is most popular for image generation, it can also be used for audio, music, video, animation, 3D, and more.

1:57 As long as a model can be connected through nodes, ComfyUI can be used as the interface to control it. These models by themselves are like an engine. They are very powerful, but you cannot really use them directly. To work with them, we need an interface. An interface is what allows us to send prompts, images, and settings to the model, and then receive results back. There are many free interfaces that let us interact with these models. Some popular ones are ForgeUI, SwarmUI, Invoke, Focus, and of course ComfyUI.

2:20 They often use similar models, but they work in very different ways. In this course, we are going to focus on ComfyUI. ComfyUI is different because it is node-based. Instead of hiding everything behind buttons and menus, it shows you exactly what is happening step by step. You can see how prompts, models, samplers, and images are connected together, like building a system. Think of it like this. The AI model is the brain. The interface is how you talk to that brain.

3:08 ComfyUI is like building your own control panel, exactly the way you want. Do not worry if this still feels complex. Understanding comes from seeing things connect, not from memorizing nodes. In this course, I will explain what each node does, why it exists, and how everything connects together. Before we install anything, we need to talk about how you will actually run ComfyUI. There is more than one way to use ComfyUI, and the right choice depends on your system and your expectations.

3:31 Let's go to the official ComfyUI website to see the available options. The official website is comfy.org. If we go to the product section, you can see that there are two main options, ComfyUI Cloud and local ComfyUI. ComfyUI Cloud runs online on their servers, and it is a paid service. This option can be useful if your computer is too old or not powerful enough to run AI models locally. Local ComfyUI is free and runs directly on your own computer, assuming you have a reasonably capable system.

3:54 This is the option we will focus on in this course. So, let's click on local ComfyUI. Here you can see three main installation options. Download for Windows, for macOS, and install from GitHub. All of these options install ComfyUI, but there are important differences between them. In this course, I will focus on Windows operating system using the portable version of ComfyUI. All the workflows, tools, and installers I show are tested on Windows using an Nvidia graphics card.

4:42 On AMD graphics cards and on macOS, performance is usually slower, and some features or custom nodes may not work exactly the same way. So, if you are using Windows with an Nvidia card, it will be much easier to follow this course step-by-step as I show it. Because there are many different AI models, hardware requirements can vary a lot. Some models are small and can run on a graphics card with 6 to 8 GB of VRAM. Other models are much larger and may require more than 24 GB of VRAM.

5:04 For this first episode, I tested the workflows on two different systems. One system uses an RTX 2060 with 6 GB of VRAM and 64 GB of system RAM. The second system uses an RTX 4090 with 24 GB of VRAM and 128 GB of system RAM. For the workflows in this episode, a graphics card with 6 to 8 GB of VRAM should be enough to follow along. In later episode, we will explore newer and larger models that may require more powerful hardware.

5:54 Now, let's talk about which version of ComfyUI you should install. As I mentioned earlier, I am using the portable version of ComfyUI. If we click on install from GitHub, we are taken to the official ComfyUI GitHub page. Here you can find detailed installation instructions, but they require more manual steps and setup. Over the past year, I have been using a portable version of ComfyUI that includes additional tools to make the installation process much easier. This installer installs the original ComfyUI, but it also adds helpful tools so you can get up and running much faster.

6:17 This installer is called ComfyUI Easy Install. You can find it on this GitHub page. You can find the creator on our Discord community under username Ivo. Thank Ivo for this installer. This entire course is built around this version of ComfyUI. You can still use ComfyUI Desktop or ComfyUI Cloud, but some things may look different or behave differently compared to what you see in this course. If you want the exact same setup that I use, and the easiest way to follow along, I recommend using the same version.

7:03 Let me show you how to install ComfyUI. So, we are on the Easy Install GitHub page. This is the complete link. If we scroll down, you can read more about this installer. Even if you might not understand what each of these things means yet, it will make sense later as you learn more about it. I will talk later about the Pixelama Discord server, where you can get more help and answers to your questions. So, this installer will install Git, which is a tool that tracks changes to files.

7:26 In ComfyUI, it helps developers safely update the main app and custom nodes, fix bugs without breaking everything, and lets you update or roll back to an earlier working version if a new update causes problems. Then it will install the ComfyUI portable version. A portable version means the program is fully self-contained in one folder, does not need a normal system install, and can be run, moved, backed up, or deleted without affecting the rest of your computer. In ComfyUI portable, this means Python, libraries, models, and settings all live inside the ComfyUI folder.

7:48 So, you can copy it to another drive or PC, update it safely, and avoid breaking your system Python or Windows setup. Python embedded means ComfyUI comes with its own built-in copy of Python already included inside the ComfyUI folder instead of using the Python installed on your system. Then it will install all the nodes that are useful and that I tested over the last year. It might not make sense for you yet if you are a beginner, but do not worry.

8:35 Take it as general knowledge for now, and it will make sense later. Then it will add an add-ons folder with more advanced stuff we can use later to speed up our generation, plus some extra tools that can be useful, and then more technical stuff explained for each one. But all you need to know for now is where to download it from and how to run the installer. It is important not to run it as administrator. That means you just double-click to run it, not right-click and run as administrator, but you will see that in a minute.

8:58 Also, avoid system folders, and make sure your Nvidia drivers are up to date since some things work only with more recent Nvidia drivers. Okay. Let us go back to where it says Windows installation, and let us download the latest release from here. Then, depending on how your browser is configured, it will either download it to the downloads folder or ask you where to download it, and you can decide where to put it. As you will see over time, it needs a lot of space if you download big models.

9:21 So, I suggest downloading and installing your ComfyUI on a hard disk that has a lot of free space and preferably on a solid state drive because it will load the models faster. I will go to my D drive and I will create a new folder called ComfyUI. But, this does not really matter. The name can be anything easy to remember. Sometimes I put ComfyUI followed by the month, so I know when I downloaded and installed it.

10:06 So, I will save this zip archive in that ComfyUI folder. Now, let us go to the place where we saved the file. Since this is a zip archive, we need to unzip it. You right-click on it and depending on what you prefer, you can use the Windows integrated option and select extract all. I like to delete the folder name at the end, so I do not end up with a folder inside another folder. When I click extract, it will extract these two files.

10:31 Let me delete it really quick and show you. If you have WinRAR like me, you just choose extract here and it does the same thing. Once we extracted everything, you can delete the easy install zip file. Now, we are left with two files. A bat file that is the installer and a zip file that contains extra resources that it will use. When you run it, you might get a security warning. That usually happens with bat or executable files because they install files on your system.

10:54 This one is safe. I installed and tested it and I personally know Evo, the creator of this installer. You can right-click and scan it with your antivirus and you will see it is clean. So, let us double-click on the bat file, then press run and it will start installing. If you already have Git installed, it will update it. If not, you will get a window like this and you have to press yes to continue the installation.

11:16 After that, it will continue the installation of ComfyUI and everything it needs to run. You can take a break for 3 to 5 minutes depending on your internet speed and your computer. So, how do you know when it is ready? You'll see a message that says installation complete along with the time it took. On my PC, it took 247 seconds. After that, you can press any key to exit. So, let us recap really quick. From the GitHub page, you download the zip archive of the easy installer.

11:39 You create a folder named ComfyUI and place the downloaded zip archive in that folder. You extract the contents of the archive. You run the ComfyUI easy install.bat file, and if it asks to run GitHub, you press yes. In a few minutes, the installation is complete and you get this screen. Now, after the installation, we can see that inside our ComfyUI folder, a new folder appeared called ComfyUI easy install. This is portable, which means you can copy this entire folder and move it to a different drive or folder, and it will still run ComfyUI.

12:27 Basically, after you install any ComfyUI portable version, you should end up with a similar folder structure to this. Since ComfyUI is based on the Python programming language, you will see many Python files and bat files inside these folders, which are used to run those Python files. The easy installer will also create some shortcuts on your desktop. If you use other versions of ComfyUI, they might not do this, and you would need to create the shortcuts manually.

12:52 That is one of the reasons I prefer the easy installer. It makes everything easier for us. Basically, we just extracted an archive and ran a bat file, and we now have ComfyUI. If we right click on this shortcut and go to properties, we can see the target of the shortcut. If we open the file location, you can see that it is connected to this bat file that starts Comfy UI. In other versions of Comfy UI, the name might be different, like run Nvidia GPU or something similar.

13:16 Are you ready for your first Comfy UI launch? To start Comfy UI, you either use this bat file called start Comfy UI, or from the desktop you use this shortcut called Comfy UI Easy Installer. E and Z stand for easy, and I stands for installer. So, double-click on it, and when it starts, it looks like this. The first time it will be a little slower, but after that it will start much faster. If you are curious by nature, you can find all kinds of information about your Comfy UI and your system when it runs.

14:01 For example, you can see what operating system I am using, what Python version is running, and where that Python is located, what the path to the Comfy UI folder is, where the user directory is, how much VRAM you have, how much system RAM you have, and the PyTorch and CUDA versions that are running. When it starts, this command window will be minimized to your taskbar, and Comfy UI will open in your default browser. The first time it opens, it will show you some templates made by Comfy UI that you can load.

14:25 If you have run a workflow before, it will open that workflow by default. So, the workflow you see open is the last one you used. A workflow is a set of connected nodes that tells Comfy UI what to do step-by-step. Let us close this for now. You can open these templates or workflows from here later. Comfy UI is made of a few main areas. You do not need to memorize them. I am naming them so we can talk about the same things later.

14:48 If we go to the top, you can see it says unsaved workflow. Basically, it is like a document that is empty at the moment, since we did not add any nodes yet. You can have multiple documents open, similar to what we have in Photoshop and other programs. We can click on this plus icon to create a new blank workflow. All these tabs on top are open workflows, and we can close, save, and edit each one. Now, this grid-like empty space is called the canvas.

15:34 Instead of drawing on it, we will arrange blocks or nodes, like using LEGO pieces, and connect things together to create a working workflow. You can use your mouse wheel to zoom in and out on the canvas. Then we have this top bar. Depending on what extensions you have installed, it might look different and have more options. Things like the manager or the run button, which lets us run workflows, are usually here. On the bottom right, we have view controls.

15:57 For example, we have a select tool that lets us select nodes and a hand tool that lets us navigate the canvas. You can fit a workflow in view, but right now the canvas is empty. We also have different zoom controls that you can use if you do not want to use the mouse wheel or if you do not have one. For me, the mouse wheel is the fastest and the one I prefer. Then we have the show mini map option.

16:20 This shows a small map that we can use to navigate when we have very large workflows. There's also hide links, but since we do not have any nodes or links yet, we will see that later. An important one is the main menu, which you open by clicking on the letter C, the ComfyUI logo. We also have more options on the left sidebar for nodes, models, and workflows, which we will explore soon. Back to the main menu.

16:43 If we click on the C, we get this menu. New creates a new workflow, but it is faster to use the plus sign from the top bar. For file, it allows us to open, save, and export the workflows we create. For edit, you can undo actions like moving nodes or changing something in the workflow, clear the workflow, and unload models. For view, we can enable and disable different panels, and we also have zoom in and zoom out controls.

17:06 Just like in Photoshop, we can do the same things in multiple ways. It is the same with ComfyUI. For theme, you can change how it looks, but at the beginning, I suggest leaving it on default, so it is easier to follow tutorials. Nodes 2 is in beta at the moment of this recording and still has some bugs, so I suggest leaving it off until it is more stable. You can browse templates and open settings, which we can explore later.

17:50 For now, the default settings work fine. Templates and settings can also be accessed from here. So again, there are multiple ways to access the same things. In some newer versions, some people might use a newer manager, and it might appear somewhere else instead of here. For now, I am using the old manager, which appears here. Under help, you can also find help options, but you will see later in this video how to ask questions and get help.

18:13 We also have a console, sometimes called the bottom panel, where you can see exactly what has happened since we opened ComfyUI. If we look at the taskbar and open the command window from there, it shows the same information. One view is at the bottom and the other is in the taskbar. To close ComfyUI, I recommend opening the command window from the taskbar and closing it. You will then see a reconnecting message in the browser, because it cannot find ComfyUI running anymore.

18:35 After that, you can close the browser window. You can also close the browser window first and then close the command window. It is time to test our