ComfyUI Tutorial: Install, Generate, and Upscale (Full Transcript)

[00:00:00] Speaker 1: In this video, I will show you how to use Comfy UI step by step. Comfy UI is one of the most popular free and open source tools for creating AI generated content on your own computer. People use it to generate images, create videos, enhance photos, and build powerful AI workflows using a visual node-based interface. In this tutorial, we'll start from the very beginning. We'll install Comfy UI, set up an image model, generate our first AI image, and then learn how workflows and nodes work so you can start building your own projects. I'm Kevin, and let's dive in. To get started, head to the following website. You'll find a link right here at the bottom of the screen. Here on the homepage, you'll notice that Comfy UI offers both a desktop version and a cloud version. For this tutorial, we'll use the desktop version since it's free to use, runs directly on your own computer, and gives you full access to your models and workflows. The cloud version can be a good option if you don't have a powerful computer or maybe you'd rather not manage the hardware yourself, although there is a cost associated with it. Let's click on download and then download again. Once the installer finishes downloading, run through the install process. After installation completes, launch Comfy UI, and you'll see the following screen. For this tutorial, we'll select local so everything runs directly on your own computer. Down below, I'll agree to the terms, and then let's click on continue. Next, let's give this installation a name. I'll set mine to Comfy UI. For most people, you can leave the remaining settings as is, and then down below, let's click on continue. Once Comfy UI launches, you'll land on the main getting started screen. Now, right away, you can see that Comfy UI supports much more than just image generation. Over here on the left-hand side, you'll find categories for images. Here we have videos, audio, 3D models, large language models, and more. To help you get up and running quickly, let's go back to the getting started section, and let's start with text to image. I'll click here. Comfy UI now loads a complete image generation workflow. Now, notice in the top right-hand corner that Comfy UI is telling us that we're missing a required model. Think of Comfy UI as the application that allows us to build and run AI workflows. The model itself is what actually creates the image. To see what we need, let's click on show missing models. Right here, Comfy UI shows us the exact model required by this workflow. To install the model, right over here, let's click on download all. Right up on top, we can now check on the download progress. Now that the model has finished downloading, let's generate our first image. Let's take a quick look at the workflow. You could zoom in and out using your mouse wheel, and you could press the middle mouse button to pan around the canvas. You can also use the mini map in the bottom right-hand corner to quickly navigate around larger workflows. Now, notice that this workflow is divided into several different sections. Right over here, we start by loading the model that we just downloaded. Next, we enter a prompt, and this describes the image that we'd like to create. Now, I want to generate an image of a chocolate chip cookie, so right over here, I'll enter in the following prompt. Now, below that, we have the negative prompt. This tells the model what we'd like to avoid in the generated image. So over here, I'll enter in my text. If you want to follow along, you'll find these same exact prompts in the description of this video. Next, we can configure the image size. Here, we could set the width, we could also set the height, and also the number of images we'd like to generate. Now, I'll leave the default settings in place since they work well with this model. Next, let's go up and we'll look at the sampling section. This is where Comfy UI generates the image. You'll notice several different settings here. Now, as a beginner, I recommend leaving these at their default values. Now, there is one setting worth knowing, and that's the seed. You'll find that right here at the top of the list. Changing the seed allows you to generate different variations of the same image all while using the same prompt. For now, let's leave all of these settings as is. Then, in the top right-hand corner, let's now click on Run. And there we have our first AI generated image. It also automatically saved the image to your computer. Now, you might be wondering, how do you view all of your generated images or assets? Well, over on the left-hand side, we have this button for assets. When I hover over, you'll also notice that there's an associated shortcut key. You can press the letter A. When I click on that, here I can see all the different images that I've generated. If I right-click over here, I could save it elsewhere on my computer, or down below, I could also delete the asset if I'd like. Right up above, I'll close this view. Now that we've successfully generated an image, let's take a closer look at how this workflow works and how we could modify it to produce a different result. Now, right now we're generating just a chocolate chip cookie. Let's see what happens when we make a few changes. So right over here, we have the prompt field, and let's update this text to instead create, let's say, a double chocolate chip cookie with a melted chocolate center. That sounds so good. Now, if we scroll down just a little bit, here we have the image size. And previously, we went with a 512 pixel by 512 pixel image, and that produced a square image. But let's actually try a different aspect ratio. So right over here, let's change the width to, let's say, 768. I'll enter in that value, and then right up on top, let's try running this. I'll click here. And there, we have a double chocolate chip cookie, and now it's also a wider image. Now that we've generated a few different images, let's take a closer look at how this workflow is organized. One of the things that makes Comfy UI unique is that the workflows are built using nodes. Each node performs a specific task. Here's an example. We have our first node, and that loads the checkpoint, or the model that we downloaded. And together, all of these different nodes create the workflow that generates this final image. Now, as you look across this canvas, notice how all of these different nodes are connected together with all of these different lines. Data flows from one node to the next until we eventually arrive at the finished image. Now, the nice thing about this approach is that we're not limited to the workflow exactly as it exists today. We could easily extend it by adding new nodes. I'll zoom in right here on this last stage. To add a node, you could simply right-click anywhere on the canvas, and here at the top of the menu, you have the option to add a node. I'll click on that, and that opens up a menu with hundreds of different nodes that you can use to customize your workflow. Now, I'd like to add an image upscaling node. So here within the image, at the very top of the list, we have the option for upscaling. And right here, we can upscale the image bio. I'll click on that. Right over here, it's now added a new upscaling node. And let's try upscaling the image, let's say by 2x. Here, I'll enter a two and click on OK. However, right now, this node is not connected to the overall workflow. So over here, let's take the output of the decoding node, and here I'll feed it into the upscaling step. And over here, I'll take the output of the upscale step, and let's feed that into the save image step. That way, we'll upscale the image before we end up saving it. Now, let's run the workflow again. In the top right, I'll click on Run. And right there, you can see that we've now enhanced the image simply by adding a single node to the workflow. Right down below, you'll now see the new dimensions of this image. We've effectively upscaled it by two. This is the core idea behind Comfy UI. By combining and connecting different nodes, you can build workflows that generate images, edit photos, create videos, process audio, and much more. Now that we've customized this workflow, let's save it so we could use it again later. Right up on top, you'll see all of your different workflows. I'll right-click on this one, and right here, we see the option to save. I'll click on this. Right here, we could enter in a file name. I'll call mine the Upscale Image Generator, and down below, let's click on Confirm. To get back to your workflow, over on the left-hand side, click on Workflows, and there you could see the workflow that we just saved. You can also share workflows with other people, which is one of the reasons Comfy UI has such a large and active community. Instead of building everything from scratch, you can start from an existing workflow and then modify it to fit your exact needs. Here, I'll close out of Workflows. Before we wrap up, let's take one more look at the template library. Right down here at the bottom, let's click on Templates to open it. One of the things I like about Comfy UI is that you have a large collection of starter templates to get you up and running quickly. Now, in addition to text-to-image, which we just used, here you'll find other fun ones, like image-to-video. Here you have image editing. Here we have image-to-video again, image-to-3D. You even have text-to-song, so lots of different options and fun things that you can play with. Now, I recommend clicking in, downloading the models, and then exploring how they're built. Every workflow here is made up of nodes, and you could use these templates as a starting point and then customize them to fit your own needs. And that's how to use Comfy UI. Now that you understand the basics, go ahead and experiment with the built-in templates and start building workflows of your own. If you'd like to learn more about AI tools and how to get the most out of them, consider subscribing. Thanks for watching, and I'll see you in the next video.