ComfyUI AI Video: Free Generation — Step-by-Step Guide & Transcript

[00:00] Today, I’ll show you how to generate AI  videos for free on your own PC using ComfyUI.   We'll install everything from scratch,  generate videos from simple text prompts,   animate existing images, and by the end of this  video, you'll have everything running on your PC  

[00:16] with no API keys, subscriptions, or usage credits.  I'm Kevin, and let's dive in. First, head to the   following website to install ComfyUI. You'll find  a link right here at the bottom of the screen.  

[00:28] ComfyUI is a free, open-source application that  lets you run AI models directly on your own PC.   That means once you install it, there are no per  video charges. Today, we'll use it for AI video  

[00:41] generation, but it also supports image generation,  image editing, audio generation, and much more.   On the website, you'll see options for desktop  and cloud. We'll use the desktop version since   everything runs on your own PC. If you don't have  compatible hardware, the cloud version is a good  

[00:57] alternative, but it requires a paid subscription.  For this tutorial, I recommend using a Windows PC   with an NVIDIA RTX graphics card, since that's the  easiest and most reliable configuration to follow   along with. Click Download Desktop, then click  Download Desktop again to download the installer.  

[01:14] Once the download finishes, run the installer.  After installation completes, launch ComfyUI,   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  

[01:29] Continue. Next, let's give this installation a  name. I'll set mine to ComfyUI. For most people,   you can leave the remaining settings as is, and  then down below, let's click on Continue. After   it opens, you'll land here on the home screen.  Over on the left-hand side, you'll see all the  

[01:44] different types of AI content that you can create  with ComfyUI, including images, videos, audio, 3D   models, and here, even large language models. Now,  today, we'll generate AI videos, so right up here,  

[01:57] let's click on Video. Here, you'll see many of  the different video models available in ComfyUI.   Some run directly on your PC, while others use  cloud services that require a paid subscription.  

[02:09] Since we want to generate AI videos for free,  let's filter this list to only show models that   run on your own PC. Under the Runs On dropdown,  let's click on that, and then select ComfyUI,  

[02:21] and there, we see all the free models. Next,  let's filter by task. I'll click on this dropdown,   and if I scroll down the list, let's start with  the basics of text to video. I'll select this,   and now, you'll see several free models. Right up here, we have the LTX 2.3, WAN 2.2,  

[02:36] Hunyuan, here we have Kandinsky, and SVD, lots of  excellent models. For this tutorial, I'm going to   use LTX 2.3, and the reason why is it's fast, it  produces excellent results, and it also runs on a  

[02:51] wide range of RTX graphics cards. Of course, you  can always come back later and experiment with   all these other models. So right over here, let's  click on LTX. This drops us on the ComfyUI canvas.   If you've never used ComfyUI before, this might  look a little intimidating, but don't worry,  

[03:07] we'll make sense of it. First, in the upper  right-hand corner, you'll notice that ComfyUI   shows a few errors. That's expected since we  haven't downloaded the LTX 2.3 model yet. Right   over here, let's click on Show Missing Models.  In my case, there are five missing models. You'll  

[03:24] also notice that they're fairly large. Together,  they require about 40 gigabytes of storage,   so make sure you have enough available disk space  before continuing. Now right up here, I’ll click   on Download All. ComfyUI will now automatically  download all the files required for this workflow.  

[03:40] Depending on your internet connection, it may  take a little while, but you only need to download   these models once. To track the progress of the  downloads, you can click on this icon right up   above. Once the download finishes, we're ready to  generate our first AI video. At the center of the  

[03:54] screen, you'll see the LTX 2.3 workflow. First,  you'll see some helpful tips and information   about the workflow. Next, you'll see two nodes.  The first is where we'll configure and generate  

[04:06] the video, and the second automatically saves the  finished video. Let's start with the first node.   Here, I'll zoom in so we can see it a little bit  better. Although there are quite a few settings,   you only need to change a handful to get started.  At the top, enter a description of the video  

[04:20] you'd like to generate. The more descriptive your  prompt, the better the results will typically be.   I'll enter in my prompt. I want a Kevin Cookie  Company blimp flying over a city at golden hour.   Below the prompt, you can adjust the width  and height to control the resolution. Here,  

[04:35] we have the duration to control how long the video  is and the frame rates to control how smooth the   motion appears. For this first example, I'll leave  the default settings, but feel free to experiment   with them later. Now, in the top right-hand  corner, click on Run to generate the video.  

[04:50] Depending on your graphics card, this may  take a few minutes. And right down below,   we could see that it's now finished. Let's preview  how it turned out. The journey is just beginning.

[05:02] The aesthetic of the video is very nice, but  it's a little bit more of a cross between a   hot air balloon and a blimp. Of course, I could  always go back and make my prompt more specific.   ComfyUI automatically saves the generated video  for you. To view all of your generated content,  

[05:17] over on the left-hand side, click on Assets. And  from here, you could preview your videos, and if   you right-click, you could save it elsewhere on  your PC, or down below, you could also delete it.   I'll click out of Assets. In addition to  generating videos from text, ComfyUI can also  

[05:32] animate existing images. Let's now head back to  the templates gallery. Down below, I'll click on   Templates. Let's go back to the tasks dropdown.  And over here, let's deselect Text to Video,   and if we scroll up just a little bit, here  we'll find the image to video. I'll select that.  

[05:47] And right over here in the top left-hand corner,  we have LTX 2.3. Let's select that since we've   already downloaded those models. Now, you'll  notice that this workflow looks very similar   to where we just were, except now we can start  with an image. So over here, I'll drag and drop  

[06:03] in an image of the Kevin Cookie Company blimp,  or at least what I think looks more like a blimp.   Then right over here, we could enter in  a prompt describing the motion that you   would like to see. So over here, I'll type  in the prompt. Now, for this first example,  

[06:16] I'll leave all the remaining settings as their  defaults. And in the top right-hand corner,   just like we did before, let's click on Run.  Down below, let's preview how this turned out. We've only scratched the surface of what ComfyUI  can do. In addition to Text to Video and Image  

[06:35] to Video, you'll also find workflows for video  editing, camera control, inpainting, upscaling,   and many other AI tasks. You'll also notice that  ComfyUI supports many different AI video models.   Today, we used the LTX 2.3, but you can also  experiment with other models to compare the  

[06:51] results and find the one that works the best for  your projects. As you become more comfortable   with ComfyUI, you'll also be able to customize  these workflows, combine different models, and  

[07:03] create even more advanced AI videos. And that's  how you can generate AI videos for free using   ComfyUI. If you have any questions, or if there's  another ComfyUI workflow you'd like me to cover,  

[07:15] let me know in the comments right down  below. I'll see you in the next video.