Hello, I'm Choi Don-hyeon of
STABLE DIFFUSION KOREA and SOY.LAB. Nice to meet you all. Today, we're going to talk
about the ComfyUI STARTER KIT. Let's see how to install and
use ComfyUI in Windows. There was something I really wanted
to thank you for before we started. We finally have more than
1,000 soy_lab subscribers. It's a channel that just
started, but I'd like to say thank you to all the
subscribers who support us. I'm really happy. Thank you so much. I'll try to continue to give
you good videos and content. It was a few days
before we started talking. A member of STABLE DIFFUSION KOREA on
Facebook saw the ComfyUI related content. I want to know more about the basics. Can you tell me
about this? Like this. There was a story like this. I kindly said, I'll explain the related part in a video. There was an issue that I said. So, we're running this content from the point of view of
our soy_lab. It was so exciting. Wow, finally look at ours. There's someone who's curious. Then, can't we make this? I really liked it. So, Let's make it now. That's what I thought and prepared. But now these requests. Little by little, it builds up At some point, There are so many parts. Many people. There are so many people who want to know. We've come to know
that there are many people. I thought a lot about what to do with this. Anyway. I don't know if I should cry or laugh. This happy feeling. How should I do this? I think there was a point
where the fear was so strong. But what can I do? I've been preparing really hard. We're going to answer you. I've been preparing. And finally. I've prepared a content that I can explain. Today, we're going to tell you. About installing ComfyUI on windows. I'm going to explain it. If you click on the path here. In this way, Connected ComfyUI installation techniques. You can check it out. Even if you look at the content here. You can check out all the related content. The most important thing is this. What is ComfyUI? There may be people who think that, but ComfyUI may be a little
unfamiliar to those who are watching for the first time, but it is
another platform of Stable Diffusion. Among the contents I explained
before, there was a thing called A1111. It's a node-based
platform that uses the same Stable Diffusion
process as this platform. So, in the case of A1111,
if it is a fully set platform, it is a node-based platform
that can be used by the user to assemble parts and parts of
the features more professionally. I think you can understand it like this. The related content can be a
little overwhelming for those who are starting for the first time, so we
have prepared the contents like this. Now, the information related to Windows, the
information related to the release, the examples related to UI, and the issues related to blog
updates can always be seen on this channel. In addition, there is a link
to the manager-related part. Since I left a link to
this part, I think it will be helpful if you check each
of the parts here one by one. Like this, information about
the part related to Windows installation or information
related to this release. What part can be updated and received? Currently, if you look at it
now, it will be a page where you can see the release
information related to Portable. I'll explain this later. And now, a page where you can check
the examples related to the ComfyUI. You can check all the
relevant functions here. And in this way, there is an
update related to the ComfyUI. And I prepared a page
that allows you to check the update issues
related to the ComfyUI. I also prepared
information about the most important manager-related
installation circuit. Now, check it out later like this. First of all, there are a few unique
things related to the ComfyUI. The important thing is that you can
use it in Windows, Linux, and MacOS. It's a great advantage. And the video video of NVIDIA and AMD. In addition, there are also
several areas related to Intel Arc. It is said that experimental
support is possible. And now, in the part related to Mac OS, I used the AMD graphic
card on the 2019 Intel MacBook model that
I recorded yesterday. I was able to check that
it accelerates here, too. So, when I used the LCM LoRa, I checked that I could create a
picture in 15 seconds based on 512*512. So, the installation content
related to Mac OS will be introduced to you later. Anyway,
in order to use ComfyUI well, Python should be installed on the PC. Well, it's not a must, but we recommend
that you install it as much as possible. And since you need to install
Python related to A1111 as well, I hope you will install
Python as much as possible. You can install this 3.11
version or less version. So, what I usually recommend
is to install version 3.10.6. Of course, if I explain it like
this, Oh, how do I install this? There may be some people
who are wondering what to do. Slightly sweating and saying,
No, this is not easy to explain, You can do it, so don't worry. What we are going to
explain today is very easy. The content to explain is I will explain the part related to
ComfyUI, and the portable install. What is this part related to the portable? If we have a specific
device already installed such as Python, Or
parts related to Git, There are various pre-installed information
and preparation situations to be installed. There are complicated things about this. It's very easy to install. Just download the
portable version and run it. That's all. I'll show you this. First, if you click the download link here, You can see that the
download is going on right away. Yes, if you get all the downloads, You can see that this type
of file has been downloaded. You can unzip this. Once you're done unzipping, You can go
in here and check out the related folders. You'll be able to check out the folders. If you look here, it says
Comfy Windows Portable. The folder here is like this. If you look here, you'll see
Run NVIDIA GPU, Run CPU. You'll see these two files. You can see that it's
made up of these folders. Once you've installed it this far, Everything starts. This is what I'm looking at right now. I'll explain to you
that it's a root folder. In other words, based
on the structure here, The part I just explained
is about this root. And below that, There's Comfy
UI, Custom Model, and Manager. This manager that we're
going to install later. The related content will be installed. I'll explain all the routes
related to the model here. First of all, before you run the
ComfyUI, I have to explain this. If you're already using the
A1111 version of Stable Diffusion, There's going to be a lot of model files. If you put those files in the model
folder that exists in the ComfyUI, If you put it in separately, There's
gonna be a lot of capacity. Now, in that case, You don't have to copy
and paste the same thing. You can modify the file here and use it. I'm going to check
the files in there again. If you go in here, There's a file called
example related to this. You can modify this. Now, please delete the example. And if you open it using a
notepad or other editing tools, In this way, the path will come out. It's going to be like this. You can use the existing path as it is. You're not using it as it is. In this way, If you match the path in
which the model is installed, If you save the related content, You can use it right away. This is the basic folder where
I'm collecting the models right now. You can save the
information about the folder where the actual model
is saved in this way. It's very simple. So, you can modify the
content you see here. In this way, I just saw this part. I've modified this part. I've set up this path
with the actual model file. In this way, the basic
model path path is all set. When you run the related part after this, If you're using an NVIDIA GPU, You can run it with this file. If this is a part related to the run CPU, If it's not an NVIDIA graphics card, You can use the CPU-related part to run it, But it's very slow. In that case, it's a little
hard for me to recommend it. If you're using an NVIDIA
product, you must use it. But if you're using an NVIDIA
product, you must use it. Of course, in these related parts, There's information that you can install
even if you're using an AMD graphics card. If you look at the parts
related to this ROCM, It's the part that accelerates. But there are some parts that
you need to look at separately. Actually, we don't have a
separate graphics card right now. I haven't tested it, but if I
can get ready in the future, I'll do a separate review of that part. Now, let's get started. If you run the Run NVIDIA GPU, It's done like this. The advantage of ComfyUI is that
we'll be able to explain it to you later. There are some parts that
are good to use professionally. Because it's not a
perfect one-build program. Pre-visualization of related parts. There are some parts that can
be implemented as a node-based. I can use it very professionally. I think I can explain it to you. First of all, it's normal
for it to run so fast. If I press Create in
this state, it's created. I'll load the model once. You can see the results
coming out like this. That's it. This is the start of the
current basic ComfyUI. If you keep pressing Queue Prompt like
this, The picture keeps being created. It can be created infinitely like this. All right. Now, ComfyUI installation is over. You've seen this interface. If you look here now,
there's a fox, there's # and there's
something like this. And this graph is in line here. I'll explain how to do that part. First of all, I'll explain
about the shortcut first. This ComfyUI has a shortcut. If you look at it now, the shortcut-related
part is also written in this way. We've copied this and pasted it. I've translated this into Korean, too. I couldn't memorize
all the content here, but I'll explain the
parts that I use a lot. There's something like this. For example, I just want to copy
the node I'm looking at right now. Then, while pressing Alt, drag. It's the same thing. There's a way to do Ctrl+C and Ctrl+V. There's a way to do Ctrl+C and Ctrl+V
by selecting several while pressing Ctrl. And while pressing
Ctrl+Shift, drag them all. There's a way to delete these. If you look here, Ctrl+M, a voiceover. Ctrl+B. This is called bypassing. It's like this. You don't let this work at all. Ctrl+B bypassing in this way. It's a function that allows
you to skip without using it. The most important thing now. I'm sure you'll be using it a lot later. Make sure you remember this. After pressing Ctrl+C here, when
you press it, it's going to be like this. You can just drop it like this. If you press Ctrl+Shift+V, you
can attach the nodes at the back. There are people who really need this. It's very difficult because of this. Oh, when do I copy
this, when do I connect it? You can do this, but there are times when you use this case
sampler in various stages. I think it would be good if
you use this function a lot. And it's a bit weird to say a few
tips, but if you're related to connection, There are some parts
that can be disconnected. You can connect it backwards. You can connect it in normal order. Like this. And it's probably one of
the most important parts. If you double-click here, All the nodes you can use. You can check out the related information. And just search like this. You can come up. And now I'm moving this canvas. Basically, left-click
and hold it and press it. You can drag it. You can do it the same
way even with the default. Yes, I'll look over the shortcuts. This time, I will look
at the interface options and icon processing in the
menu on the right bottom. If you press this, there are
options to set related areas, and there is a link-render mode below. The This part is
originally a spline. So if you look at it as a spline, It looks like this at first. It's a little more messy than you think. It doesn't look good. So in my case, I'm using it in a straight line. It's neat, isn't it? I think you can use it like this. Other than that, For some basic options, You can adjust it here. After that, we'll install a manager. If you install other extension-related parts, The convenience features will
continue to come into this setting. So what you're looking at
right now is completely pure. I think you can think
of it as a basic setting. When we use a platform called A1111, If you install a related extension, The number of options increases. I think you can think of it
as something similar to that. Now, through this method, I've learned how to
render links in a neat way. This time, I'll take a
look at this workflow load. What's a workflow load? I've made it like this. If you create it here, At the
bottom of the appropriate folder, This is how the image builds up. Can you see it? With this content, If you drag and drop
it like this, Like this. The settings I used
earlier are the same. Even UI-related interface. I've saved it like this. If you say you created it, The picture will come out in this shape. I'll try it. I purposely changed the location like this. Let's press the Create button. Now, there's an image, right? And the image is added here now. Now, let's turn it around. It's like this. If you put this like this,
Remember the location. The workflow is organized. So, among the contents we made before, I've explained the model
called Stable Video Diffusion. I've explained the
related content like this. How did you explain this? I'm going to drag this image to this node. It's written here to drag and drop. If you do this, the
workflow will be loaded. Like this. It's very easy so far. And for those of you
who are using this for the first time, I can
tell you what it's about. For those of you who have already used it,
What should I do when there's a problem? Some of you may be thinking about it. I'm going to show you an example of that. I'm going to show you some
of the content we filmed before. I'll call you here. Like this. If you look at it now, I called you. There's an error right away. If you look at it now, it's like this. You can't bring in parts of the node. You can see that. It's an error. There's a custom node
that I don't have right now. The way to solve these related parts is... For those of you who are using it
for the first time, it's very difficult. It's not working at all. There's a red light. There's a way to solve this very easily. When you call a custom node like
this, You don't have to shed tears. I'll tell you how to solve it easily. It's very simple. The solution is ComfyUI Manager. I'm going to use an extension program
called Manager to solve the problem. If you click on the link here now, You can see the related Git. If you just install this
Git in the extension, You'll be able to see this. There are two ways to install this. One is to go to the link we gave you. There's a way to run this git clone
on the custom node and install it. Another is to run the related part
through this batch file provided here. It's running. First of all, I'll show you. I'll show you how to install it with Git. In this document we just
gave you, Copy this part here. I'll run the related part in CMD mode. In this state, If you copy and paste the related part, The installation is over. Other than that, this site. There's a script related
to this installation. Go to this portable directory and run it. If you look at it now, Among the things we showed you earlier,
There was something called a root, right? There's a part we explained as a root. You can put a related
batch in here and run it. It's simple. But now I've already
installed Git like this. So what are you going to do in this state? If you look at it now, There's going to be a
CMD related to the ComfyUI. I'll close this first. I'll close it all. This is going to be an
error, so I'll close it first. All right. I'll try running it again in this state. Then it'll come up again. The ComfyUI will automatically install
the related extension, and it will run. It's all ready now. If you look at it like this, it's added. The manager wasn't there earlier. So if you click on this
part, you can see this menu. This is how we load the basic workflow. The route around the content
we made before will be on the way. We're going to make a
content called Toy XYZ Camera. I've shared it with you before. At that time, there may have been
people who had errors related to it. Of course, there are
people who install the git, but there are people
who don't know well. I just put in the related workflow. Why is there an error? I'm sure there were people like this. So, in this case, it's going to be
helpful for us to explain it right now. In this state, if you look
at it now, run this manager. Here, press Install Missing Custom Node. After pressing the
install related part once, I'm going to install it like this. After it's installed, press Restart once. In this way, you can see
the cleanly organized situation. So you don't have to be afraid
of this custom node-related part. Of course, if you go to
a site related to workflow, There's too much content. One of the things that
you might find a little difficult is there may be things
that don't work out here. Because now you have to go and get it. Or you might have to make it yourself. There's a downside to that. Well, if you look a little bit more
depending on your needs, You can solve it. The related node usage is like this. You can drag and drop and solve it. In the current case, there
are a few related parts like this. At this time, install the
Instant Missing Node. You can use it. In this state, I'll press
the Restart button. Then you can see the
results come out like this. Of course, the parts
related to the models here It could actually be a
model that I don't have. You have to check all the models. Don't forget that. If you look at it now,
it's just hidden like this. The node-related parts
are all connected in a line. So now I'm going to explain
the contents here one by one. So now I'm going to find
a node that wasn't there. But basically, I'm going to use
the Install Custom Node like this. We've been talking about this before. It's not just finding it automatically. We're going to search for related content. There are functions
that can install that part. So if you put the keywords in here, You can find these parts and install them. You can solve the errors
in this part you saw earlier. In this part, There are quite a few cases
where you don't know the route of the parts related to the
model that we need to install. There are also functions that
can be found and received here. Basically, Models that
are detected or Models related to upscale You
can check it out here. If you need it, I think
you can see it here again. And now, Parts
related to updates. Updates related to this ComfyUI. Parts related to the
entire update extension. Or Parts that receive updates
related to some patches. You can check it out here. Or We've installed
it as a missing node. If not, In the workflow
related parts, I'm here. I used a separate git to make it. If the git address is given
separately, You can enter it here. There's a way to install it. In addition, You can check out the information related
to the community manual and node info. And here, In the ComfyUI workflow gallery, I've checked it out before. You can check out this content. Check out the various workflows. How to use it and make it. You can check it out. And this is the part. I have to give you a design. If you look at it
now, I've installed the custom node. I just did. In this way. But, Now it's coming in here a lot. So what is this WAS Suite? What is this YMC Suite? There's a lot of stuff like this. If you look inside, There's another content. It's not like I'm basically providing it. So how to separate this. You might have a problem. At the time, In the Badge here. What Nickname? ID, What ID Nickname? In this way. You can see the relevant contents. So this is where this is made. This is the basic node. Custom node you have
to look at this and check it out. Without this, you don't know
which one will pane areas later. So you must use this. I explained that you can check this out by using a good Badge of
these relevant decades. Finally, You just look at the basic
procedure and finish it. One of the basic things we should know is One of the most important
things is how to apply the text-to-image and the
image-to-image process in A1111. I'll add the basic nodes
related to this to the PDF link, so I'll test the
related part right now. Basically, when you create it,
this is the text-to-image state. Now, the way to change this to
image-to-image is to change the latent image part of MT, which is
connected to the case sampler below. First of all, load the
image here, and it's done. In this state, VAE encode. change this to a latent
space in the pixel space. You can think of it as changing this
to a latent space in the pixel space. But what's important
is that you have to connect the VAE
related part here like this. Don't forget this. You can create it in this state,
but now the denoise strength is 1. If you look at A1111, there's 1 state when we go to I2I. It's too strong. If you want to go with this
state, you can put it in 0.5. If I press Queue Prompt, you can
see what you want slowly. You can see it. That's right. I'll keep turning this on purpose. Should I keep increasing it? Like this? It's fun, right? It's too much. I'll lower it a little bit. I'll fix the seed. If you keep going up here,
you can see where you can go. Or you can get a little
closer to the character. You can lower it like this. All right. You can change the image like this. You can keep changing it randomly. You can stop where you want. All right. You can use it like this. If you look at it now, what I can
explain is basically this road checkpoint. Call the model here. Here's the model. One of them will be
connected to the KSampler. Now that the KSampler is
connected, the model-related thing is over. And now this clip. This clip texture is
basically connected here. You can connect it like this. One is positive. One is negative. Negative is this way. Positive is this way. You can connect it this way. And this latent-related
thing is basically made. Whether you're going to use a
latent or an image-related thing. Whether you're going to mix it in here. You can decide. And if you want to use the parts related to
VA separately, you can use the VA loader. You can connect it like this
using the most used loader. It's simple, isn't it? You can also connect it separately here. Now, the important thing here is this. When you enter the KSampler standard, when you enter the image,
you have to encode it. And when you take it out
of here, you can decode it. You can save it or save it. You can connect it to the preview image. And if you don't want
to use it, you can mute it. If you press Queue Prompt again like this,
you can get a preview image. You can enlarge the
image like this and see it. It's very good. Basically, I want to go back to T2I. Then you can report the Load Default. Yes, that's good. This is the part related to I2I. The node you just saw was organized. I'll organize it and connect it and use it. I'll give it to you. The color here is related to I2I. I explained it earlier, but
I want to separate this. I want the positive to be blue. I want the correction to be something else. You can put the color in here like this. It's simple, isn't it? You can use it like this. Yes, that's all for today. Next time, there's a lot
of stuff you've asked for. I'll organize it step by
step and show it to you. Thank you so much for 1,000 subscribers. I'll come back with a
better look in the future. Don't forget to subscribe and like. Thank you.
