Hello, today, it's been a while since I've uploaded a video. It's about comfyUI. Controlnet became available in ComfyUI some time ago, and another tool called Control Lora, which abbreviates this and can be used effectively, has begun to appear. Developers seem to be really grateful. You can use this with the SDXL model, but it doesn't work with SD webUI yet, right? I know it's not working yet, but I'd like to briefly explain how you can easily install this through ComfyUI and how you can use it, starting with installing ComfyUI from the beginning. Now, let's go right away. Then, let's start installing ComfyUI right away. I'll leave a link, but if you go all the way down, you can download this file. There are also shortcuts like this, so it's good to know a shortcut as well. There’s a direct link right here, right? If you click here to download, it will be downloaded. The file has been downloaded here. Now you need to unzip it. Go to Show Additional Options and unzip it with Bandizip. It has been unzipped like this and I will move it to an appropriate drive so I can use it. I will cut this so I can move it to the C drive. Once it's moved like this, all you have to do is run it. It's so simple. If you go here, there is now a batch file. If you use an Nvidia graphics card, you can use this. You can run it like this A cmd window will open and the run screen will appear automatically. It looks like this. The ComfyUI screen that you can use will appear right away. The big advantage of ComyUI is that it is so fast and you can customize what you want and create a work flow. And it certainly seems to take less load on the graphics card, so there is the advantage of being able to use it more comfortably. It may seem simple now, but once features are added, it may seem complicated again. I think that this is a problem that will be solved as you gradually get used to it. Then, we will learn how to install the Control lora that I mentioned earlier. I think Control net is a very useful extension program for posture control or any other form of control. This is the Depth model. Through the Depth model, an image can be drawn in a certain form. And Canny picks the outline, so images with these outlines are created. Next, the sketch colorizser is a black and white sketch like this, and it becomes a controllet model that is made in various colors. In the case of revision, you can think of it as a control net in the artistic field that can create a variety of styles by combining two styles of paintings. And now, if you go back and look at it, the Control Net is actually a very large capacity for the SDXL model, about 4.7 GB. If you receive it all, it will be a bit of a burden, so I was actually starting to hesitate, but fortunately, I was well trained in the Control Roller. It was reduced to 738MB so we could use it. Since the functionality is not significantly reduced, I don't think there will be any major problems due to the reduced capacity. If we go into File and Version, we can receive the files we want. If you look here, there are two. In the case of 128, you can think of it as half of the 700MB. Then, in the case of 256, the file becomes about 700MB in size. I guess I can get this much. So, you can take the four in turn like this. I'll accept it like this After accepting this, we now have a workflow. I briefly mentioned this earlier, but since these work flows are saved, if you take them and use them right away, you can easily adapt to them by using what someone else created without having to create any work flow, so it can be seen as a necessary task. I will also accept this one by one. I accepted it like this, and then there was something called revision that I mentioned earlier. Would you accept a case like this? You can accept it like this In this case, the capacity is a bit large. You can accept it. Since the Json file is something that we can now show as a work flow in ComfyUI, I will explain this again later. Yes, you can receive it like this and use it. Later, you will need to take the files you received to the appropriate location. Yes, I'll go right away. If you look on the left, you'll see the ComfyUI folder. When you go here, I'll take a look. Now, this is the executable file, and it goes into another folder called ComfyUI. When you enter, there is a model and a part where you can install extensions, as we can see in the Stable Diffusion web UI. Here, Costum_nodes can be thought of as an extension program in the web UI, and the model can be thought of as a model of the web UI. Then, here’s what I received earlier. I received these four models called Control Rolla. This is how I got it Where should I put what I received? If you look here, you can see it here. There is a control net You can go here and move it as is. Then it came right here. In this way, the transfer work is completed. Then, in the case of the revision from earlier, you can put it here. I have a clip vision. I think you can just put this in the clip vision. In this way, the process of inserting it into the model is completed. Then, I received the json file earlier. These are the json files. Now, just insert them into ComfyUI and the work screen will appear. Yes, I'll show you right away. Now, this is the ComfyUI screen. When the screen is displayed, let's move this json file. Let's move this Laura's Depth work screen. If you move it like this, the work flow will appear like this. The advantage is that you can get the desired result without having to write anything by just inserting the model and inserting the prompt. You can use the JSON file like this. Now, what do we need to do to properly utilize this feature? I didn't insert the model right away. The basic model has not been inserted, and only the model called Control Roller has been inserted. That's why the basic model, that is, as written here, is not currently inserted. You need to insert the sdxl base model. There was a base model and a refiner model. You can also get this from CivitAI here If you search for this, it comes up. If you search for sdxl, it comes up here. The 1.0 base model is right here, and you can get it like this. Right-click and click Save Link. Here, it will pop up. If you look closely, you can just connect directly from the C drive and go in. The ComfyUI folder is here, right? You have to go in here and keep going and go in and enter the model, right? Where do we go into the model? Yes, this is the checkpoint model right here, so you can just put it here. You can receive it by putting it like this. Likewise, you can receive and insert the refiner model in the same way. The file will be there Not from CivitAI, but you can search for it on Google and get it from the place that pops up here. This part was covered a lot in the last video, so I will skip it. And when I get here, I'll try it again. Now the checkpoint model is in here. In this way, you can check that the two received items are displayed. And then the important thing is from now on The process so far does not actually apply to those who have already installed it, but it is explained under the condition that it is your first time using it. Then, I inserted the basic model like this and also inserted the Control lora model. From then on, it's an important process, so I hope you'll pay attention and listen again from here. It was quite easy to install extensions in our StableDiffusion WebUI. However, it can be difficult to know what to do to install an extension in ComfyUI. There is a program that can help you with that. It is a program called ConfyUI Manager. Let me try this out Now this is the screen. Now you need to install this. Here's how to install it: Earlier, I pointed to Custom Nodes and said that they function like extensions. To put it simply, yes, by the way, if you go here, git clone, and enter this address, the installation will be completed right away. It's really simple. Let’s try it together. From this screen, go to ComfyUI and go back to custom nodes. Go here and click at the top. I'll press cmd. I'll press cmd. What did I tell you to input earlier? I just copied this phrase from here and there is a CMD window here, right? Just copy and paste it here. In this way, a program called Manager is installed. Now, you have to turn it off and run it again to see the screen. Yes, then I will turn it off and run it again. Let's go ahead and do this again to add the execution screen. Now, I opened it like this, and a warning window like this pops up. A red warning window pops up, so you have to pay close attention to this part. It says that the following cannot be found when loading this graph. They say they are using a depth model, but there is no preprocessor. The lack of a pre-processor means that if we think about the control net in the web UI in detail, the pre-processor and the model work together to create an image. This is also true here, so it cannot function properly because there is no pre-processor. That's what I'm talking about So what I want to say now is that what we need to do next is to get a preprocessor. To do that, I installed this manager that I mentioned earlier. Now, let's click on the manager. When you click, this will appear: Here we click Install Custom nodes again. Then you can see that there are many different programs like this. By installing everything we need like this, we can conveniently implement the functions. Yes, then shall we type pre to pre-process here? When you hit Free, the first time you play is a Free Processor, and the second time, another Free Processor appears. It is usually better to accept the second one, so if you click install, it will be installed. Even though ComfyUI looks complicated, in some ways, installing it is very quick and convenient, so this can be seen as refreshing. Okay, I'll close it now. Yes, then to re-recognize this, close it again and run it again. Now, I'll upload it here again. Ah, this may not work because there is another CMD window open. In this case, you need to close all existing CMDs and run it again like this. Then, the preprocessor installation is now progressing like this. If the process is successful and the installation is successful, you will see it correctly on the operating screen. Yes, then I ran it again and now the warning window does not appear, which means it was done correctly. Now, we need to insert the controlnet model here again. Of the four we received earlier, since we are using Depth, select the Depth model and then load the image. Click on it and import it. Yes, I'm going to create a Depth model with this. Probably because it is a preprocessor, if you return this, the analysis will appear here, and the final image will be created like this. You can change the rest of these settings, but in this case, the settings are in good condition, so you can use it as is. This means that the workflow is good. Shall we just change the next prompt? Since it's a woman, let's change it to Woman. Let's call it Dress. As for the rest, let's just create it like this. But the important thing now is to check how it is created, so now press Queue Prompt. This is how the work is progressing. Here, the depth model has been created like this, and the progress is like this on the next screen. When the (border) color changes like this, you can tell what stage is in progress. I think it's almost done. Yes, we're now in the sampler and work is progressing. And I'm telling you, if you're curious about what this is, you can right-click here and press Collapse to check what function this is. In order to see it briefly, I am minimizing it through Collapse like this. Now, the depth analysis is not accurate based on that kind of perspective, so I don't know how it will come out. It was completed like this. Uh.. Okay, these poses are roughly similar, right? Similarly, it was completed like this, and we confirmed that SDXL can sufficiently create images using Controllet. This installation process may seem a bit complicated. If you repeat this a few times, you will understand why you need a manager. After that, if you look at each item one by one, you will be able to realize once again that there is not much difference in the end compared to when using the web UI. And you got what you received here earlier, right? If you want Canny, you can insert Canny json like this and create another picture according to the work flow. If you want to do something other than the SDXL base model, you can insert another model and do it. CivitAi has many SDXL checkpoint models for masks, so you can get them and use them. Also, there continues to be news about ComfyUI. We will be able to do additional video work related to this. And I would like to tell you one more thing, and I will tell you about that later. So, if today’s video was helpful, please like and subscribe. I'm a bit busy with my day job these days, so I can't make videos, but I'll do my best to provide you with information. thanks for watching
