Hello everyone, today we will talk about AI painting. The ComfyUI you see now is also an open source project . This project has released a video before, which talks about the content of real-time painting. And the effect that ComfyUI can achieve is far more than that. There are still many effects that can be achieved. Considering that more content will be involved in the follow-up , I still want to introduce some related content of ComfyUI to everyone . Then this must first be downloaded and installed. Then there is an installation here. You can click this. Then when you come here, there is a link. Just click on this link and it will automatically download a very small 1.4G size. You can use it after downloading. I have put a package here for you. You can use it directly. You can also download this package by yourself on this web page . Then we will see how to install it. Just click on the decompressor to decompress it to the current folder. As you can see, we have now decompressed the file, and then this file is opened. Let's update it first. There is an update folder here . Just click on the first update. Then the second update involves other things. As long as we can run the program, the second update will not work. If necessary, we can only keep the first update. Then come back. Both of the following two files can be started. If we are A card users , you can use this CPU to start. Then of course we use the graphics card, which is definitely the best. Okay , then we double-click this to start. Then we will come to our user interface. There are some settings here. Of course, it is in English. Then the first thing we do is that we need to add a plug-in to make it Chinese . It is the Chinese plug-in we want to use. It can be converted into other languages . If we download it, it is very simple. Click this code and there is a download here. Then we download it to a file like this. Then we also give this file here. Everyone is ready. In this node, there is a Chinese version . This is what we downloaded. Then copy it and put it in this package. Then there is a folder here. Pay attention to the name. Then we put Just paste it in and then unzip it to the current folder. In this case, our Chinese plug-in is installed and this can be deleted. We can remove the following string of characters from it and then let's start the program again . You can see that on our side, it has been changed to Chinese and has been successfully translated into Chinese. Then it is more convenient for us to use it. There are also some settings here. If there are some settings, we will install more plug-ins here later. There are some related settings that we will mention later. Now let’s turn it off first and then here, because I have prepared a few plug-in nodes, which are the most basic plug-ins. These things can be done according to the method I just mentioned. Then put it in our corresponding folder. I won’t go into details. We want to do AI painting. So first of all, in addition to this tool, we definitely need models. We need a lot of large models, small models , and others. The model is then used to achieve our painting effect . So how do we download the model here? After everyone downloads the model , it is still the same as other painting tools. After the model is downloaded, it is still placed in this model folder. This folder is fine. Everyone should know about enlarging the model , and then some corresponding small models and some other models VAE models are all in this folder. You can download them and put them in because we may have used the SD and painting before. This is the tool , and we have downloaded a lot of plug-ins in it. If we download it repeatedly, it will definitely not work. Then if you copy the file again, it will actually waste space. So what should we do? This provides a new usage : We can use the model here directly and there is no need to download or copy it. You can see that there is a file here. Let's remove the latter one and then go here . Then we double-click to open this file . I see that there is a setting here that allows our ComfyUI to share the model here . Of course, first we have to change this address. Then this address is based on the file here. Then we take a look at this file and copy this. Just put this address and then paste it here and then save it. You can see that the model below is this big model, right? It is this address and then add this address and it will be a position of our big model. Then the same is true for the following. Then you can follow the location of this folder on your computer and copy it. In this way, we can call the model in this SD folder through a combination of two addresses , so we don't need to repeat it. Download and there is a ControlNet ControlNet below. Please note that you need to check whether the model involved in the ControlNet related to your file is in this folder. If so, then you do not need to change it. If not , then you can adjust it again. Everyone must pay attention to this position of yours . This must correspond to one -to-one. This is a common usage of our model. In this case, up to this point, we have all the installation and use of this and some of the things involved. We have finished talking about the use of this tool. It seems relatively simple. There are not many things that need to be changed. There is only one file like this and nothing else . Then we start our program and now it comes in. Because we just installed the Chinese plug-in , we can see it here. If you add other plug-ins, If it is installed , then we can click this gear to open its settings interface. Most of the options in it remain the default and there is no need to adjust them. Then what about here? There is a language , and you can choose the language you want. Generally speaking, we use Chinese. Then there is an image panel position here. By default, it is at the bottom, which is the line below. What it means is the image you finally made . It will be displayed in the column below . Of course, we can make changes. For example, we can change it to the top or to the left. It depends on which direction you like. Then we can change it to which direction. This is actually all our settings. After setting them , here Another very important thing is the ComfyUI manager , which involves the installation of all nodes and the installation of some models, which are all here. We just mentioned that for the installation of this node, we can go directly to its web page. After downloading, put it in the corresponding folder. If you don't want to download there, you can also download the installation node here. It will automatically detect the database and open the installation interface of a node. Then you can Search here yourself. This is also a way to install nodes. Let's wait a moment. Opening this library requires an external network. For example, for the plug-in we just installed, we can copy its name and then we can do it here. If you search it , you can see that you have found it here, right? It’s the plug-in we just downloaded. If you don’t download it on the web page, then you can install it here. You can just click Install here. Then the same thing below If there are missing nodes and you have installed the model, you can download them through the method here. What is a missing node? Sometimes we directly drag in the workflow shared by others and I will find one for you. You can see that there are so many nodes and I have not installed them here . This is After the missing nodes are reflected , your side is red. Let's take a look at it. There should be this color here. Everyone sees that this color is red. It means these nodes are not installed here. Do you want to install them at this time? If so, just click here to install the missing node. It will automatically detect what is missing on your node and then automatically detect it and then install it. So this is a very convenient function. This is something that everyone needs to understand. Then the following Let's make a simple example. Let's make a picture. The process of making a Vincent diagram. First of all, we definitely need a model. A large model. Here, we need a loader. As you can see, we have a loader here. It is used Let's load our model , and then we can choose a model here, which is the model on your machine. If not, it means that the path of your file is not correct , and you need to check it again. This is because we first load a large model and then After the model is loaded, we definitely need prompt words and then use this clip text encoder to have a positive prompt word and a negative prompt word. Then we have two. The top one represents the positive and the bottom one represents the negative prompt word. Then this is what we Next, what should the prompt word do? We should enter the sampler and connect a sampler to the positive prompt word. We also connect the negative to the node and connect it here . Then the model is directly connected from here. Connect it over because we definitely need a large model, right? And what does this thing do here ? We need to provide the sampler with a most basic image because this is the overall AI painting. We pass this text two Text to get the final picture. So what about the text prompt? It requires a blank canvas. If you don’t have a canvas , then what should we draw on the content of our text ? So here we first have to give The sampler has a blank picture , so we click on this to pull it out . If there is nothing here, we have to find it here. We need a blank canvas now , and then we give him a blank canvas. The width and height of this canvas are batches, which means that several pictures will be produced at a time. Meaning , then we follow the most basic 768, the most classic size. I just searched for this node for a long time and couldn't find it . So is there a quick way to find it? There must be one. For example, we can find it in this blank space. Double-click your left mouse button. You can see that we can open a search box and then you can search for your node. Of course, you need to remember some simple names of such nodes. For example, ours is this one, right? There is also an English one. We can directly enter this. You can see. We already have it here . Of course , it's no problem if you can remember this node. Now we provide a blank canvas for the sampler , and then we perform a series of activity samples in this sampler. After that, we need to decode the information of this image. Look at our decoding node . Then we finally output an image. We can click this to save the image. We can also preview it directly because if we say we don’t want to save it, then this You can preview directly here and you can also save it. If you save it, it will be saved directly to this out. This folder will be saved to your output folder. All your pictures will be there. Now, for the sake of simplicity, we will not save them. Let's just preview it directly. Then let's click Run. How to run? There are two methods. The first method is to directly click on the prompt word queue or the second method is to use the Ctrl key and the Enter key on your keyboard to run. Let's run it. You can see that now there is an error message saying that the VAE model is not mounted. If you are not connected, it will prompt you. So it doesn't matter even if you connect it wrongly. Then we can connect it to the VAE node of the large model. Large models usually come with this VAE model , so we can just connect it directly and let us run it. Well, if you look at the process, it starts here and then goes through this and then enters the sampler and then decodes and then outputs this picture. It comes out. It is so simple. The overall process is roughly like this.
