I have created a video for those who are starting with Confi UI for the first time. A1 qu Confi UI provides a UI package that allows everyone to use Stable Diffusion. The concept of using Stable Diffusion is the same, but there are pros and cons for each in terms of UI. When you first see Confi UI, it may seem quite complex, but once you understand it a little, it is really easy and simple. Even for those who had no plans to use Confi UI, after watching this video to the end, their perspective on Confi UI may change a bit. First, let's install Confi UI. For those who use Stability Matrix as explained in the previous video, you can simply add Confi UI within it, and that's it. Alternatively, you can install Confi UI separately. Here is the official Confi UI website where the installation instructions are provided. If you are using Windows, there is a portable version available. Just download a file, extract it, and you're done.

You can also use it only with CPU instead of GPU, which seems to be good news for those who couldn't use stable diffusion due to the weak GPU performance. However, the speed with GPU will be slower. Click on this link to download the ConFi-UI package file. Unzip the file. Note that the compressed file is not a G file but a 7Z file. If you don't have a program to unzip the 7Z file, you can download and use it from this link. After unzipping, when you go into this folder, you will see these files. There is a file here that is very important to read. Let's open and read it. If you are using an NVIDIA GPU, run the NVIDIA GPU batch file, and if you are using CPU, run the CPU batch file when running ConFi-UI. If you see various things in red when running ConFi-UI, check if there is at least one checkpoint in this folder. It says that if there is no checkpoint model, of course, it won't run. The most basic is SD 1.5.

You can download the model here. To update the Confi UI, it is said that you should run the update Confi UI batch file in the update folder. Also, when there are other stable diffusion packages besides Confi UI, for example, if you are already using a11 and want to install Confi UI additionally, it is recommended to share the model files in these two packages to avoid wasting space. Let me show you directly. Go to the Confi UI folder and delete the end of this file. By doing this, it becomes a YAML file. This file is read as a YAML file. Open this file with a text editor. Add the A1 folder name to the base paths inside. Since I have a11 like this, I will copy and paste the folder address. After doing this, starting from the folder where the checkpoint models are, including Vera and embedding, will be included directly below.

The folder containing the model is used. The default folder names given are already correct, so there is nothing to change. If needed, you can change the folder names. In particular, if the folder name of the control n model is different, you can change it here. Don't forget to save the file. In the confi UI folder, when you run the run nvdia GPU batch file, the confi UI will start shortly after. If the web UI does not start automatically, you can type this address in a web browser. It's the most basic screen. If you have already tried a1l, you will feel that the confi UI starts really fast. It is known that the confi UI is much lighter, faster, and uses fewer resources compared to a11. Pressing the queue prompt here will create an image. Now, let me explain the basic structure here. I will explain it compared to A1. If you are familiar with A1, you will understand it easily. If you have used a1l before,

Even if you have never used it before, you can easily understand each one shown here is called a node. The entire UI is composed of nodes. On the far left, there is a checkpoint node. You can select the checkpoint model to use from here. Currently, the Aniverse model is selected, but you can change it to the desired model. In the case of a11, selecting a checkpoint model here is the same. There are two clip text encode nodes here. One is for providing prompts, and the other is for providing negative prompts. In the prompt, it requests to draw a beautiful glass bottle landscape and a purple galaxy bat. In the negative prompt, it requests to remove the text watermark. There are many connecting lines here. Where the line connects to the left of each node is where the input goes into this node, and the part on the right is where the output comes out of this node.

So, in the Road Checkpoint node, this model is selected, and the clip information of this model goes into the Prompt node on the right. This prompt goes as positive for the K-sampler, and this prompt goes as negative. The model to be used by this K-sampler is connected to come in from this node. This node provides image information, and the current image size is 512 * 512. If you change the image height to 768, the generated image size will change. This flow is the actual process of stable diffusion, and if you use the PUI, you can directly configure that process. Giving prompts and negatives here, setting the image size here. As seen so far, just setting what you want in A1, you can configure the process directly. Where each setting information comes from and how it goes to where in the UI in A1.

You can see directly if it's flowing in. At first, this may seem a bit complicated, but you will quickly get used to it. Once you get used to this, you can see the flow of stable diffusion a, so you can know more accurately how to set what you want. By using Confi, you will realize that this node method allows for a much wider range of settings. Information coming out of the Load Checkpoint node goes into the Decoder, and the results from the K Sampler also go into the Decoder. Therefore, the Decoder generates an image. The generated image goes into three Image nodes, is displayed on the screen, and is saved to disk. Looking at the K Sampler node again, you can also provide a seed value here, decide whether to randomize the seed values, or keep using the same seed. You can set the steps to take, provide cfg values, select a sampler, and set the denoising value.

You can provide the same value at a11. You can connect the nodes as you like. Currently, this is being used as a positive and this is being used as a negative. I can change this for you. I will revert it back to the original state and you can give a name to the node. When you right-click, an additional menu appears. If you select the Properties panel here, it will show like this. You can change the name here. I will provide a positive prompt and then a negative prompt. There are many useful features when using the Confi UI node, and I will show you just a few simple ones. When you right-click, an additional menu appears. Let's add one more key sampler. It will share the same prompt as the one above and share the same image node.

I will attach a decoder to create an image for this sampler. You can attach it from the additional menu or copy and paste the decoder node by pressing Ctrl + C and Ctrl + V. You can share the decoder from the same checkpoint model that is already in use and connect it. Then, connect the sampler created earlier. To save the generated image, attach a Save Image node. By doing this and pressing the Queue Prompt button, two images will be created with the same settings. The reason the images are different is that each sampler used different seeds. Now, make the seeds the same, fix them so they don't change randomly, and create the two images again to make them identical. Create another Late Image node and make it longer this time. Connect this node to the second K-sampler, and with this setting, you can create a 512 * 512 image.

When going into the sampler and setting the 512 * 768 image, it goes in with the key sampler below. When you do it like this, a longer vertical image is created below. However, the image above was not recreated. It's one of the advantages of Confui. Just now, an image was already created, and it's meaningless to create the same image again with completely the same settings. In this case, Confui does not create unnecessary images. So, it only creates a longer vertical image below. We will change the settings to 30 for the steps below and 25 for the steps above to create both images again. A green border appears to show which node is currently running. By using Confui, you can visually see the detailed content of the workflow. Both the images above and below have been newly created.

This time, we will create images using a different checkpoint model for the top and bottom. Add a Load Checkpoint node, then select a different model you desire. If there are many nodes and it looks cluttered, you can press Ctrl+C on the keyboard to minimize the selected node. Copy the Prompt node and Negative node. Connect the clip information of the Checkpoint node to the Prompt, then connect the model to the Key Sampler. Also, connect it to the Batch. By doing this, when you press Queue Prompt, new images will be generated vertically. The image above is created using the Aniverse model, and the image below is created using the Reanimated model. If you wish, you can easily change the node names for distinction. In this way, there are separate groups for generating different images above and below. You can group each of them together. Right-click to add a group. By expanding the group, you can place them inside.

All nodes that go in are automatically included in that group. Create a group below as well. This way, two different groups, one above and one below, are created. In this way, you can create a unique configuration workflow. You can share this workflow with others or reload it later for use. To do this, you must first save this workflow. When you press the save button here, it will ask for the name to save. If you give the desired name, it will be saved under that name. By pressing load, you can reload the saved workflow. To share with other users, you just need to share the saved file. By pressing load default, it reloads the default workflow that appeared first. Press clear to erase everything and start from a blank space. This image is created using AI. When you drag this image, the workflow settings used to create this image appear. If you want to recreate an image created by someone else or want to know how.

When you wonder if it's created, dragging the image will bring up the workflow. By using Compy UI Manager, you can use Compy UI more effectively. Go to the Compy UI Manager corporate site, click on the download button here to download the file. Unzip it, copy all the files inside this folder, and paste them into the custom nodes folder inside the Compy UI folder. Do this and restart Compy. A manager button has been added here. In the install custom nodes, you can install additional nodes. Think of it as similar to A1's extension program. For example, if you type animate, it will show the animate node that you can use in Compy UI. When you click the link to open the site, it will look like this. When you load a Compy UI workflow created by someone else, the custom nodes that person used will not be available to me.

You can do that. In that case, if you press the Install Missing Nodes button here, it will show me nodes that I don't have and install those nodes for me. You can also update by pressing this button. In the Confi UI Workflow Gallery, you can download various types of Confi UI workflows. If you are curious about how other people have created various images or videos, you can download and try the workflow you want. In this video, only the basics of the Confi UI were explained. In future videos, we plan to upload UI content together whenever there is an opportunity.

ComfyUI를 시작합니다.

안녕하세요? 디지클로 주인장, '인장' 입니다. 체널을 방문해 주셔서 감사합니다.
AI, Stable Diffusion, Automatic1111, Forge, ComfyUI 그리고 이미지 영상 제작관련 Adobe등 다양한 기술/방법을 설명합니다.



ComfyUI를 시작합니다.

ComfyUI를 시작합니다.

스크립트