Unlike the words I brought, you can see that it is written in text form. Hello, nice to meet you. Today, I will explain about two buttons that you can check once in Image to Image. If you run Stable Diffusion like me, you can come to the Image to Image tab. But unlike Text to Image, Image to Image has two other buttons next to the Create button. Interrogate Clip and Interrogate Deep Blue. So, let's take a quick look at what these buttons do. Interrogate Clip and Interrogate Deep Blue are similar to the expansion function I explained before. If you go to the Tagger tab, When we select an image like this, I explained that it is a tagger to analyze this image and select the most similar tag. So, I explained that it is written in the tags that collect the most likely tags by analyzing the image on the left. So, based on this, we can move it to the prompt when we use Image to Image, Or you can use it in Text to Image. But you can use it without a tagger. That's Interrogate Clip and Interrogate Deep Blue. So, what's the difference? It's simple. Interrogate Clip comes in a sentence form. So, upload the image, analyze it, and then write it in a sentence form on the prompt. The Deep Blue below is the Deep Blue that comes out in the form of a word, just like we used in the tagger. Then I'll apply it. The way to use it is to upload a picture to the Image to Image. You can select a random picture and just press this button. Let's press the clip first. But if you press this at first, it can take a long time. It takes time because you have to download other models according to this clip and deep blue, as if installing another expansion function. So, if it takes a little time, take a look at the prompt window. If you look at the prompt window, you can see that it is downloaded like this. After this is over, analyze this image and write it in a word form or sentence form. Yes, the model is downloaded and I'm done analyzing. The time this model is downloaded and analyzed can vary slightly depending on the graphics card and PC specifications, so please note it. Now, if you look at it, you can see that the form is written in a sentence form, unlike the words we just brought from the tagger. A woman is standing in front of the Eiffel Tower Tower in France. In this way, you can see that most of the tags are created in a long sentence form. Now, let's take a look at the Deep Blue below. Deep Blue also takes time because it takes time to download and analyze the model when you first click it. Let's do it once. Deep Blue click Now, then, Deep Blue also downloads the model like this, and then starts the analysis after it is downloaded. Interrogate Deep Blue also downloads the model like this and checks the result value, so you can see that this prompt window has changed. When I just clicked Interrogate Clip, I could see that it was not so long and was not a word, but a sentence form. Looking at Deep Blue, you can see that it is almost a word form by cutting one word at a time. So, depending on what image you create, it can be an important point when entering this prompt. Some images may be better in sentence form, and some images may be better in word form. For example, the image of something moving seems to have a better sentence form, but the image of just a fixed feeling seems to have come out better when the word form prompt was made. Of course, this is a little different for each person, so I think it would be better to try it myself and feel the difference. Like this, the difference between the Interrogate Clip and Deep Blue is the difference between the words and sentences. You can think of it as the difference between the sentence and the word that analyzes this, and it is quite similar to Tagger, but the biggest difference is that there is no other example when using the Interrogate function. It just analyzes and tells you one result. In this way, it just throws a prompt like this, saying, I got this result, so use this result. But Tagger is a little different. Tagger shows an example that these tags came out with the highest probability among these possibilities, and it also shows various examples that there may be words like this even if the probability is low, such as 0% or 1%. So I think it's important for you to try it and find the best extension function for you. That's it for today's video. I'll see you next time with more useful content. Thank you.
