 One of my colleagues is giving his third talk today. So I think he gets the crown for most talks at Price News and Group Meetups this whole year, maybe even in the history of Price News and Group Meetups. So he's going to show us 30 things that we can do. Let's have a round of applause for Aisheng and see what we can do. Hi, everyone. My name is Yixuan, or YS. And then I'm going to be talking about 30 things we can do in Titan. As you guys probably know, Titan is very famous for two things. One is server-side code. And the other is working with data, like data work, data analysis, data visualization, all this data science. But is that all we can really do with Titan? I don't think so. So we're going to look at what else we can do with Titan. The format here is that for each thing, I'm going to go through exactly what our target is, what we are going to do. And then I'm going to show you the code. We're going to try to run the code if we have time. I'm going to highlight the key functions which we are executing for each task. And then I'm going to say the task is finished. And then we'll move on to the next task. Also, before I start, can I get a show of hands who here is relatively new to Titan? Oh my god, it's a very, very, a lot of quite new faces here. So very good. I'll try and keep things slower, not slower, but simpler. If you don't understand the code or anything, just think about the objective and what we are doing in each thing, what we input and what's the output. OK, let's start. So first thing we're going to do is, OK, so welcome to the world of Titan. So we're going to start in the plains of web development. We're going to make our way across the mountain range into the large lands of data science. As you can see, the lands of data science is the largest area in this world of Titan. It's rapidly expanding. From there, we're going to move on along the coast down to the province of NLP. And NLP is, of course, natural language processing, which is basically trying, computers, trying to make sense of language, which is basically just strings to them. And from there, we're going to sail east towards the Isle of Chet. The Isles of Chet, they were going to move north past the Peninsula of App Development, the islands of Images, the domain of physical locations onto the other lands of the miscellaneous tasks that we can do in Titan. And we're going to end our merry journey across the mountain ranges again at the volcano of Mezzan, the things that you can do with your Python code in Python. OK, so let's begin in the plains of web development. Great. OK. OK, this is not looking exactly like how it worked on my computer. OK, so this is Instagram. Instagram is powered by Django. And Django is based in Python. And what that means is that Python is determining what posts get shown to you as you scroll down your feet and what happens to your images when you post them onto Instagram and everything about your account. Everything, the logic is handled by Python. OK, so how do we build our own Instagram? I'm not going to do that in one minute. I will show you of that. But here, we're going to start a server. There's two libraries that we can use to do a web server side code. I'm sure many of you know. There's Django, and then there's Flask. So this is Flask. At the bottom, at the top right here, this is our code, is our actual code. At the bottom, there's a terminal here, which I'm going to make it bigger. And then we're going to do a Flask run. By doing this, we have now run this code. Basically, the important thing here is this thing here. So the app.root saying that whenever someone goes to this root, root is basically like a URL. When they go to this URL, we will serve them some data. And the data here is Slytherin Slytherin. I forgot to mention that as many of our examples here will be snake themed, because it's Python after all. So we can see our data that we served at $5,000. And you can see the Slytherin Slytherin. OK, so that's the first thing done. That's how you make a web server in Python. Second thing is we're going to do data science. We're going to go into the lens of data science. We're going to predict the survival rate of someone who has the same profile as me. If I was on the Titanic. As you guys may or may not know, the Titanic was a situation where they had lost people. The ship was sinking, but they didn't have enough lifeboats. So some people had to die. Some people had to die. So who did you shoot to die? So let's figure it out. So this is data science. First thing we're going to do, main things we do is import NumPy, a pandas, as Klan. I won't go into exactly what they are. So here we load in the data. But before I talk about loading in the data, what is data science? Data science is basically answering questions using data. And usually this comes in the form of making predictions or judgments. And first we have to feed them data. So here we're going to feed them data. And the data here is in a CSV form. Even if you're not doing programming, you know what a CSV, because it's like a lousy Excel. It's Excel without formatting. So if we feed in the data here, what does this data look like? Now we can see. This data is each row is a passenger. Then we can see all this data about them. We're going to use this data to figure out who lives and who dies. Then we manipulate the data a bit here, ignore, ignore, ignore. The most important thing here now is we're going to create a classifier. This is like a model. What are we going to use to? What type of model are we going to create to predict who lives and who dies? So this is a decision tree classifier. There's many types of classifiers. We're going to use a decision tree classifier. Then we're going to fit the model. This is to train the model with the data. And then we're going to use the model now. So let me quickly load all of this, run all of this code. This is a Jupyter Notebook. It's the same as what the earlier speaker was using. So we're going to use the model. CLF.predict is a key function. This is predicting a person with these stats. These stats are basically my stats. I'm 30 years old. And then all these things are what class I would have bought. So it says zero means I would have died. But we can do better than that. We can predict the probability at which we will that I will have died. So the probability is that I would die at 60% chance. So you say I probably died. So then we can also save the model we predict survival using data science and Python. Let's go to the next thing. Facebook improved upon this. They said that we can. They may even easier to predict things. But they make it easier to predict time value data. Data that is along a time series. So here, they developed this thing called profit. So we import FB profit, profit. Then we also load in some data. What does it look like? It looks like this. Every month, there are some sales. These are the sales of snakes. Not very ethical, but OK. Then we have the shape. This is how many rows are there. This is how it looks like. And this is 293 rows. So we have data from 1992. All the data predicts 10 more years of the data. So we're going to do that. And then wait for a little bit. And then it was really out there, didn't I? So it shows us immediately. With so few lines of code, it immediately shows us it predicts that this is the range of outcomes that the sales could tend towards in the next 10 years. So they made it extremely easy. Then after that, Google made it even easier by you don't even have to know Python. So that's what we have done to predict future sales using Python. Let's go on to the next one. It's using deep learning now. So deep learning is using neural networks, but it's also machine learning. All this is machine learning. That's our science thing. We create models to predict things, to predict things, or to make judgments and things. So here, we have a model. That has multiple hidden layers. So what does this mean? It's like, let's say you feel hot on a certain day. Does that make you happy or sad? But to a human, you won't really know. That does make you happy or sad. Maybe if you're in a sauna, hot is good. So you must know sauna plus hot, then it's good. So does this hidden possibility of sauna, which is like a hidden thing you must account for, there's a context there. And all these hidden layers can account for the context. So it's like kind of how our brain works. It tries to model how our brain works. But anyway, let's do Keras. We import Keras, I won't talk about CV2. We load all of this stuff. All this is not very important, not very important, not very important, not very important. So we can see model sequential. This looks familiar, right? This is like the decision tree classifier. It's just a different type of model only. It's called a sequential model. Then we compile the model and we fit the model on the training data. But then here we fit the model and it's going to train over all these cycles. And when it finishes, then it will be fully trained. You can see the loss here as it tends towards zero or it tends lower, then it will be more and more accurate. How this works, I will not go into it. Okay, so next, I'm going to have time for this nonsense. So we're going to load a trained model. So you can load a trained model. This is just a file. You can just load a model and immediately use it. So now in my folder here, you can see here. Oh yeah, sorry, I forgot to even tell you what we're even doing this time. We are using deep learning to identify handwritten digits. Handwritten digits. Okay, so some people write four, look like five, right? Also, people write five look like S, right? All these things we are trying to get the machine to identify it. So now I have this thing here. You see mystery number dot PNG. See, you all don't know what it is, right? Computer also don't know what it is, but computer is going to find out. So the computer is going to, CV2 is going to read the PNG and then using the model that is loaded is going to predict what the, is going to predict what the, okay, let me cancel this. Sorry? How do I cancel this? Okay, okay. Okay. Ah, okay, so it says, I think this digit is a four with 100% confidence. Then we can go back to our mystery number and indeed it looks like a four, right? So think about this. It took a number, a handwritten number, and it figured out that it is a certain digit. So it's pretty cool. Okay, so that's, we have learned how to identify digits with Python. Let's continue on. We're going to make our code, write code. Okay, so how does this work? This is recurrent neural networks. So recurrent neural networks are kind of like machine learning as well. The only difference is that the recurrent part means that they have a memory of the inputs that were given to them earlier. So you give them a string of inputs and then they have a memory of all the inputs which were given in the past, right? Or not all the inputs, but a number of inputs which were given in the past. This probably doesn't make sense, but it's just a brief overview. Okay, so this is very useful for strings. When you feed it a long, long string, it can figure out the, it has all the memory of string of characters that you fed it previously, gives it some context for what this machine learning model should do, okay? And of course, as I said, we are gonna make our code write code. So hope for the best, okay? No, okay, so here we're gonna go cancel this guy. So, okay, so we do train.py, right? Basically, I have a huge file of code. My entire code base is inside here. I'm not gonna show you because my boss would kill me, okay? So this entire code base here is gonna train, but of course I'm not gonna let it finish because then it will be like tomorrow already. So we are gonna do another thing which is sample.py, sample.py. And then sample.py, we have to tell, what does this mean? And extremely important thing I didn't mention is that this recurrent neural network is a generative network, which means that it tries to generate new output, which is similar to what you have seen before, okay? So it's going to generate to the train. I created some, I see some code naturally, but it's like, so you can see that it's trying to import stuff from the network, and then import some stuff. It's trying to import stuff that you don't like, but it also has to assign stuff and do all the functions. It's not too bad for a vending machine. It's not too bad, but of course, as you generate it, you get that, right? So that's really our code right code. So next thing we're gonna do is beta visualization. It's gonna show off beta in a nice way. And we're gonna do this with OK in Jupyter Notebook. Let's go back. You know what, we're gonna select close, close, close. Okay, close like this. Okay, so here we have, okay, we're going to import andus, and that's what it's going to have, a really like data-free like data. Then we import OK, which visualizes the data for us. And then after that, we import in the chart data. This is all the chart data that we're looking at. The rest is all styling and creating a chart itself. So this is, you can see the pivots, these are pivots, and these are the damage. We're gonna create a vertical bar chart. And these are colors that we use in the vertical bar chart. The important function is, this creates the entire chart, without the data you don't create each chart. And after that, you also need another one, which is the bar, this creates the vertical bars. It takes in your values, which are your top, is your value. So how high the vertical bar reaches is the top. Okay, then after that, we output Notebook and show it. Then we can show our nice visualization. It's on just the numbers above. So now we have visualized stuff, we can impact it. So now let's go on to number seven, which is scripting. I will talk about this in my previous talk. So when scripting is basically normal, normally in the whole style, you want to find out about architecture. You go into your main book. You click this one, then you click five seconds. Then you click the next one, then you click inside. Then you can really do it yourself. Then you go in, copy the address, copy the name, all these things up. Then you copy the company profile, all these things. So you can click this thing and then copy that in your next second standard. But then, now we're going to do that in half-new style. So now the new style right is, use scripting. So let's select the data that programically. And then you use a spider, the main file, and actually writing it is this one. So the green book spider, we start with this URL. So that is exactly the same URL we started with alone. It's going to go into every single link from that URL at the start. Now that's going to go into every single link from the next sub division. And then it's going to pass each of those HTML and find the related data that you want, the phone number, text number, the emails. And if you run it, it still looks something like, if you run it, it's going to look something like this. So it's going to go in its case and then it's going to call its website and it's going to find the links and it's going to go to those websites, then it's going to go to those web pages and find the links and then you wait for one day and you come back and have all your point of data will be in the form of, where the data will be in the back. This, right? So the data will be in the CSV, very nice deal. Okay, so that's scripting. So we have scripted data and link like that. Then next, we go to World Cloud. So we can visualize a lot of that and then we can do that. And it's called NLTU. I feel like it's processing. So let's see how we do that. Let's cancel this data. Okay, we cancel it again. Okay, so this is a World Cloud. So the important thing, by the way, I think that it's called World Cloud. It's not very creative. So World Cloud, then we're going to open the text. What is this text? I got it from a project, we have lots of free books because I don't want you to use like free books. So this is an entire book. And it's all about, you can guess it, it's all states. Right, so the states of Europe. And we're going to look in the entire book into this text variable. And then we're going to open a mask, which is the Python logo. Then we're going to set this. We're going to see in time what is thought, but thought was obviously back. The words in English, which are not very useful, and it's not very meaningful. We're going to remove all those. And then the rest of them, we're going to create a World Cloud. And we're going to show, we're going to start with a cloud in states.png. So let's do that. Actually, I'll be running if you don't know something like this. So you can see, you don't know a bit more. I cannot do it all. But it was to be like the backup over. So you can see the video on the side there, the words which appear along with it. So that's a World Cloud. So now we have visualized text in a World Cloud with Python. Next up, we're going to look at this, continue our nationality processing, but identify key phrases in a piece of a passage of text. So this is a passage of text that is from UTDM. So this is about states, of course. We're going to use a library called text.ac, text.ac, it's derived from another library called spacec. But they just have a wrapper on top of spacec. So this, okay, and then the key things that we're going to do here is, actually, there's not much to see here because it's already wrapping spacec. So if it's calling this module, we get the key terms, I mean rank, then in terms of significance, and rank means we're going to get phrases up to four words long, okay? And we're going to find which are the most in a significant basis. So let's run this guy, okay? So after I run this guy, it shows us this text, and then it shows us, so here's the text. And it shows us what are the key terms. Key terms, that's a key factor. There's more than they do, and it moves sensitively sectors. Okay, it's about all the nonsense, like have all the stop words, some of these things, they do not have which are the most important phrases, or significant phrases. It also tells us some interesting stuff, like the readability. The most interesting thing I think here is the fresh key, the grade level, which is basically saying, what age of a kid in the US would be able to read this confidence. So here it's something like a long drawn pen, which is around centuries before our Singapore context. Okay, so now we have identified the key phrases in the passage, in fact passage, from one thing at least, just a string to the computer, but we're turning it into an in mind important phrases. What are the phrases for the significant text? That's a good question. I don't even know. I will find out what you want to find out in the passage. So, then next time we can read that up, we can summarize the text straight from the HTML. Straight from the HTML, we can use this thing called SAMI. So we do SAMI, and then we talk about this stuff. So first thing we're going to do is get it in language of English. We're going to summarize it with five sentences. So the use case here is when your friends in New York, or your boss in some capital, but then you just throw it out and it's a damn long run. So you're going to get it, but then you're going to know it's not. So if you want to ask you, it's going to know that. So you, this is going to go around, you get this cluster. So this cluster will tokenize it, tokenize it, all of the words inside that. And then you, of course it has text, and then you do tokens, which are words or pockets of meaning. Then SAMI is like, this family here eating, the stem is it. So the key meaning is it. So you'll stem all the words, and then after that you'll summarize everything. You won't want to stop words. And then after that, you will give you a bunch of summarized sentences. Okay, so here, let's see what happens when we run that. Change. Okay, so then it gives us, still a bit longer, but at least, there's only five sentences, so you read. So it tells you there's two, if we send one of the professor on 10 plus a letter, and the other one, it's quite easy to read. So basically, yeah, it's easy to read. But it tries to cut out. So we have extracted, it's still some meaning from a long passage, using Python. Then next one we're going to do a chatbot. Chatbot is, you know, change so a chatbot is this one. Close this one first. Okay, so a chatbot is a library of chapter one, and you can import chatbot. First thing you need to do with chatbot is to train it. Right, give it some responses, else it's not very interesting to take empty. It's like a baby, you talk to the baby, but baby don't know what it's in. So here we have a chapter one of the trainers. If you're the trainer of chatbot, you can also train it manually with a list. So if you say hello, you can say hi, hi, and you'll see how it works. And then you can also train it with a Twitter data. That's what people read, learn about or reply to an input into that and read from that, okay? Rescribe just so I can do it. So chatbot, this is the name of chatbot, and then we train it. After training it, can we run it? Let's see what happens. Okay, so now it's going to train one. I don't think this one's quite fast enough. So we'll just bury with it. So yeah, it's a bunch of topics that you can talk about. So once the trainer will ask me for some input, I'll say, how are you? Then you'll say, how can I do well? What is the best way to do this, okay, thank you. So then, you know it's not enough. So then you can, so now we have a chatbot we can chat with. Okay, so next thing we're going to do is to make a user tweet faster than that or NLP, the comments of the NLP. NLP, NLP, and this is going to be one next level, I think. So this script is going to go into Twitter, script, users, tweets, and then after that it's going to mark all the final, which basically is like generate rendered tweets from, yeah, yeah, original tweets. Maybe some of these tweets, generate fake tweets from someone. Okay, so the main thing is, you know, we mark all five texts. So the mark all five texts is to create a random change, which will be like new tweets. Okay, so let's run that. This is going to be by loss. Number of tweets, five tweets. So it's going to get tweets from my boss's Twitter account and then trying to pretty much give examples of what he would tweet like. Okay, so it's just, we don't know how to use this NLP, it's sort of a work use thing, which I can identify with. Okay, then I'm going to get pretty good, pretty good, it's like something I've been asking. So we have many someone's tweets using pattern. Next thing we're going to do is to use Slack programming. I think we can use a Slack or outside of work, one person, two person, three person, three person again. Okay, but the rest of them don't use Slack. Or if you don't know anything about this, just listen. So we're going to import the Slack client and from here we use a token. We give it a token and this is the key part. So we create this SC, which is a Slack client. And then the key things here is the whole call. SC.hti call, you can post message in the channel info and you can add reactions. So I'm going to show you what the Slack is like. Currently the Slack is, yeah, it's just empty, empty, I'm going to run this guy, okay? And then he's going to say some stuff, okay? And he put up, up, and up, down and down, right? And then if you like it, you can have someone and go down and down. Okay, so that's what we mentioned. So it's just posting some random messages. Okay, but we think of that one and that one. We can create a Slack squad, which is what we're going to do next. So we have created the app, we've extracted the Slack programmatically. Now we're going to create a Slack squad, which is going to respond when you call a message key. Okay, so how do we do that? See. So now it's no longer happening on this video. I'm going to say like a snake's hide and see this. This is at least one of them and I'll come out with one because I've been here for 30 years. Okay, so this time it's not working, although it was working just now. I'm not sure I can't be back now. So I'll show you the code in a minute. So the code here is, so if you did some authorization, some authorization details, don't copy mine, then again you're using the Slack find, then now we're doing a Slack event adapter. So that it can respond to events. So if you message the robot, it will reply to you. So it says, if someone says hi, then it should reply to it alone. And then if you do a click on message, then it's actually hi. On message, you know, you will say your first message, hello. Hello, hello, say hi. And then if you do a click on message, not working, never mind, let's confirm. Okay, so that's creating us like what? Next thing we're going to do is create, we're going to move straight into the realm of app development. First thing is desktop apps. Let's talk about, we're going to create a custom work app. This is what it's going to look like, or just not what it's going to look like. This is the code. You can use time queue files. This can also be checked in. So time queue files, and then you will, it's basically a GUI library. How do you create graphical user interfaces for your programs, right? And then the thing about time queue write is that it has a lot of very convenient all of the box stuff for you to use. So there is a plainly text editor, which is kind of a non-packing, but now we're just going to customize it. I'm going to customize it by adding our own little button, which is going to say, gets made, gets made. And this is the icon, this is the icon, which is going to symbolize the button. It's going to say it's gets made, it's going to call an action, it's self-examined, and it shows self-examined, which is here. Right, it's going to insert something text, and it's listening. Okay, so let's run this guy. So we have to get fast, we have to make it go fast, we're really short of time. Okay, so this is our sleep pad, we can add in stuff to it, it gets more, it's more, if I try to do something, it'll see, it's listening. So we have a custom, custom desktop app, look back, where you put it in, so that's how you do that. Next thing I'm going to do is a look back at, so now right, desktop app is not the most profitable thing, the most profitable thing is a look back at, everything also can be put into the HDL on top. So we're going to do that. Okay, then we use a library called flex, and when you pop that's our X, and then we're going to define what text input looks like, text input has a name and a text, there's some input that you type into it, and then it has a name, okay? And then whenever you, then it's important also, whenever you say when we are done with this text input, if you update its text variable, it's set text, it's measuring things to this text, you update its text variable into your input text box value. And then after that we're going to create this thing with three of these as inputs one, two, and three, and then from there we're going to write up letter, letter template in, okay? So let's run this guys, okay, so this guy runs. So now we have this thing, this looks like the desktop app by the time, and we can export it into a HTML app and it still became the model, okay? So I'm going to add in the mission. So, so now it's built in this, that's the template and then you can complete this. And now we are created a web app within Python and this can be straight exported into a HTML, naturally, okay? They just convert Python into JavaScript, it's quite amazing, okay? So the next thing that we're going to do is create a visual model using VenPy. So I'm going to do a big game to do VenPy, here we're going to go and see how it's created. So let's launch the project first, okay? I'm not sure the code was there. So it's cool, okay? It's a, you know this thing is an RGY file, it's not a RGY file, so it's a RGY file, it also accepts Python, regular Python. So we see something that used to be a Polish stuff looking in there, but it also accepts regular Python. And then you define some characters and you define the model, you define some variables, and then you have a root drop-down model. And in your root, you can have a menu and two have different choices. And you jump to the table, and go to a certain angle, okay? So let's run this game and see what happens. So it's the Snake, it's that. That, it's that, please. It's that, it's there, yes. You can have options, So that's all. So that's all. We just created a visual novel using iPad. So next we will move on to the realm of the islands of images. In the islands of images, the first thing I'm going to show here is this one. Okay, so now we're going to manipulate images. Here we're going to draw an image from nothing. We're going to draw an image just from data. And so M2 is the PI now. M2 is the pattern image library. And the key thing to do is to draw an image. Okay, so all the data we can use to create this object, this image. And this is just some setup to create an image. Now we're going to create a new image using image.new. This is how big the image is in terms of the image, right? Then we're going to draw some lines in the images and then save the image right here. Okay, after we save the image, we're going to do snakey.png. And we're going to go and find out what snakey.png is like. So we get Python. And then from just numbers, we can create this shape which describes the profile of this object. The length is this one, this length is this one. So I think we just created, we just drew an object on just a label. And then some numbers. Okay, it's right here. Next, okay, we're going to edit API. We're going to download subreddit images using a Reddit API. This is pro. It's a Reddit API wrapper. So now it's going into, so this red API, it is actually letting us go into a certain subreddit or whatever subreddit we choose. And then get all the submissions, or accept all the submissions from this subreddit. So I'm writing the submissions by thumbs up. So if there's no thumbs up, then we're going to download those images. So this is going to be very nice. Let's check it out. We can see that it's all that nice scenery. Okay, let's continue on. Let's download subreddit images with Python. Then we have to create number 15, which is downloading Google images with Python. Same here. We're just going to run this. You can see it's super easy. It's really good in wrapper. Just use Google in your download. Keep on with your postnates. And it's going to download Biotopsnick, which is not going to be recognized. But there's Biotopsnick pictures in our downloads folder. Okay, then the next thing we're going to do is it can also create PDF reports. So we're going to create a PDF report. So what it looks like is we're going to use a library called fpds. So I'm going to give you a PDF here. So we're going to import this thing, which allows us to create PDFs. So from here, all we need to do is tag in some text, and then it can create the rest of the PDF for us. I'm not going to demonstrate. But the key thing is we can create PDFs from the library in Python. And we can figure out the page for these images, all these things, create them for us as necessary. Okay, now we're going to move into a bit of physical locations. It's next distance is, let's show this one. Okay, so all I have to do is reference the institution. The two inputs are reference the institution and home, college view. And then it suggests for the exact address, it gives us the effect, the relevant address. And then it gives us the coordinates in latitude and latitude. And it also gives us the distance from one location to the other. So let's start creating distance using Python. We're going to create the next one. We're going to use the Google API to look for a place. So we're going to look for a place called 6 Singapore. You may not be able to find it. So the key thing here is... So we pass in the parameters and we send a request. So you may know this library called a request. So we send a request to the API URL to the key parameters input. It's missing a box. And then we go get output, which is the place. And it is place. And we get a place ID. This ID, you can find details for the place. So let's go to here. You get details for the place here. And it tells us that it's a clothing store. And it gives us whoever who has created upon it. This guy has created a one-star rating. So let's see who can get details straight from Python. Just by passing in strings. So that's pretty amazing. You can also get directions. I'm not going to show this. We're going to send emails using standard library. It's SMTP library. We create a mind-boggling path. So this is the data that we're going to send. We attach the message. And then we're going to log in with our username and password. And then we're going to send the mail to this type of TTP. Send the mail. Send the mail. So that's that. I'm not going to show that detail. We can operate it point-nigh. Definitely don't have time for that. But it is a good way to learn how to define a human's mind-boggling path. The mind-boggling of T1 and the wallet of T1. This period is what a bit of cryptocurrency is about. It's about mining, processing the transaction, and making the transaction. So you can see what happens after you transaction for the miners and for the people with the wallets making the transactions. I won't go through that. This one is super interesting. So I'm going to go through it super quickly. This is Linus, then this is HiTaserec. It allows us to read text from images. So HiTaserec, I'm going to use this again. So you can see that it's... Why? I don't think this should be here. So it's just like that. Okay, so we're going to run this guy to 8. Change. Change. And then it's going to tell us my encode. It must have gotten messed up somewhere here. So my encode is messed up here. So basically what it's doing... The key thing we call is HiTaserec images to string and then your image. Image is not open. It's using the HiTaserec image library to open the image. And hopefully it can open this image to string. So when this works, as you will have the full effect string from the image. What it looks like is this. And in my casting reports, it converts this 100% accuracy. So that's pretty good. What is the difference? It can work with different models, but with limited capability. So if you're giving it something like... something like... Give it something like... This guy, then it's going to be... But so a lot of once people is very, very long. And you can tune and optimize the results as well. By providing more parameters for more images. The last two things is building the bokeh. We can mix lookbooks and scripts. And add them together. So I'm going to show you one last thing. So I'm going to skip the last one. We can create a script also that we can share our HiTen code with people who don't even install HiTen. So we can share it. Then we can run it without even installing HiTen. But I'm not going to go to that. You basically just run HiInstaller and then run HiTen Power. Okay, but this one is quite interesting. So I'm going to go through. You can mix the lookbook and script. A lookbook is basically the script that I'm not going to practice. Like what we're showing you in the top. And just now, the script is like this. Just like this regular text. It's easy to do this. You can do this and find out what the difference is. You can also use your favorite editor upon script. But then there's also two things about lookbook. And lookbook is good because... Yeah, the lookbook is because it lets us view outputs of codes it is very easily. You can find out why it's going on. And Jupytex allows us to combine code. You can just install Jupytex. And then you go into... Here... I did lookbook. I did lookbook. I did lookbook. I did lookbook. And here you... In your Jupytex... Option, you just put in... Formats. Then what will happen now is... If I type in... Something new here. Okay, I don't know. And then I change this. If I go back into my... If I go back into my... Type in... It actually sends it... Type them out. In my... So it has created the duality of this path. It has created an area exception. Every time I save the lookbook, it will also save a .py file. So... You can just call me the .py file in Jupytex. Okay, so that's pretty much it. I've gone through most of everything I wanted to cover. So it's 30 things from my technique. So that's all. If you have any other questions... Or me, any questions... Sorry, I overestimated my time.