 Okay, got it. So hello everybody. So at first, we actually pivoted a lot too. So at first, if anybody knows or has watched anime before, as in this voice actor is actually very, very popular. As in she has starred in a lot and a lot of anime before. Like really a lot. So there won't be a chance that you never heard her voice before. So we thought about like what can we, what if we can make a speaker recognition? So apparently we found out that speech recognition and speaker recognition is a different thing. So at first we thought about, okay, so what if we take some of, so can we actually like when we watch an anime, can we know that whether she's voicing an actor, that she's voicing a character in that series. Then we realized that we don't have any capabilities in doing audios. So what happened is that we, so this is one of the character that, this is one of the character that she voices in this anime. And like this character has a very iconic tuturu, like tuturu songs to the point where as in there's just like 10 hours and this video alone has been watched like 1.6 millions. And there's like a lot of things like this too, for example. All right, so that brings us to the question. Like who can actually make the best tuturu out there? And then like you don't want to rely on subjective, like unsubjective answers. We want the objective data. So what happened was that we actually run, it's something, something collected, something collected, wow, simple. So this is the, like this is the original tuturu. Wait, wait, all right. Oh, it's, wait, wait. So this is the original tuturu, right? And then like, and then to quantify and to build our mortal. So what happens is that we actually have to, so we started recording our voices and our voices and like use this API to actually find out. So we use this API called the, oops, not this one, this one called the audio analysis. And we found out that actually like, how can we actually quantify a sound? It turns out that they're about at least 34, like 34 per meters, like 34 per meters just to like measure whether like a voice is similar to something or not. And like we spent about four hours trying to find, like listening to tuturu over and over again. Like listening to tuturu over and over again and then trying to find out like out of all these APS stuff. So we do all this analysis like for example, wait, that's a little thing. Yeah. So we tried to do all these analysis like for example, that's tuturu, that goes out. Then afterwards, like, so we record ourselves, we record like our own, we record our own, like this is the, like this is the model and this is the sample to, that's my tuturu. Then we realized that so like out of all those 24 per meters, like what are these things? How can this be a full red? Well, this is yellow. And then after going through all the data and spreadsheet, after going through all the data and spreadsheet, where is it? And like even like, we have to go through all these facts for grounding. And then, and we are very proud to announce that we have found out the formula to actually find this car. So this is my tuturu and this is Hanazawa Kana's tuturu. So this is like, so you have the zero crossing ray energy. And so after like, after we try to compare everything, only the energy column makes sense to us. And then, and then after that we made this one. So this is the, so this is how like for and then starting from this data, after we got this one, then we start building a scoring algorithm. So, so scoring algorithm is built by this guy over here, who is a credit analyst by job, by job. So afterwards, what happened is that currently we have this API. So every time you like, just as long as you have the, so yeah. So that's the, so zero means really good. Zero means you're Hanna and like one means you're really bad. So the closer you are to zero, the better you, like the better your score is. So if anybody wants to try to speak like, to like record the tuturu, we can give you your score. Anyone want to try now? We can actually do that. Are you going to do like really high pitch or are you going to go like... So let's mute this one first and mute this one first. Okay, so let's make sure that this is there. All right, you're, you're free. You're free. You're free to come closer here. Okay. And like by the kind of tree, you can say tuturu. Okay. One, two, three. All right. And then what happened is that we are going to take this audio track, like we are ready to select this audio track, file, export the, export the selected audio. Wait, tuturu joy. Okay, tuturu joy and save it to, as a graph. We continue. Then after that, we, we do the analysis. And we do, we do the extraction. Okay. It's left. Sorry. We wanted to make an application like, but we just didn't have the time. All right. So this, so the file is now generated and then we go to the file. Like, and then we go, this is WFE and this is joy. All right. And then we run the file. Oh, you're very close. So we actually got a score of 0.56. But yeah, as in, so that's all for our demo. Is, is the tune like stuck in your head now? So like six, six hours, six hours. Constantly, yes? I mean, so is this potentially scalable for other characters? Actually, you can sync up with the other hobby. So you can mimic like your favorite character's voice. So at first we wanted to do Pikachu. So like, after, after knowing, like after knowing about this, like, oh, we can actually do, like do this contest. No, we think about what if we make like, what if we expand this to like all the Pokemon characters? And then we realize that it's very hard to find a sample. So because we have to go to the video and cut it and cut them. Well, this sample is now as lossless. So we go for this one. But yes, you can try to compete to be the best as a voice maker for your favorite character. You mean like to be the best like no one ever was? Any other comments, questions? Anyone? Okay, enough. Thank you very much. Sure.