Speaker 1: Hello everyone, welcome to this new video on this channel. Today we'll do something really special. We will code a complete text-to-speech synthesizer with Python. And in the next video, I think it's the first two videos, in the next video I will deploy it online so you can play with it. So I will show you the process of creating a really good text-to-speech synthesizer and the process of deploying it online on a web server and then sharing it with your friends or whoever can play with it. So, let's get started. In order to create a text-to-speech synthesizer in Python, we need some basic modules. And to install those modules, we need Pype. I hope you have Python installed on your computer. So, Python commit Pype. So, the thing that we have to do is just install tts5py, for example. Right now, I have tts on my computer, so I don't need to install it anymore, but you have to install this before. And then let's create a file called tts5py. This file will just remind you to install a requirement. Requirements. And this will be py install tts. Okay. Okay. And the next step is to import all the modules that we will need to use. Perfect. So, we have to go in the tts module and import a synthesizer. We have to do from tts.utils.synthetizer. Import. Synthetizer. I hope I wrote it right. And then the synthesizer with 3s. And then hash here. Yes. And then kinda. Okay. Great. And we need to define some variables, like the path where synthesizer is actually installed on my computer. Let me check. This is here.

Speaker 2: This path. Path. Home. Slash home. Slash ltthing. Slash anaconda. Slash lib. Slash python3. Slash style packages. Yeah.

Speaker 1: And in this folder, there's a file called module.json that we need to import. And this will give us some parameters for the modules. And after that, we need the manager. The manager will help us to load some module, to download the module, actually, and to download some vocoders. So we need to import the module manager from tts.utils.manage. Import module manager. Okay, great. And we just call manager here. Module manager. No, no. Module manager is equal to module manager path. Yeah. Okay, great. And then the next step is to load the module path. Then the config path. And the module item. And for that, we go to the manager. And then we call the function called download module. And for that, we pass the link or the directory where we have the modules. And in this case, I think the default directory will have modules in tts. We can see it on GitHub. It's actually tts.modules.language. We can just see it by typing here tts after you have installed tts.listmodules. And we will see a list of all modules. You will see you can just take the one you have downloaded. So we'll just take this one. Okay. Yes. And then perfect. Okay. And then after that, we need the synthesizer variable. We need to create a synthesizer instance. I don't know. Just call it sent. It's equal to synthesizer. No, no, no. Synthesizer. And for that, we pass a lot of parameters. The first one will be tts.checkpoints. And we pass the module path that we had previously. We have to pass the tts.config file also. We pass the config file that we had previously. Okay, great. And then here, we can pass some text. For example, text is equal to I am a text read by a computer. And then we continue by defining the output as being our variable sent. It's called sent. Yeah. .tts. And we pass the text. And then we just do sent.savewave. Save waves. Outputs. And then we pass a file, a random file, like audio.wav, for example. We save it. And then we try it here. By typing python tts.py. Okay. Synthesizer. I think the problem comes from the way I actually import it. Synthesizer or synthesizer. Synthesizer. Synthesizer. Something like that. Yeah. Synthesizer. So. No, no. That's not the way. It's small. Yes. And then the model works. And we get an audio. Let's play this audio.

Speaker 2: One minute.

Speaker 3: I'll browse to the audio so I can play it. It's actually really bad.

Speaker 1: Yeah, it's working, but it is actually really bad. We have to add a second module out of it. I mean, we have to add a vocoder to make it sound more human. And for that, we will import vocoder here. We'll use the same model manager to download the vocoder. We call it vocoder. Voc path. Voc config path also. Yes. And then a random. Random. Just like this. Is equal to manager. Download. Model manager. Download. Model. And in this case, we use the model item. And the model item, the model item we had previously. And in this model item, we get the full vocoder name. Default. Vocoder. Yes. So, we can now pass the vocoder here. Call vocoder. Vocoder checkpoint. Yeah. Vocoder checkpoint is equal to vocoder path. Voc path, yeah. And then vocoder config path. It's actually the same thing somehow. Vocoder config is equal to voc config path. I think it's called voc. Let's correct that. Voc config path. Yeah, perfect. So, let's try that again. Perfect. Let me play it. Seems better. Seems far more better. Okay. So, let's test it further. So, what I will do is actually paste a really long text. It comes from the last blog post on ulife.ai. I'll just try to read it. And see what it gives. It will take time and it will use computer resources. Let me play it for you so you can see the output.

Speaker 4: It seems not bad at all.

Speaker 1: I think there's some punctuation problems. But I will check again with another text. To see if the problem comes from the text or the synthesizer. So, let's test that. Okay. Play again. Wait.

Speaker 4: It will launch the notebook. And you will be able to write and debug your app easily. Firstly, we need to import the packages that we will be using.

Speaker 1: Great. I think that's perfect. Okay. So, I think it's all for today. It's really simple. I will post this code on gist and put the link down in the description. And in the next video, I will just try to optimize this synthesizer and deploy it online. So, you can test it yourself. So, see you in the next video. Bye. Bye.