😊Introduction
It is also might say a title like, "how can I make my own voice models and contribute them?"
Last updated
It is also might say a title like, "how can I make my own voice models and contribute them?"
Last updated
Please note that some guides are not finished yet, but although it's available in public, take a look around and I'll add some in the future.
Hello and welcome to the voice model making tutorial.
This is where we will be a walkthrough on how to make your own voice model with Tacotron 2 and uploading it to the Uberduck website.
But, you've probably never heard of those, so I'll explain one by one as much as I can.
Oh, forgot to introduce myself. Names Awesome Face (aka awesomeface349) and I created this guide for those who want to make a voice model for the first time.
I made the Johnny Silverhand voice model that's in public on uberduck (or Keanu Reeves whatever is your taste).
We get those every time in the #inquiries channel section who wants to do it for the first time, but people are literally confused and don't know how things work. Especially some people, asking for a voice to be in uberduck in every channel which, annoying but most likely may get less chance to get created.
Thankfully I had to make this guide for YOU! the user, who wants to make a voice model for the first time.
It'll cover everything with gathering dialogs and training stuff.
So let me explain what is Tacotron 2 and Uberduck.
From what I understand is that Tacotron 2 is a training and synthesis of AI text to speech.
It'll be able to clone the voice of the character that you want and then turn it into a cloned voice. And you can make them say whatever the hell you want.
Although sometimes they say not correct words and you might need to fix them a little bit with pronunciations.
But also one more thing. It requires A LOT of GPU power. Your gaming GPU may not work, especially Intel Graphics HD (ew why.)
Just so you know when you want to training someone's voice or your voice you possibly do need more GPU video memory or else that ain't gonna work like that, My guess RTX 3000 series ain't gonna work like that eh?
So as I said before, you do need an expensive video card to work like this, for example, the NVIDIA Tesla series.
Which those graphics cards will work for training voices, but they cost expensive.
However, if you are lucky enough to get one, you can basically train voices on your own end, but requires an OS like Linux, but not sure if Windows is gonna work. And also programming knowledge like Python, which that's what they use.
(Update: However it does require a lot of VRAM for you to train voices like that so (Like RTX 3K series with 24 GB or 30 GB), just so you know that)
But enough of this let's talk about Uberduck.
So, you've probably heard it or never heard of Uberduck right?
Well, let me explain how this works. Remember 15.ai or vo.codes (aka FakeYou)? that's the one you remember. But also in the fact that 15.ai is still on hiatus at the moment (which is retraining some voices to a new version) (it's now out) and FakeYou still work like a charm, we have a new website called uberduck.ai.
So how's this website works?
You first need to log in to your Discord account, or Google account.
As matter of fact, that's why Uberduck got so much popular (it's because of a one video TikTok that got millions of views on it, especially Twitter) traffic is just going too far and they had to create a login page so that it won't gonna have huge spike lags on their end.
But anyways, you select a character and type what you need and press synthesize.
Please note that this gonna take a while, depending on the servers.
If that took too long then that's probably many people using Uberduck.
But you can do whatever you like with Uberduck, making skits, memes, heck even make a song for it (but they won't sing so you gonna have to do it pitching manually). Uberduck has implented TalkNet.
They also have a Discord server which, joining a server is recommended for assistants and server statuses. Uberduck has also a Twitter account so you can give em a follow for it.
Oh one more thing, if they gonna notice that you used Uberduck voices but didn't credit them, please do so credit so that people would know what the heck is it.
Great now that is getting out of the way let's do things step by step on how to make your own voice model.
This is gonna be a written tutorial but if you don't like reading and you only want is to watch a video tutorial for it then here, you can watch it right here:
This guide may also be complicated for phone users, but I've covered some for the web versions of that.
