Get Started with Voicebox: Open-Source Alternative to ElevenLabs Tutorial

Quick setup guide for Voicebox: the free, open-source voice synthesis studio! Clone voices locally with Qwen3-TTS, edit multi-voice stories in a timeline

*https://www.youtube.com/watch?v=RRRBxNXgeKQ
**

пересказ видео

00:00:00 Введение в VoiceBox

  • VoiceBox — программа для клонирования голоса.
  • В отличие от Eleven Labs, VoiceBox бесплатен и не требует интернета после установки.
  • Ссылка на скачивание доступна в описании видео.

00:00:59 Установка и первые шаги

  • Процесс установки прост и аналогичен другим программам.
  • Рекомендуется ознакомиться с управлением моделью перед использованием.
  • Упоминается необходимость скачивания Quen TTS 1.5.0.7B.

00:01:58 Выбор модели и создание голоса

  • Возможность выбора между разными моделями генерации голоса.
  • Рекомендация скачать Whisper Base для упрощения процесса.
  • Два варианта создания голоса: импорт или создание нового.

00:02:38 Запись и распознавание голоса

  • Три варианта получения аудиосэмпла: из аудиофайла, с микрофона или системного звука.
  • Максимальное время записи — 30 секунд.
  • Использование Whisper для распознавания текста по аудиозаписи.

00:04:25 Генерация речи

  • Загрузка Quen3 TTS при нажатии на кнопку «Сгенерировать речь».
  • Проблемы с загрузкой и их решение путём принудительного закрытия и перезапуска программы.
  • Локальная обработка может занять около минуты.

00:05:51 Примеры генерации голосов

  • Пример генерации голоса Стинки Скраблета на основе одного аудиоклипа.
  • Возможность добавления нескольких образцов для повышения качества.
  • Пример генерации голоса Алекса Джонса.

00:06:51 Создание нового голоса

  • Запись звука с компьютера для создания нового голоса.
  • Проблемы с распознаванием голоса Тео Вогана из-за необычных звуков.
  • Попытка улучшить точность распознавания путём записи дополнительного аудиоклипа.

00:09:32 Функция «рассказ»

  • Возможность добавления нескольких голосов и создания диалога.
  • Пример добавления реплик разных людей в один аудиофайл.
  • Подчёркивание преимуществ программы: бесплатность и локальный хостинг.

00:10:31 Заключение

  • Благодарность зрителям и пожелание хорошего дня или вечера.

В этом видео

Эпизоды
Расшифровка видео
Поиск в расшифровке
Эпизод 1: Introduction to voicebox
0:00
0 секунд
Hello everybody, Stinky Scrubble here and today we’ll be looking at a software called Voice Box. It’s a voice cloning software. Uh you may have heard of other
0:09
9 секунд
ones like Alone uh sorry 11 Labs which is on the browser. Um I believe they have a free trial still. It’s been a
0:16
16 секунд
while since I’ve used it, but eventually you do have to pay for it.
0:20
20 секунд
As well as that, um I do believe recently they made it so you can’t clone um famous people’s voices, actors,
0:29
29 секунд
uh basically public figures, uh for legal reasons. So that’s something to think about there. Uh so with this
0:37
37 секунд
software, like I said, it’s completely free hosted on your computer. Uh don’t need the internet uh after you have
0:44
44 секунды
everything downloaded. So it’s really cool. And yeah, you’ll scroll down here and you’ll see the downloads
0:52
52 секунды
link is in the description. Not sure if I mentioned that. Either way, there’ll be different ones for different operating systems. I’m on Windows. So,
0:59
59 секунд
it Oops, misclick. So, it’ll be this setup.exe here. Um, the setup wizard is
1:08
1 минута 8 секунд
really easy. It’s basically like every other software that you ever downloaded.
Эпизод 2: Setting up the models
1:12
1 минута 12 секунд
You’ll choose probably the location. uh if you want a desktop icon, that kind of stuff. It’s really self-explanatory.
1:20
1 минута 20 секунд
So, I don’t need to show you the install, but once you have it downloaded, it will look just like this.
1:27
1 минута 27 секунд
Uh well, almost like this. You won’t have these voices uh that I’ve already cloned here. So, once you’re in here, you’ll see a few different things.
1:36
1 минута 36 секунд
And I would recommend going into the model management before you start using it. It’ll just make the process simpler.
1:44
1 минута 44 секунды
uh quicker. You’ll want to download uh Quent TTS 1.7B or whatever the newest
1:51
1 минута 51 секунда
one is. Um it might be changed by the time you watch this video. And you’ll go ahead and download that.
1:59
1 минута 59 секунд
You can also download this model, which I believe is a lot smaller of a size, uh
2:05
2 минуты 5 секунд
but it’s not as accurate as the Quen TTS uh voice generation model is. So, I just went with the best one and downloaded
2:13
2 минуты 13 секунд
it. And then I would also recommend downloading Whisper Base. Uh it’ll make things a little bit simpler and easier.
2:20
2 минуты 20 секунд
Uh which I’ll show you why right now.
2:24
2 минуты 24 секунды
So, we can go ahead and go to uh there’s two options. There’s import voice, which I’ve not tried yet. I bet you can have a
Эпизод 3: Cloning your own voice
2:32
2 минуты 32 секунды
friend share a voice and just send it over. And then there’s also create voice.
2:38
2 минуты 38 секунд
When you’re creating a voice, there’s three options where you’ll get the audio sample from. Um, an audio file, a
2:46
2 минуты 46 секунд
recording directly from your microphone, or system audio that captures straight from the uh system of your computer. So,
2:55
2 минуты 55 секунд
basically, you can choose either of these. As you can see, it says max clip size 30 seconds. You could add multiple
3:02
3 минуты 2 секунды
clips, but the max uh length of those uh clips is 30 seconds. So, we can go ahead
3:10
3 минуты 10 секунд
and go ahead and basically clone my own voice. Let’s go ahead and try that now.
3:16
3 минуты 16 секунд
Hello everybody. This is Stinky Scribblelet. I’m currently doing a small recording for the YouTube video uh to
3:24
3 минуты 24 секунды
clone my own voice. Um you can say whatever you want in this. So there’s nothing specific that you have to say,
3:30
3 минуты 30 секунд
but yep, I am recording my own voice and we’re going to see how this does.
3:36
3 минуты 36 секунд
So, uh, as you sorry, as as you saw before, um, I said
3:45
3 минуты 45 секунд
you should download the whisper. And the reason that is is because after you record your voice, it’ll ask you to
3:52
3 минуты 52 секунды
insert the text spoken in the audio so it knows what you’re saying. Um, with the whisper you can hit just hit the
3:59
3 минуты 59 секунд
transcribe button and boom, it has everything accurately
4:06
4 минуты 6 секунд
uh what I said so it knows what I said and uh yeah, it makes it a lot simpler so
4:14
4 минуты 14 секунд
you’re not trying to remember what you said basically. Um yeah, so it’s all there and you can create profile. Boom.
4:23
4 минуты 23 секунды
Scrub is released and we can go ahead and generate speech using Scrub
4:32
4 минуты 32 секунды
and we’re going to go ahead and generate it. So, if you didn’t download the Quen 3 TTS yet, right as you hit this
4:42
4 минуты 42 секунды
generate speech button, it’s going to go ahead and download it. Mine did have a little bit of an issue. I don’t know why. Um, basically download wasn’t
4:52
4 минуты 52 секунды
going. Uh, it does look like real download progress tracking and accurate percentage and speed info was in the latest update, but I got it after that.
5:02
5 минут 2 секунды
Mine seemed a little wonky, so I forced quit out and uh restart the program and it did end up downloading uh very
5:08
5 минут 8 секунд
easily. So, we can go ahead and type in anything that we’d like here. Hello.
5:17
5 минут 17 секунд
This is stinky. Oh god, mistype. Stinky scrubblelet.
5:25
5 минут 25 секунд
Thanks for watching my video.
5:29
5 минут 29 секунд
Hopefully this sounds like me. All right, so I’m going to go ahead and generate that. Uh, it is doing the work
5:38
5 минут 38 секунд
locally, so it might take longer than, you know, something like 11 labs. Uh, it might take about a minute here. So, I’m
5:45
5 минут 45 секунд
just going to go ahead and skip ahead to where it generates it. Hello, this is Stinky Scrubblelet.
5:53
5 минут 53 секунды
Thanks for watching my video. Hopefully, this sounds like me. All right, so that was basically from one audio clip or
6:02
6 минут 2 секунды
actually was from one audio clip. And as you can see, you can add more samples to um basically increase the quality and
6:10
6 минут 10 секунд
increase uh the likeliness or how much it sounds like you. So, that worked
6:17
6 минут 17 секунд
really well from just one small audio clip.
Эпизод 4: Cloning celebrity voices
6:21
6 минут 21 секунда
Um I’ve done a few different voices like Alex Jones. Um I only have one audio clip of him as well, so it might not be
6:29
6 минут 29 секунд
as good as like I said if you add more audio clips. So we can go ahead generate speech using Jones. This is just going to be a quick example.
6:51
6 минут 51 секунда
Hello, this is Alex Jones from infowars.com.
6:55
6 минут 55 секунд
There you go. There’s Alex Jones. So I’ve done a few other ones. I’ve done Rogan, but I can show you what it’s like creating a new voice. Let’s say I
7:04
7 минут 4 секунды
haven’t done Theo Bon and here’s a video of him just himself talking. So we can go ahead and say Theo
7:13
7 минут 13 секунд
and you can capture from your system audio. So let’s just go to a certain point in the video. No.
7:20
7 минут 20 секунд
Oh, [music] there’s music there. We don’t want that.
7:24
7 минут 24 секунды
You for a while and quartet. Okay, there’s no music here. That little perverted butterfly hook.
7:30
7 минут 30 секунд
So, we’re just going to do start system capture.
7:35
7 минут 35 секунд
Just let them. Some people don’t like a damn hug.
7:42
7 минут 42 секунды
Some people don’t like a hug and sometimes I don’t. Sometimes I’m a cactus.
7:51
7 минут 51 секунда
Don’t touch me.
7:54
7 минут 54 секунды
But sometimes I just want to I just I just want a hug. I won’t ask.
8:02
8 минут 2 секунды
So that that’s kind of a weird audio clip, but let’s go ahead and transcribe it because he’s making noises and different things. So I might uh we’ll do
8:10
8 минут 10 секунд
that. Create profile. And I’m going to go back and maybe do one more audio clip just because that was kind of, you know,
8:18
8 минут 18 секунд
he’s a weird guy. Uh, let’s do one more to see if we can get some good accuracy on his voice.
8:25
8 минут 25 секунд
Throw a hug though.
8:28
8 минут 28 секунд
That’s who I am. I ain’t no hoe. You feel me? Ho ho. Sorry. I know Santa gets offended if you say I ain’t no hoe cuz
8:36
8 минут 36 секунд
that’s his thing. But uh yeah, you know, I won’t ask for a hug. I’ll stand by you
8:42
8 минут 42 секунды
and look like I need a hug or do some or just do some like this, you know?
8:51
8 минут 51 секунда
What’s up?
8:53
8 минут 53 секунды
See, that’s probably not good. He’s doing a bunch of noises. He started doing a impression. So, that might be weird. Let’s see.
9:04
9 минут 4 секунды
That’s fine, though. Let’s Let’s just try that. So, now let’s do one more. uh example for Theoon.
9:32
9 минут 32 секунды
This is Theo Vaughn. Welcome to my podcast this past weekend. Okay, so that one didn’t work as good as the other ones, but he’s acting a little weird.
9:40
9 минут 40 секунд
You know, if I had regular audio, it’d be better. Uh there’s another thing here called story. Um this is cool because
Эпизод 5: Creating stories and closing
9:48
9 минут 48 секунд
you can add you can basically make a conversation between uh multiple people. So you could
9:57
9 минут 57 секунд
add Theo Vaughn, you could add Alex Jones, you could have Joe Rogan and basically add them um
10:04
10 минут 4 секунды
all in one audio clip. So you can add Jones saying something, then Rogan, then Theo Vaughn, and basically create a
10:11
10 минут 11 секунд
conversation. I’m not go through uh not going to go through with that in this video, but it’s really simple and straightforward to do. Uh either way, uh
10:20
10 минут 20 секунд
cool software, free, like I said, hosted locally. Um, so you won’t have to be worried about, you know, paying a
10:27
10 минут 27 секунд
subscription or anything like that. I hope you guys enjoyed this video. Thank you guys very much. And I hope you enjoy the rest of your day or night.
Показывать комментарии к текущему моменту видео

Поделиться: