Voice Cloning For Any Language | Fine-Tuning Tortoise-TTS | Part 1

In this video I will show you how to fine-tune the Tortoise-TTS model to generate speech in any language! If you want to explore the realm of text-to-speech models beyond English, this video is for you. In this video I will show you a step-by-step process for adapting the Tortoise-TTS model for your native language, allowing you to create high-quality speech samples in your language. From acquiring or creating a suitable dataset to adjusting the fine-tuning code, everything will be covered. Plus, don't miss out on the chance to win an NVIDIA RTX 3080 Ti GPU! I hope you enjoy this video which hopefully allows you to generate speech in your language.

Register for GTC 2024 and win an NVIDIA RTX 3080 Ti (Deadline March 22nd):

Send Your Proof of Attendance:

GTC sessions mentioned in the video:

What’s Next in Generative AI

The Fastest Stable Diffusion in the World

Human-Like AI Voices: Exploring the Evolution of Voice Technology

Code Used in This Video

My Medium Article for This Video:

My Workstation

00:00:00 Intro
00:00:21 Promo
00:01:27 Prepare Your Dataset
00:06:00 Adjust Fine-Tuning Code
00:13:04 Clone and Install Adjusted Repository
00:14:15 Train Tokenizer For Your Language
00:18:39 Adjust Sampling Rate
00:19:32 Fine-Tuning

Stay in Touch

Medium

LinkedIn

YouTube
Of course, feel free to subscribe to my channel! :-)

Of course, financial support is completely voluntary, but I was asked for it:

[carlosedubarreto]

Wow. incredible. I was trying to make a Tortoise TTS to work in portuguese and was lost, now I have a way to do that, thanks for sharing this info. Now I just have to wait for the other parts, and find free time to do that.

that is an amazing effort from your side, since its a very complex topic 👏👏
Congrats;

[bouchrasaidi]

Hello, thank you for the great tutorial and i wanted to ask when the part2 of this please ?

[olcaybuyan]

Great video. Looking forward to the custom dataset video.

[shovonjamali]

Another great one, really useful but I have a question though. The dataset you used, has different speakers (like maybe even male or female too), right? So, for training the model, we can put all the wavs from different speakers under a single wavs folder, we don't need to create/manage different ones for different speakers, is my understanding correct?

[FAITHseek]

Please make a Fine Tune guide for MetaVoice 1B TTS

[albertigle]

Nice video Martin! How long did it take you to train the new language?

[dogfoxpodcast]

Hi. Great job. I'm encountering the same problem over and over, though: ModuleNotFoundError: No module named 'unified_voice2'. Any idea of why this happens?

[Athelstanovsky]

Hi, Great video, Does the TTS work with an RTX2060 ?

[awnyfaris]

Hi Martin,
Is it possible to train a voice in the Arabic language and then use that voice to read English text ?

[bobsmithy]

Who's the lucky new owner for the 3080Ti?

[tempertephra]

I used files from your fork with audiobook maker fro jarod mica. Quite passable but longer sentences corrupt.

[DM-dyvn]

Quite a few subsets in that German language data are of peculiar quality. Anastasia Solokha gave me shievers ))

[MightyMindsDev]

Hello. I would like to hear how to create a dataset for your language

[BoskaPalma]

My transcription txt file is around 1GB, i am running tokenizer now for about 30 minutes and don't see any progress 🤔
Running locally on m1 max mac studio

[chiyanchandru]

how can i run this code local machine ?

[bouchrasaidi]

Hello, can i fine tune turtoise for English speech?

[ashuu]

heyy, did you implement this without any gpu?

[timothymaggenti]

WHY do you always use someone else's computer, what is the point of this....

[miwoj]

WHERE IS THE PART WHERE YOU SHOW US HOW IT CAME OUT ??
WE WOULD REALLLY LIKE TO HEAR IF RESULT IS WORTH ALL THAT EFFORT OR IF WORKS AT ALL.
dislike for being useless and wasting my time.