Mms Tts Eng

🚀 Massively Multilingual Speech (MMS): English Text-to-Speech

This repository provides an English (eng) text-to-speech (TTS) model checkpoint. It's part of Facebook's Massively Multilingual Speech project, aiming to offer speech technology across diverse languages. You can find details about supported languages and their ISO 639-3 codes in the MMS Language Coverage Overview, and view all MMS-TTS checkpoints on the Hugging Face Hub: facebook/mms-tts. The MMS-TTS is available in the 🤗 Transformers library starting from version 4.33.

✨ Features

• Multilingual Support: Part of a project aiming to cover over 1000 languages.
• End - to - End Synthesis: Utilizes the VITS model for direct text - to - speech conversion.
• Stochastic Duration Prediction: Allows for different speech rhythms from the same text input.

📦 Installation

MMS-TTS is available in the 🤗 Transformers library from version 4.33 onwards. To use this checkpoint, first install the latest version of the library:

pip install --upgrade transformers accelerate

💻 Usage Examples

Basic Usage

from transformers import VitsModel, AutoTokenizer
import torch

model = VitsModel.from_pretrained("facebook/mms-tts-eng")
tokenizer = AutoTokenizer.from_pretrained("facebook/mms-tts-eng")

text = "some example text in the English language"
inputs = tokenizer(text, return_tensors="pt")

with torch.no_grad():
    output = model(**inputs).waveform

Advanced Usage

Save the resulting waveform as a .wav file:

import scipy

scipy.io.wavfile.write("techno.wav", rate=model.config.sampling_rate, data=output.float().numpy())

Or display it in a Jupyter Notebook / Google Colab:

from IPython.display import Audio

Audio(output.numpy(), rate=model.config.sampling_rate)

📚 Documentation

Model Details

VITS (Variational Inference with adversarial learning for end-to-end Text-to-Speech) is an end-to-end speech synthesis model that predicts a speech waveform conditional on an input text sequence. It is a conditional variational autoencoder (VAE) comprised of a posterior encoder, decoder, and conditional prior.

A set of spectrogram-based acoustic features are predicted by the flow-based module, which is formed of a Transformer-based text encoder and multiple coupling layers. The spectrogram is decoded using a stack of transposed convolutional layers, much in the same style as the HiFi-GAN vocoder. Motivated by the one-to-many nature of the TTS problem, where the same text input can be spoken in multiple ways, the model also includes a stochastic duration predictor, which allows the model to synthesise speech with different rhythms from the same input text.

The model is trained end-to-end with a combination of losses derived from variational lower bound and adversarial training. To improve the expressiveness of the model, normalizing flows are applied to the conditional prior distribution. During inference, the text encodings are up-sampled based on the duration prediction module, and then mapped into the waveform using a cascade of the flow module and HiFi-GAN decoder. Due to the stochastic nature of the duration predictor, the model is non-deterministic, and thus requires a fixed seed to generate the same speech waveform.

For the MMS project, a separate VITS checkpoint is trained on each langauge.

📄 License

The model is licensed as CC-BY-NC 4.0.

🔧 Technical Details

This section details the inner workings of the VITS model used in the MMS-TTS project. VITS is a conditional variational autoencoder (VAE) designed for end - to - end text - to - speech synthesis. It consists of a posterior encoder, decoder, and conditional prior. The flow - based module, which includes a Transformer - based text encoder and multiple coupling layers, predicts spectrogram - based acoustic features. The spectrogram is then decoded using transposed convolutional layers similar to the HiFi - GAN vocoder.

The model incorporates a stochastic duration predictor to account for the one - to - many nature of TTS, where the same text can be spoken in multiple ways. This allows the model to generate speech with different rhythms from the same input text.

Training is done end - to - end with a combination of losses from variational lower bound and adversarial training. Normalizing flows are applied to the conditional prior distribution to enhance the model's expressiveness.

During inference, text encodings are up - sampled according to the duration prediction module and then mapped to the waveform using a cascade of the flow module and HiFi - GAN decoder. Since the duration predictor is stochastic, the model is non - deterministic, and a fixed seed is required to generate the same speech waveform.

BibTex citation

This model was developed by Vineel Pratap et al. from Meta AI. If you use the model, consider citing the MMS paper:

@article{pratap2023mms,
    title={Scaling Speech Technology to 1,000+ Languages},
    author={Vineel Pratap and Andros Tjandra and Bowen Shi and Paden Tomasello and Arun Babu and Sayani Kundu and Ali Elkahky and Zhaoheng Ni and Apoorv Vyas and Maryam Fazel-Zarandi and Alexei Baevski and Yossi Adi and Xiaohui Zhang and Wei-Ning Hsu and Alexis Conneau and Michael Auli},
    journal={arXiv},
    year={2023}
}