PHBJT/cml-tts-filtered-annotated

Dataset Card for Filtred and annotated CML TTS

This dataset is an annotated and filtred version of a CML-TTS [1].

CML-TTS [1] CML-TTS is a recursive acronym for CML-Multi-Lingual-TTS, a Text-to-Speech (TTS) dataset developed at the Center of Excellence in Artificial Intelligence (CEIA) of the Federal University of Goias (UFG). CML-TTS is a dataset comprising audiobooks sourced from the public domain books of Project Gutenberg, read by volunteers from the LibriVox project. The dataset includes recordings in Dutch, German, French, Italian, Polish, Portuguese, and Spanish, all at a sampling rate of 24kHz.

The original dataset has been cleaned by removing all rows with a Levenshtein score inferior to 0.9. In the text_description column, it provides natural language annotations on the characteristics of speakers and utterances, that have been generated using the Data-Speech repository.

This dataset was used alongside the LibriTTS-R English dataset and the Non English subset of MLS to train [Parler-TTS Multilingual Mini v1.1. A training recipe is available in the Parler-TTS library.

Motivation

This dataset is a reproduction of work from the paper Natural language guidance of high-fidelity text-to-speech with synthetic annotations by Dan Lyth and Simon King, from Stability AI and Edinburgh University respectively. It was designed to fine tune the Parler-TTS Mini v1.1 on 8 european languages (including English).

Contrarily to other TTS models, Parler-TTS is a fully open-source release. All of the datasets, pre-processing, training code and weights are released publicly under permissive license, enabling the community to build on our work and develop their own powerful TTS models. Parler-TTS was released alongside:

• The Parler-TTS repository - you can train and fine-tuned your own version of the model.
• The Data-Speech repository - a suite of utility scripts designed to annotate speech datasets.
• The Parler-TTS organization - where you can find the annotated datasets as well as the future checkpoints.

Usage

Here is an example on how to oad the clean config with only the train.clean.360 split.

from datasets import load_dataset

load_dataset("https://huggingface.co/datasets/PHBJT/cml-tts-filtered", "french", split="train")

Note: This dataset doesn't actually keep track of the audio column of the original version. You can merge it back to the original dataset using this script from Parler-TTS or, even better, get inspiration from the training script of Parler-TTS, that efficiently process multiple annotated datasets. You can find the original dataset here

Dataset Sources

• Homepage: https://www.openslr.org/141/
• Paper: https://arxiv.org/abs/2305.18802

Citation

@misc{oliveira2023cmltts,
      title={CML-TTS A Multilingual Dataset for Speech Synthesis in Low-Resource Languages}, 
      author={Frederico S. Oliveira and Edresson Casanova and Arnaldo Cândido Júnior and Anderson S. Soares and Arlindo R. Galvão Filho},
      year={2023},
      eprint={2306.10097},
      archivePrefix={arXiv},
      primaryClass={eess.AS}
}

@misc{lacombe-etal-2024-dataspeech,
  author = {Yoach Lacombe and Vaibhav Srivastav and Sanchit Gandhi},
  title = {Data-Speech},
  year = {2024},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/ylacombe/dataspeech}}
}

@misc{lyth2024natural,
      title={Natural language guidance of high-fidelity text-to-speech with synthetic annotations},
      author={Dan Lyth and Simon King},
      year={2024},
      eprint={2402.01912},
      archivePrefix={arXiv},
      primaryClass={cs.SD}
}