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 data archives were restructured from the original ones from OpenSLR to make it easier to stream.

Supported Tasks

text-to-speech, text-to-audio: The dataset can also be used to train a model for Text-To-Speech (TTS).

Languages

The dataset includes recordings in Dutch, German, French, Italian, Polish, Portuguese, and Spanish, all at a sampling rate of 24kHz.

How to use

The datasets library allows you to load and pre-process your dataset in pure Python, at scale. The dataset can be downloaded and prepared in one call to your local drive by using the load_dataset function.

For example, to download the German config, simply specify the corresponding language config name (i.e., "german" for German):

from datasets import load_dataset

mls = load_dataset("ylacombe/cml-tts", "german", split="train")

Using the datasets library, you can also stream the dataset on-the-fly by adding a streaming=True argument to the load_dataset function call. Loading a dataset in streaming mode loads individual samples of the dataset at a time, rather than downloading the entire dataset to disk.

from datasets import load_dataset

mls = load_dataset("ylacombe/cml-tts", "german", split="train", streaming=True)

print(next(iter(mls)))

Bonus

You can create a PyTorch dataloader directly with your own datasets (local/streamed).

Local:

from datasets import load_dataset
from torch.utils.data.sampler import BatchSampler, RandomSampler

mls = load_dataset("ylacombe/cml-tts", "german", split="train")
batch_sampler = BatchSampler(RandomSampler(mls), batch_size=32, drop_last=False)
dataloader = DataLoader(mls, batch_sampler=batch_sampler)

Streaming:

from datasets import load_dataset
from torch.utils.data import DataLoader

mls = load_dataset("ylacombe/cml-tts", "german", split="train", streaming=True)
dataloader = DataLoader(mls, batch_size=32)

To find out more about loading and preparing audio datasets, head over to hf.co/blog/audio-datasets.

Dataset Structure

Data Instances

A typical data point comprises the path to the audio file, usually called file and its transcription, called text. Some additional information about the speaker and the passage which contains the transcription is provided.

{'audio': {'path': '6892_8912_000729.wav', 'array': array([-1.52587891e-...7344e-05]), 'sampling_rate': 24000}, 'wav_filesize': 601964, 'text': 'Proszę pana, tu pano... zdziwiony', 'transcript_wav2vec': 'proszę pana tu panow... zdziwiony', 'levenshtein': 0.96045197740113, 'duration': 13.648979591836737, 'num_words': 29, 'speaker_id': 6892}

Data Fields

audio: A dictionary containing the audio filename, the decoded audio array, and the sampling rate. Note that when accessing the audio column: dataset[0]["audio"] the audio file is automatically decoded and resampled to dataset.features["audio"].sampling_rate. Decoding and resampling of a large number of audio files might take a significant amount of time. Thus it is important to first query the sample index before the "audio" column, i.e. dataset[0]["audio"] should always be preferred over dataset["audio"][0].
text: the transcription of the audio file.
speaker_id: unique id of the speaker. The same speaker id can be found for multiple data samples.
transcript_wav2vec: the transcription of the audio file using the wav2vec model. Has been used to curate the dataset.
wav_filesize: The size of the audio waveform file. Has been used to curate the dataset.
levenshtein: The Levenshtein distance between the wav2vec transcription and the original transcription. Has been used to curate the dataset.
duration: The duration of the audio in seconds.
num_words: The number of words of the transcription.

Data Splits

# Samples	Train	Dev	Test
german	608296	5314	5466
dutch	309785	4834	4570
french	107598	3739	3763
spanish	168524	3148	3080
italian	50345	1765	1835
portuguese	34265	1134	1297
polish	18719	853	814

Data Statistics

Language	Duration (Train)	Duration (Test)	Duration (Dev)	Speakers (Train)	Speakers (Test)	Speakers (Dev)
	M	F	M	F	M	F
Dutch	482.82	162.17	2.46	1.29	2.24	1.67
French	260.08	24.04	2.48	3.55	3.31	2.72
German	1128.96	436.64	3.75	5.27	4.31	5.03
Italian	73.78	57.51	1.47	0.85	0.40	1.52
Polish	30.61	8.32	0.70	0.90	0.56	0.80
Portuguese	23.14	44.81	0.28	0.24	0.68	0.20
Spanish	279.15	164.08	2.77	2.06	3.40	2.34
Total	3,176.13		28.11		29.19

Dataset Creation

Curation Rationale

[Needs More Information]

Source Data

Initial Data Collection and Normalization

[Needs More Information]

Who are the source language producers?

[Needs More Information]

Annotations

Annotation process

[Needs More Information]

Who are the annotators?

[Needs More Information]

Personal and Sensitive Information

The dataset consists of people who have donated their voice online. You agree to not attempt to determine the identity of speakers in this dataset.

Considerations for Using the Data

Social Impact of Dataset

[More Information Needed]

Discussion of Biases

[More Information Needed]

Other Known Limitations

[Needs More Information]

Additional Information

Dataset Curators

[Needs More Information]

Licensing Information

Public Domain, Creative Commons Attribution 4.0 International Public License (CC-BY-4.0)

Citation Information

@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}
}