LibriTTS: A Corpus Derived from LibriSpeech for Text-to-Speech
By Heiga Zen, Viet Dang, R. Clark, Yu Zhang, Ron J. Weiss, Ye Jia, Z. Chen, Yonghui Wu. 2019
1,387 citations recorded by the source corpus.
This paper introduces a new speech corpus called "LibriTTS" designed for text-to-speech use. It is derived from the original audio and text materials of the LibriSpeech corpus, which has been used for training and evaluating automatic speech recognition systems. The new corpus inherits desired properties of the LibriSpeech corpus while addressing a number of issues which make LibriSpeech less than ideal for text-to-speech work. The released corpus consists of 585 hours of speech data at 24kHz sampling rate from 2,456 speakers and the corresponding texts. Experimental results show that neural end-to-end TTS models trained from the LibriTTS corpus achieved above 4.0 in mean opinion scores in naturalness in five out of six evaluation speakers. The corpus is freely available for download from this http URL.
Subfields: Artificial Intelligence, Speech & Audio, Signal Processing
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
In Speech & Audio
In Signal Processing
Citations in the corpus
- Neural Codec Language Models are Zero-Shot Text to Speech Synthesizers (2023, neutral): “Compared to previous TTS training datasets, such as LibriTTS [17], our data contain more noisy speech but provide diverse speaker timbres and prosodies.”
- SoundStream: An End-to-End Neural Audio Codec (2021, methodology): “For clean speech, we use the LibriTTS dataset [50].”
- FunCodec: A Fundamental, Reproducible and Integrable Open-Source Toolkit for Neural Speech Codec (2023, methodology): “For academic usage, we employ the commonly-used LibriTTS corpus [27] to train and evaluate models.”
- UniCATS: A Unified Context-Aware Text-to-Speech Framework with Contextual VQ-Diffusion and Vocoding (2023, methodology): “Our experiments conducted on the LibriTTS dataset (Zen et al. 2019) demonstrate that CTX-vec2wav outperforms HifiGAN and Audi-oLM in terms of speech resynthesis from semantic tokens.”
- StyleTTS 2: Towards Human-Level Text-to-Speech through Style Diffusion and Adversarial Training with Large Speech Language Models (2023, methodology): “When trained on a large number of speakers like the LibriTTS dataset [14], StyleTTS 2 demonstrates potential for speaker adaptation.”
- Vocos: Closing the gap between time-domain and Fourier-based neural vocoders for high-quality audio synthesis (2023, methodology): “Data The models are trained on the LibriTTS dataset (Zen et al., 2019), from which we use the entire training subset (both train-clean and train-other ).”
- NaturalSpeech 2: Latent Diffusion Models are Natural and Zero-Shot Speech and Singing Synthesizers (2023, methodology): “Model Comparison We choose the previous zero-shot TTS model YourTTS [45] as the baseline, with the official code and pre-trained checkpoint4, which is trained on VCTK [44], LibriTTS [46] and TTS-Portuguese [47].”
- Grad-StyleSpeech: Any-Speaker Adaptive Text-to-Speech Synthesis with Diffusion Models (2022, methodology): “We train our model on the multi-speaker English speech dataset LibriTTS [21], which contains 110 hours audio of 1,142 speakers from the audiobook recordings.”
- AdaSpeech 4: Adaptive Text to Speech in Zero-Shot Scenarios (2022, methodology): “We train AdaSpeech 4 on LibriTTS dataset [25], which is a multi-speaker TTS corpus derived from LibriSpeech [26].”
- YourTTS: Towards Zero-Shot Multi-Speaker TTS and Zero-Shot Voice Conversion for everyone (2021, positive): “To further test its performance outside of the domain of VCTK, we select 10 speakers (5F/5M) from the subset testclean of the LibriTTS dataset [32].”
- Meta-TTS: Meta-Learning for Few-Shot Speaker Adaptive Text-to-Speech (2021, methodology): “However, Meta-TTS with shared speaker embedding obtains comparable results on LibriTTS and better on VCTK.”
- SpeechT5: Unified-Modal Encoder-Decoder Pre-Training for Spoken Language Processing (2021, methodology): “Dataset and Evaluation Metrics We use the 460-hours LibriTTS clean sets (Zen et al., 2019), a multispeaker corpus of read English speech from the audiobooks of the LibriVox project, as TTS training dataset.”
- A Survey on Neural Speech Synthesis (2021, neutral): “05 English LibriSpeech [253] 982 2484 16 English LibriTTS [421] 586 2456 24 English VCC 2018 [208] 1 12 22.”
- AdaSpeech: Adaptive Text to Speech for Custom Voice (2021, methodology): “…a large amount of high-quality and single-speaker recordings (Ito, 2017), and has been extended to multi-speaker scenarios (Gibiansky et al., 2017; Ping et al., 2018; Zen et al., 2019; Chen et al., 2020) using multi-speaker corpora (Panayotov et al., 2015; Veaux et al., 2016; Zen et al., 2019).”
- Make-A-Voice: Revisiting Voice Large Language Models as Scalable Multilingual and Multitask Learners (2024, positive): “• Librilight (Kahn et al., 2020) contains 60K hours of unlabelled speech from audiobooks in English, and LibriSpeech (Panayotov et al., 2015), Lib-riTTS (Zen et al., 2019), Gigaspeech (Chen et al., 2021), AISHELL (Fu et al., 2021), VCTK (Veaux et al., 2017) datasets include transcriptions.”
- ViSQOL v3: An Open Source Production Ready Objective Speech and Audio Metric (2020, background): “Projects like LibriTTS [28] have curated large amounts of freely available speech data, which has been a boon to speech-related DNNs, there is yet no standard and widely available subjective score dataset that is of similar scale.”