Hi-Fi Multi-Speaker English TTS Dataset
arXiv:2104.01497
Abstract
This paper introduces a new multi-speaker English dataset for training text-to-speech models. The dataset is based on LibriVox audiobooks and Project Gutenberg texts, both in the public domain. The new dataset contains about 292 hours of speech from 10 speakers with at least 17 hours per speaker sampled at 44.1 kHz. To select speech samples with high quality, we considered audio recordings with a signal bandwidth of at least 13 kHz and a signal-to-noise ratio (SNR) of at least 32 dB. The dataset is publicly released at http://www.openslr.org/109/ .
References in corpus (14)
- HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis
- Libri-Light: A Benchmark for ASR with Limited or No Supervision
- Transfer Learning from Speaker Verification to Multispeaker Text-To-Speech Synthesis
- Common Voice: A Massively-Multilingual Speech Corpus
- NeMo: a toolkit for building AI applications using Neural Modules
- Deep Voice 3: Scaling Text-to-Speech with Convolutional Sequence Learning
- AdaSpeech: Adaptive Text to Speech for Custom Voice
- CTC-Segmentation of Large Corpora for German End-to-end Speech Recognition
- Citrinet: Closing the Gap between Non-Autoregressive and Autoregressive End-to-End Models for Automatic Speech Recognition
- MultiSpeech: Multi-Speaker Text to Speech with Transformer
- Multi-Speaker End-to-End Speech Synthesis
- TalkNet: Fully-Convolutional Non-Autoregressive Speech Synthesis Model
- MelGlow: Efficient Waveform Generative Network Based on Location-Variable Convolution
- NeMo Inverse Text Normalization: From Development To Production