SUPERSEDED - CSTR VCTK Corpus: English Multi-speaker Corpus for CSTR Voice Cloning Toolkit

By Christophe Veaux, Junichi Yamagishi, Kirsten MacDonald. 2016

1,222 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Evaluation & Benchmarks

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 Evaluation & Benchmarks

Citations in the corpus

Neural Discrete Representation Learning (2017, methodology): “We ﬁrst consider the VCTK dataset, which has speech recordings of 109 different speakers [41].”
WaveNet: A Generative Model for Raw Audio (2016, methodology): “We used the English multi-speaker corpus from CSTR voice cloning toolkit (VCTK) (Yamagishi, 2012) and conditioned WaveNet only on the speaker.”
Neural Codec Language Models are Zero-Shot Text to Speech Synthesizers (2023, neutral): “, 2022b] as the baseline, which is trained on a combined dataset of VCTK [Veaux et al., 2016], LibriTTS [Zen et al.”
Transfer Learning from Speaker Verification to Multispeaker Text-To-Speech Synthesis (2018, background): “VCTK [21] contains 44 hours of clean speech from 109 speakers, the majority of which have British accents.”
Dilated Recurrent Neural Networks (2017, methodology): “In this section, we evaluate the performance of D ILATED RNN on four different tasks, which include long-term memorization, pixel-by-pixel MNIST classiﬁcation [15], character-level language modeling on the Penn Treebank [16], and speaker identiﬁcation with raw waveforms on VCTK [26].”
FedSpeech: Federated Text-to-Speech with Continual Learning (2021, methodology): “We conduct experiments on the VCTK dataset [Veaux et al. , 2017], which contains approximate 44 hours of speech ut-tered by 109 native English speakers with various accents.”
VALL-E 2: Neural Codec Language Models are Human Parity Zero-Shot Text to Speech Synthesizers (2024, methodology): “We use LibriSpeech test-clean [Panayotov et al., 2015] and VCTK [Veaux et al., 2016] for zero-shot TTS evaluation, ensuring none of the speakers from these corpora are included in the training data.”
RepCodec: A Speech Representation Codec for Speech Tokenization (2023, methodology): “For VCTK, we follow the model architecture and toolkit provided in Polyak et al. (2021).”
High-Fidelity Audio Compression with Improved RVQGAN (2023, methodology): “For speech, we use the DAPS dataset [25], the clean speech segments from DNS Challenge 4 [10], the Common Voice dataset [2], and the VCTK dataset [39].”
Mega-TTS: Zero-Shot Text-to-Speech at Scale with Intrinsic Inductive Bias (2023, methodology): “And in terms of speaker similarity, Mega-TTS signiﬁcantly outperforms YourTTS with +0.51 MOS-S on VCTK and +0.68 MOS-S on LibriSpeech, demonstrating the effectiveness of Mega-TTS in zero-shot scenarios.”
Audiodec: An Open-Source Streaming High-Fidelity Neural Audio Codec (2023, methodology): “The Valentini dataset [35], which is derived from the VCTK corpus [36], is adopted for the evaluations.”
A3T: Alignment-Aware Acoustic and Text Pretraining for Speech Synthesis and Editing (2022, positive): “(2021), we conduct our speechediting experiments with a single-speaker TTS dataset LJSpeech (Ito & Johnson, 2017) and a multi-speaker TTS dataset VCTK (Yamagishi et al., 2019).”
YourTTS: Towards Zero-Shot Multi-Speaker TTS and Zero-Shot Voice Conversion for everyone (2021, methodology): “, 2021) achieved promising results with only 11 speakers from the VCTK dataset (Veaux et al., 2016), when one limits the number and variety of training speakers, it also further hinders the model generalization for unseen voices.”
A Survey on Neural Speech Synthesis (2021, neutral): “ARCTIC [169] 7 7 16 English VCTK [363] 44 109 48 English Blizzard-2011 [161] 16.”
Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech (2021, methodology): “We used the LJ Speech dataset (Ito, 2017) for comparison with other publicly available models and the VCTK dataset (Veaux et al., 2017)”
Speech Resynthesis from Discrete Disentangled Self-Supervised Representations (2021, methodology): “We employ two datasets: LJ [42] single speaker dataset and VCTK [43] multi-speaker dataset.”
AdaSpeech: Adaptive Text to Speech for Custom Voice (2021, methodology): “Varying Adaptation Data We study the voice quality with different amount of adaptation data (fewer than the default setting) on VCTK and LJSpeech, and conduct MOS evaluation as shown in Figure 4b.”