CSTR VCTK Corpus: English Multi-speaker Corpus for CSTR Voice Cloning Toolkit (version 0.92)

By Junichi Yamagishi, Christophe Veaux, Kirsten MacDonald. 2019

849 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Evaluation & Benchmarks, Speech & Audio

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

In Speech & Audio

Citations in the corpus

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.”
SoundStream: An End-to-End Neural Audio Codec (2021, methodology): “We evaluate the different models using the VCTK dataset [62], which was neither used for training SoundStream nor SEANet.”
Voicebox: Text-Guided Multilingual Universal Speech Generation at Scale (2023, methodology): “The model consists of 6 blocks with hidden dimension sizes: [64, 128, 256, 256, 512, 512].”
StyleTTS 2: Towards Human-Level Text-to-Speech through Style Diffusion and Adversarial Training with Large Speech Language Models (2023, background): “Furthermore, it attains human-level performance on the multispeaker VCTK dataset [13] in terms of naturalness (CMOS = −0.”
Grad-StyleSpeech: Any-Speaker Adaptive Text-to-Speech Synthesis with Diffusion Models (2022, methodology): “We also use the VCTK [22] dataset which includes 110 English speakers for evaluation of the unseen speaker adaptation capability.”
AdaSpeech 4: Adaptive Text to Speech in Zero-Shot Scenarios (2022, methodology): “For VCTK and LibriTTS, we average the MOS and SMOS scores of multiple speakers as the ﬁnal scores.”
A3T: Alignment-Aware Acoustic and Text Pretraining for Speech Synthesis and Editing (2022, positive): “We use 24K sampling rate for VCTK and 22K for LJSpeech.”
Neural Voice Cloning with a Few Samples (2018, methodology): “7 Voice cloning is performed on VCTK dataset [Veaux et al., 2017].”
BigVGAN: A Universal Neural Vocoder with Large-Scale Training (2022, neutral): “In Appendix E, we also report the additional results including UnivNet (Jang et al., 2021) and the ablation models of BigVGAN-base on LibriTTS dev sets, unseen VCTK (Yamagishi et al., 2019), and LJSpeech (Ito, 2017) data.”
FastDiff: A Fast Conditional Diffusion Model for High-Quality Speech Synthesis (2022, methodology): “To evaluate the generalization ability of our model over unseen speakers in multi-speaker scenarios, we also used the VCTK dataset [ Yamagishi et al. , 2019 ] , which was downsampled to 22050 Hz to match the sampling rate with the LJSpeech dat-set.”
Voice Synthesis for in-the-Wild Speakers via a Phonological Loop (2017, methodology): “First, we employ a subset of the VCTK dataset [20] that contains 21 American speakers, both male and females.”
Bilateral Denoising Diffusion Models (2021, positive): “In addition to the single-speaker speech synthesis, we evaluated BDDMs on the multi-speaker speech synthesis benchmark VCTK [40].”
Meta-StyleSpeech : Multi-Speaker Adaptive Text-to-Speech Generation (2021, methodology): “We train the classiﬁer on VCTK dataset and the classiﬁer achieve 99% accuracy for validation set.”
Expressive Neural Voice Cloning (2021, methodology): “We follow the speaker encoder architecture described in Wan et al. (2017); Louppe (2019b).”
NExT-GPT: Any-to-Any Multimodal LLM (2023, background): “302 Table 10: Text+audioto-audio generation (textconditioned speech editing) results on VCTK data [83].”
Emotional Voice Conversion: Theory, Databases and ESD (2021, methodology): “However, existing databases, such as VCTK database (Yamagishi et al., 2019), CMU-Arctic database (Kominek and Black, 2004), and Voice Conversion Challenge (VCC) corpus (Toda et al., 2016; Lorenzo-Trueba et al., 2018b; Zhao et al., 2020), are designed for speaker voice conversion.”