VCTK

The VCTK Corpus is a publicly available dataset designed for research in speech synthesis and voice conversion. It was created by the Centre for Speech Technology Research (CSTR) at the University of Edinburgh. The corpus consists of speech recordings from 109 native English speakers, each reading out a set of 400 sentences. These sentences are designed to cover a wide range of phonetic contexts, making the corpus particularly useful for training and evaluating speech synthesis models.

The VCTK Corpus is notable for its high-quality recordings, which were made in a professional studio environment. The recordings are sampled at 48 kHz with 16-bit resolution, ensuring a high level of audio fidelity. The dataset also includes additional metadata, such as speaker demographics and phonetic transcriptions, which can be useful for various speech processing tasks.

The corpus has been widely used in the research community for tasks such as text-to-speech synthesis, voice conversion, and speaker identification. Its comprehensive coverage of phonetic contexts and high-quality audio make it a valuable resource for developing and benchmarking speech technology systems. The VCTK Corpus is freely available for download, encouraging its use in both academic and commercial applications.