15 datasets found

Unified codec language models for speech recognition, synthesis, and translation.

Neural codec language models are zero-shot text to speech synthesizers.

A large language model that can speak and listen.

A comprehensive collection of contemporary codec models, resulting in the creation of the CodecFake dataset.

Minimally-supervised speech synthesis method based on diffusion models with minimal supervision. Introduces the CTAP method as an intermediate semantic representation and uses...

A text-to-speech synthesis system typically consists of multiple stages, such as a text analysis frontend, an acoustic model and an audio synthesis module.

Text-to-speech (TTS) models have achieved remarkable naturalness in recent years, yet like most deep neural models, they have more parameters than necessary. Sparse TTS models...

The CSTR VCTK Corpus is a dataset of speech recordings of 109 speakers, each with 20 utterances.

The VCTK dataset is a large corpus of speech recordings, each containing a single speaker and a single sentence.

The LJSpeech dataset is a collection of audio recordings of a single female speaker reading aloud.

FastDiff is a fast conditional diffusion model for high-quality speech synthesis. It employs a stack of time-aware location-variable convolutions with diverse receptive field...

The LJ speech dataset is a dataset of speech samples recorded from a single speaker reading passages from 7 non-fiction books.

The Hi-Fi Multi-Speaker English TTS dataset is used to generate training, validation and test inputs for the audio splicing detection and localization task.

The dataset used in the paper is the LibriSpeech dataset, which contains about 1,000 hours of English speech derived from audiobooks.

A popular text-based VC approach is to use an automatic speech recognition (ASR) model to extract phonetic posteriorgram (PPG) as content representation.