LibriTTS-R Filtered Speaker Descriptions: Annotated English Speech Corpus

An annotated version of the LibriTTS-R corpus, which is a multi-speaker English speech dataset. The dataset provides approximately 960 hours of read English speech at a 24kHz sampling rate, originally published in 2019. It was annotated by parler-tts and updated on the Hugging Face platform in August 2024.

Use Cases

• Training controllable TTS models based on natural language speaker descriptions.
• Researching the relationship between acoustic properties and descriptive speaker characteristics.
• Developing voice cloning or voice conversion systems using annotated speaker data.
• Benchmarking speech synthesis quality on a large-scale, annotated English corpus.

Strengths

• Approximately 960 hours of English speech data provides a substantial audio foundation.
• Audio is sampled at 24kHz, which suggests high sound quality for speech synthesis tasks.
• Includes natural language annotations describing speaker and utterance characteristics.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

An annotated version of the LibriTTS-R corpus, which itself is derived from the LibriTTS corpus.

Collection Method

Annotations were generated using the Data-Speech tool, as referenced in the description.

Freshness

Last updated 2024-08-08 12:56:46; freshness should be verified.