Korean Single Speaker Speech Dataset for Text-to-Speech

The KSS Dataset is a Korean text-to-speech dataset consisting of audio files recorded by a professional female voice actress, with aligned text extracted from books. The dataset is the first publicly available speech dataset for Korean, released by the copyright holder.

Use Cases

• Train a text-to-speech model on Korean audio files with aligned text for speech synthesis
• Fine-tune a speech synthesis model using expert-generated audio recordings from a single professional speaker
• Develop Korean speech processing applications using a monolingual dataset of aligned audio and text pairs

Strengths

• First publicly available speech dataset for Korean
• Audio recordings created by a professional female voice actress
• Text annotations and audio files are expert-generated
• Dataset size is categorized as 10K<n<100K samples

Limitations

• Dataset contains only a single speaker, limiting voice diversity
• Text source is limited to specific books, potentially restricting vocabulary coverage
• Last updated in July 2022, which may not reflect current speech patterns

Provenance

Source

Bingsu

Collection Method

Audio files recorded by a professional female voice actress with text extracted from books

Freshness

Last updated 2022-07-02

Geography

Korean language dataset