Arabic Professional Voice: Single-Speaker TTS Dataset with Full Diacritics

A high-quality Arabic Text-to-Speech dataset recorded by a professional male speaker. It contains 439 utterances in Modern Standard Arabic, all transcriptions include full Tashkeel (diacritical marks). The dataset was created by NightPrince and last updated on Hugging Face in March 2026.

Use Cases

• Train neural TTS models based on the high-quality, single-speaker audio recordings.
• Develop Arabic speech synthesis systems based on transcriptions with full diacritical marks.
• Benchmark TTS model performance on Modern Standard Arabic based on the professional speaker recordings.

Strengths

• Contains 439 high-quality audio utterances recorded by a professional speaker.
• All transcriptions include full Tashkeel (diacritical marks), reducing the need for text normalization.
• Audio is recorded at a 16 kHz sampling rate, suitable for TTS model training.

Limitations

• Dataset is limited to a single professional male speaker, which may not represent diverse vocal characteristics.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale training.

Provenance

Source

Hugging Face dataset by author NightPrince.

Collection Method

Recorded by a professional speaker.

Freshness

Last updated 2026-03-02 19:03:39; freshness should be verified.