Breathy and Semi-Modal Phonation: High-Fidelity Human Vocal Preview

Harmonic Frontier Audio released this preview dataset (v0.9) in March 2026, providing under 1,000 high-fidelity human vocal recordings of breathy and semi-modal phonation. It serves as a technical reference for the Human Vocality Primitives series, showcasing specific formatting and metadata for expressive voice modeling.

Use Cases

• Training expressive text-to-speech models using breathy phonation samples
• Analyzing turbulent onsets for phonetic speech research
• Developing voice conversion algorithms for semi-modal phonation

Strengths

• High-fidelity audio quality designed for AI training
• Specialized labels for extended vocal techniques like turbulent onsets
• Standardized metadata conventions for expressive voice modeling

Limitations

• Small sample size of under 1,000 records
• Restricted to specific phonation types rather than general speech variety
• Language is categorized as undetermined (und)

Provenance

Source

Harmonic Frontier Audio

Freshness

Last updated March 2026.

Geography

United States