General notes about synthesized voices

Synthesized voices offer a helpful reference when a human recitation isn’t available, giving singers an accessible way to hear text spoken aloud across many languages.

These samples often reflect spoken-style synthetic voice output, even though we prompt the AI to incorporate lyric-diction practices where possible. Spoken and sung diction often differ—especially in vowel length, resonance, and consonants such as R—so the results may not fully capture the nuances of classical sung pronunciation.

No current synthesized-voice system fully supports all Latin or regional pronunciation variants, and output may vary accordingly. SongBase also allows users to attempt synthesis in additional languages, though results will vary depending on phonetics and the system’s language support.

At this time, SongBase supports left-to-right (LTR) languages only; right-to-left (RTL) languages such as Arabic, Hebrew, and Farsi are not yet supported.