Pronunciation lexicon

Invented names, places and jargon are where text-to-speech stumbles. The lexicon swaps a term for a respelling in what's sent to the speech model, while your script keeps the original spelling.

Adding an entry

Open the project's Settings → Pronunciation lexicon and add a term and its replacement:

Write the replacement the way a voice actor would read it off a script: plain syllables, with capitals on the stressed one. The respelling is read with the pronunciation rules of the line's language.

How matching works

• Whole words only. Xyl matches in Xyl's blade but not in Xylophone.
• Case-insensitive by default. Tick case-sensitive to match only the exact capitalization, so the acronym US doesn't change the word us.
• Apostrophes are interchangeable. ', ’ and ʼ all match each other, so curly quotes from a word processor don't break an entry.
• Longest term wins. With entries for Ashfall and Ashfall Keep, the text Ashfall Keep uses the longer one. Each part of the text is replaced at most once, so one replacement is never rewritten by another entry.
• Tags are never touched. Text inside vocal tags (<sigh>) and backchannels (|mhm|) is left alone.

Per-language entries

An entry applies to every language unless you choose one. In a multilingual project, add language-specific respellings where it matters: an entry for a specific language replaces the all-languages entry for the same term in that language.

Effect on renders

Adding, editing or removing an entry marks every rendered line whose spoken text changes as stale, in the affected languages only. Run a Stale and draft render to apply it.