How to fix pronunciation in text-to-speech (names, acronyms, brand terms)

Every TTS engine eventually mangles something — a founder's name, an acronym read as a word instead of letters, a brand term with an unusual spelling. Retyping the word phonetically every time you write a new script doesn't scale, and it looks wrong on the page even if it sounds right.

The "sounds like" dictionary

GetMeAudio's Pronunciation panel is a saved list of rules: *this word* → *say it like this*. Add "UPI" → "U P I" once, or "Swasti" → "Swaa-sthee," and every script you write afterward — this one and every future one — uses the correction automatically. You don't edit the visible text at all; the correction is layered on underneath.

How it applies

• Whole-word matching by default, so fixing "read" as a name doesn't also rewrite "read" the verb elsewhere — you can also allow partial or case-sensitive matches per rule if a term needs it.
• Works the same across Draft, Preview, and Download — the device voice (Draft) gets a literal text substitution, and the cloud voices (Preview/Download) get an SSML alias that keeps the correct written word while changing only what's spoken.
• Unicode-aware, so whole-word matching works correctly inside Hindi, Tamil, Telugu, and other non-Latin scripts too, not just ASCII text.
• Rules are saved to your account, not the project — fix "Swasti" once and it's correct in every script you write from then on.

Where to start

Open Pronunciation from the script toolbar, add a rule for anything that came out wrong in a Draft or Preview play, and re-play — you'll hear the fix immediately without touching your script's wording.