Pronunciation Best Practices

Our voice model is pretty good at predicting pronunciation, but a few things can trip it up: unfamiliar or rare nouns (i.e., company names, words borrowed from other languages, or specialized terminology), homonyms (read vs. read), acronyms, and numbers. With the following best practices, you can get words spoken just the way you want. 

Acronyms as words

For an acronym as a word, try spelling the word the way it sounds. For example, FEMA sounds better when you enter it as "feema" or "::FEE-ma::"

Word Pronunciation

When the voice avatar needs help predicting a word, we can provide it with a Replacement spelling to get the correct pronunciation. Actually, you can replace any set of characters with another set of characters, not just regular dictionary words.

You can add Replacements to your script in one of the following ways:

1. Respellings is a unique way to format a word by breaking down each syllable and which syllables should be emphasized.
   
   Example: For edamame, you can include "ed-uh-MAW-may or"  "try-yuh-uh-doo-THY-ruh-neen" for triiodothyronine.
2. Creative spellings is an alternative way of spelling a word in order to prompt the voice avatar to give the performance you're looking for.
   
   Example: For edamame, you can include "edamawmay" or "try idoe thighro neen" for triiodothyronine.

Emphasize a word or phrase

Add quotation marks ("") around a word or phrase to direct the voice avatar to pay particular attention. Learn how to get the right inflection.