Guides
Pronunciation, pauses and timing: the controls that matter
Ninety per cent of "this voice is bad" is one mispronounced name and a missing pause.
Last checked 2026-08-14
In short
Punctuation is your pause control: a comma is a beat, a full stop is a breath, a paragraph break is a rest. Names and acronyms go into the workspace pronunciation dictionary once, spelled the way they should sound, and every future render on the account inherits the fix. Length is controlled per line with speed, not by rewriting the whole script.
Punctuation is the pause control
It works in every engine, which no markup does. A comma gives a beat, a full stop a breath, a new paragraph a rest. Two full stops in a row do not double a pause; splitting the line does. If you need a specific silence - three seconds after a question in a training module - make it a separate line and leave a gap in the edit.
The pronunciation dictionary
Add the word and the spelling that sounds right (Naadly to Nard-lee), and it applies across the workspace: every voice, every project, every retry, every teammate, every language path that uses that word. Do it before a long render, not after. Common entries worth having on day one: your company and product names, your founders' names, the acronyms you want read as letters, and any place name your market cares about being said correctly.
Hitting a length
Video does not care how you feel about a sentence; it has 4.2 seconds. Three fixes in order of preference: cut words, slow or speed that one line, and only then change voice - a faster voice reads the same script fifteen per cent shorter, which is usually cheating rather than solving. If timing has to be exact, act the line yourself at the right length and convert it with the voice changer.
Numbers, dates and units
- Write what you want heard: twenty twenty-six or two thousand and twenty-six, not both from the same digits.
- Currency and units read better spelled out in narration (fifteen dollars a month).
- Phone numbers want spaces or dashes so they are read in groups.
- Ranges want the word: ten to twelve, not 10-12.
Questions
- Do you support SSML?
- Punctuation, per-line speed and the pronunciation dictionary cover what SSML is usually used for, and they behave identically across every engine, which SSML support does not.
- How do I make a voice say my brand name correctly?
- Add it to the pronunciation dictionary spelled the way it sounds. One entry fixes it everywhere on the workspace.
- Can I add a long pause?
- Split the line and leave the gap in your edit - it is exact, which a pause tag never is.