Rules for writing numbers in text-to-speech scripts
Speech models interpret raw digits inconsistently. By following three baseline rules, you will never get a mispronounced number in your voice tracks:
Rule 1
Spell out years as pairs
Write "nineteen ninety-two" instead of "1992", and "twenty twenty-four" instead of "2024".
Rule 2
Add currency words
Never leave lone $ or € symbols. Replace "$25" with "twenty-five dollars" and "$25.50" with "twenty-five dollars and fifty cents".
Rule 3
Separate phone digits
Add commas or spaces between phone groups so the AI breathes naturally between numbers.
Common examples & quick cheatsheet
| Raw Expression | How TTS often reads it | Recommended phonetic fix |
|---|---|---|
| 1999 | "One thousand nine hundred ninety-nine" | "Nineteen ninety-nine" |
| $350.25 | "Dollar sign three fifty point two five" | "Three hundred fifty dollars and twenty-five cents" |
| 2nd | "Two nd" | "Second" |
| 3/4 | "Three slash four" | "Three fourths" |
| 1-800-555-0199 | "One eight hundred five million..." | "One, eight hundred, five five five, zero one nine nine" |
