TTS Number Pronunciation

Most TTS models rely on normalization rules that guess context. When an engine encounters "1995," it often guesses "one thousand nine hundred ninety-five" instead of the year "nineteen ninety-five." Writing numbers out phonetically removes this ambiguity.

Write the year as two pairs of words. For example, write "nineteen ninety-five" instead of "1995", and "twenty twenty-four" instead of "2024".

Spell out both the main unit and fractional cents. For example, rewrite "$49.99" as "forty-nine dollars and ninety-nine cents" so the voice doesn't say "dollar sign forty-nine point nine nine."

Separate every digit with a space or hyphen, or spell out each digit ("five five five, zero one nine nine"). Grouping digits with commas adds natural conversational pauses.

Write "1/2" as "one half" or "half", "3/4" as "three fourths", and "15%" as "fifteen percent." This prevents engines from reading the slash literally as "one slash two."

Yes. Once you convert numbers into plain English words, every text-to-speech engine (including Kokoro, ElevenLabs, Amazon Polly, and browser speech) will read them with complete consistency.