My text-to-speech API is mispronouncing names, what can I do?

If your text-to-speech (TTS) API is mispronouncing names, you usually have a few practical fixes:

1. Add pronunciation hints

Many TTS systems support:

• SSML (<phoneme>, <say-as>, <sub>)
• pronunciation lexicons or custom dictionaries
• override rules for specific words

Example with SSML:

<speak>
  Meet <phoneme alphabet="ipa" ph="ˈdʒoʊnˌsən">Johnson</phoneme>.
</speak>

2. Spell the name differently

If SSML isn’t supported, a phonetic spelling sometimes works:

• “Siobhán” → “Shi-vawn”
• “Nguyen” → “Win” or “Ngoo-yen” depending on the intended pronunciation

This is a workaround, but it can be effective.

3. Preprocess names in your app

Create a mapping of names to preferred pronunciations:

{
  "Siobhán": "Shi-vawn",
  "Nguyen": "Win"
}

Then substitute or annotate before sending text to TTS.

4. Use a custom voice or model

Some platforms let you:

• fine-tune a voice
• train pronunciation of names
• add user-specific vocabulary

5. Check language/locale settings

Mispronunciations often happen because the engine is using the wrong:

• language
• accent
• voice
• region

For example, an English (US) voice may handle names differently than English (UK).

6. Insert punctuation carefully

Commas, hyphens, or apostrophes can change prosody:

• “O’Connor” may be read better than “Oconnor”
• “Jean-Luc” may be read better than “Jean Luc”

7. Test with the exact voice

Different TTS voices pronounce the same word differently. Try:

• another voice in the same language
• a neural vs. standard voice
• male vs. female voices

If you tell me which TTS API you’re using, I can give you the exact syntax for fixing pronunciation.