Home
Contact usLog inSign up
Release notes

Pronunciation Dictionaries for TTS

2, Apr 2026

Your voice agent has been mispronouncing your company name. Or medication. Or a customer's name. And every time you switched to a different voice, you had to fix it all over again.

Pronunciation dictionaries solve this at the source. Define how a word should be pronounced once, attach the dictionary to your assistant, and it applies across voices and providers. No re-tuning per voice. No per-provider workarounds.

What's new

  • Cross-provider dictionaries: One dictionary works across all supported TTS providers. Switch voices without losing your pronunciation rules.
  • Alias and phoneme support: Use simple text substitution (e.g., "Telnyx" to "Tel-nix") for any voice, or IPA notation for engines that support it (Telnyx Ultra, MiniMax, and Inworld).
  • PLS format import: Already have a pronunciation dictionary in ElevenLabs, Retell, or Vapi? Import it directly using the industry-standard PLS format.
  • Portal-based preview: Test pronunciation before going live. Pick a voice, model, and language, then hear exactly how it sounds.
  • Assistant attachment: Assign a dictionary in the Voice tab of your assistant configuration.
  • API access: Create, update, and manage dictionaries programmatically.

Why it matters

Mispronunciation breaks trust faster than latency does. A caller notices when your agent says "Ngen-X" instead of "Engine-X" every single time, or when a healthcare assistant stumbles on drug names. Pronunciation dictionaries let you fix this once and carry it forward.

Example use cases

  • Customer support: Support agents pronouncing customer and company names consistently across every call.
  • Healthcare: Assistants reading medication names, procedures, and acronyms correctly.
  • Financial services: IVRs handling product names, ticker symbols, and acronyms without mispronunciation.

Getting started

  1. Go to AI Suite in Mission Control Portal and open Pronunciation Dictionaries.
  2. Create a dictionary and add entries. Use alias format for most words. For engines that support it (Telnyx Ultra, MiniMax, Inworld), IPA phoneme entries work via PLS import.
  3. Select a provider, model, and voice to preview how it sounds.
  4. Save the dictionary.
  5. Open your assistant, go to the Voice tab, and assign the dictionary under Pronunciation Dictionary
  6. Test using the Sample Voice button or a live assistant flow.

Learn more in our developer documentation or contact your Telnyx team.