Voice

Account-wide voice settings shared across every agent, including pronunciations that control how words are spoken and recognized.

Voice settings apply across all your agents. Configure them at Settings > Voice.

Pronunciations

On the Settings > Voice tab, you can add custom pronunciations that apply to every Voice agent in your account. Each entry serves two purposes:

• Controls how the agent speaks the word. The pronunciation drives text-to-speech, so "SQL" → "sequel" or "kubectl" → "kube control" make the agent say specialty terms correctly.
• Strengthens how the agent hears the word. The source term is also added to the speech-to-text key term list, biasing recognition toward that word. This is especially helpful for specialty terms that sound like common English words. For example, mapping "Paychex" → "Paychecks" both makes the agent pronounce the product correctly and helps the model transcribe "Paychex" instead of "Paychecks" when a caller says it.

Add an entry for any product name, person name, acronym, or industry term that voice agents need to say or hear reliably. Pronunciations apply org-wide to every Voice agent. There is no per-agent override today.

The Pronunciations panel on Settings > Voice, showing example entries that apply to every Voice agent.

Use phonetic spelling: spell the word the way it should sound, and use spaces to separate syllables or letters (for example, "A P I" for the acronym "API").