Pronunciations

What it does

Pronunciations let you correct how your AI receptionist says specific words — and how well it recognizes those words when callers say them.

Two things happen when you add a pronunciation:

1. Your receptionist says it correctly. When the AI is about to speak a word you’ve configured, it substitutes your spelling before generating audio. If you tell it that “Nguyen” should be pronounced “Win,” it will say “Win” every time — in greetings, when looking up a contact, whenever that word comes up.

2. Your receptionist hears it correctly. The speech-to-text engine gets a heads-up about your words before each call starts, so it is less likely to mishear “Kowalski” as “Kawalsky” when a caller asks to speak with her.

Both improvements happen automatically once you save a pronunciation. No further action required.

How to get there

1. Click Agent in the left sidebar.
2. Click the Pronunciations tab.

Step-by-step

Adding a pronunciation

At the bottom of the Pronunciations tab, you will see two text inputs:

• Word — the word exactly as it is spelled in your contact list or business name (e.g., Chike)
• Pronunciation — how that word should sound when spoken aloud (e.g., CHEE-kay)

Fill in both fields. A few things to keep in mind as you type the pronunciation:

• Write it as a single word, the way it sounds — not as a sentence or a description
• Use capital letters to stress a syllable (CHEE-kay, not chee-kay)
• Avoid hyphens or spaces between syllables — they create unnatural pauses in the audio
• For a soft “uh” sound, write uh or a lowercase a

Once you are happy with the spelling, click the speaker icon button next to the inputs to hear a quick preview. If it sounds right, click Add. The entry appears in the list above and saves automatically.

Using Find Pronunciation (recommended for tricky names)

If you are not sure how to spell a pronunciation phonetically, click Find Pronunciation instead. This opens a four-step guided dialog that uses AI to help you find the right sound and spelling.

Step 1 — Enter the word and a phonetic guess. Type the word and your best guess at how it sounds. You do not need to get this right — it is just a starting point. Click Generate options.

Step 2 — Choose the right sound. The dialog shows up to five options, each with a technical pronunciation code and a plain-English description of the sounds involved. Click the speaker button on any card to hear a preview. When you find one that matches how the name should actually sound, select it and click Use this pronunciation.

Step 3 — Test spelling variants. The dialog now shows up to four ways to spell that pronunciation as plain text — your original guess plus three alternatives. Each card has three numbered speaker buttons (labeled 1, 2, 3) so you can play the same variant multiple times and hear whether it sounds consistent. Select the spelling that sounds best and click Use this alias.

Step 4 — Confirm. You see a summary: “When the AI sees [word], it will say [pronunciation].” Play it one more time to confirm, then click Save. The entry is added to your list and saved automatically.

Editing an existing pronunciation

Each row in the pronunciations list has three buttons on the right side:

• Speaker icon — plays a quick audio preview of how the pronunciation currently sounds
• Pencil icon — opens the Find Pronunciation dialog pre-filled with the existing word and pronunciation, so you can adjust it using the same guided steps
• Trash icon — removes the entry permanently (takes effect on the next call)

Reading the list

The list shows three columns:

Tips

Start with your business name and employee names. These appear most often in conversation and callers are most likely to notice when they sound wrong.

Unusual spellings are the most important ones to configure. Common words like “Smith” do not need pronunciation entries. Focus on names or terms that a voice engine would not know how to say from spelling alone — foreign names, invented brand names, uncommon abbreviations.

Capitalize the stressed syllable. For a name like “Azumo” pronounced “Ah-ZOO-mo,” write ah-ZOO-mo. The AI voice engine responds to capitalization as a stress cue.

Use Find Pronunciation for names you are uncertain about. If you cannot hear the difference between two phonetic spellings in your head, the guided dialog lets you listen to both before committing.

Play the preview multiple times. AI voice synthesis can sound slightly different on repeated plays. The three numbered preview buttons in Step 3 of Find Pronunciation exist for this reason — use all three to make sure the result is consistent before you save.

Changes apply to the next call, not the current one. If you correct a pronunciation while a call is in progress, the in-progress call is not affected. The new setting takes effect when the next call begins.

Your voice settings affect how pronunciations sound. If a pronunciation preview sounds flat or robotic, try adjusting your voice speed or expressiveness in the Tuning page and then re-preview.

Troubleshooting

I added a pronunciation but the AI still says the word wrong. Pronunciations take effect at the start of the next call. If you added or changed a pronunciation during an active call, you will not hear the update until a new call comes in. Wait for the next call and test again.

The pronunciation sounds almost right but not quite — how do I fine-tune it? Open the entry using the pencil icon and go back through the Find Pronunciation dialog. In Step 3, try a different spelling variant, or go back to Step 2 and choose a different sound option. Small changes like doubling a vowel (ee vs e) or adding a silent e at the end can shift how the voice engine interprets the syllable.

The preview audio is not playing. Make sure your browser is not blocking audio. Try clicking the speaker button again. If it still does not play, try Chrome or Safari. If you have clicked the speaker button more than 10 times in the last minute, the system will briefly pause previews to prevent overuse — wait a minute and try again.

I added a pronunciation but the word column shows a duplicate warning. Each word can only have one pronunciation entry. If you try to add a word that is already in the list, the system will reject it silently. Edit the existing entry instead of creating a new one.

The AI mishears a name when callers say it, even though I have a pronunciation configured. The pronunciation entry boosts recognition accuracy but does not guarantee it. Accuracy is also affected by call audio quality, background noise on the caller’s end, and how the name is spoken. If mishearing continues, confirm the word is spelled exactly as it appears in your contact names, and check whether the contact’s name itself is listed in your Contacts — contact names get their own recognition boost automatically, separate from the pronunciation list.

I cannot find the Pronunciations tab. The tab is on the Agent page, not the Settings page. Click Agent in the left sidebar (it has a robot icon), then look for the Pronunciations tab at the top of the page.

Related

• Agent Configuration — configure your receptionist’s greeting, name, voice, and special instructions
• Tuning — adjust voice speed and expressiveness, which affects how pronunciations sound
• Contacts — contact names receive automatic recognition boosting without a pronunciation entry
• Knowledge Base — give your receptionist answers to common questions