Voice Library

The Voice Library is where you browse, preview, clone, and manage the voices your AI agents speak with. Every voice — whether it ships with a provider, was imported for your account, or was cloned from a recording — appears here in one searchable list, and from here you can assign any ready voice to an agent in a couple of clicks.

This guide explains everything a business user needs to know to find the right voice, listen to it, create a custom cloned voice from your own recording, and put that voice to work on your agents.

In this guide

• Where to find the Voice Library
• Plan requirements
• The Voice Library at a glance
• Browsing voices
• Filtering and searching
• Previewing a voice
• Favourites
• Assigning a voice to an agent
• Choosing a voice inside the agent editor
• Voice cloning: the Clone Voice wizard
• Managing your voices
• Voice statuses explained
• Tips and common workflows
• Troubleshooting
• Quick reference

---

Where to find the Voice Library

Open the Voice AI hub and look at the left-hand sidebar. Under the Navigation section you will see Voice Library (its icon is a small audio waveform).

The Voice Library has two web addresses that both open the same page:

This is why clicking Clone Voice changes the address to …/cloning — it is a shortcut straight into the cloning flow. Closing the wizard returns you to …/library.

---

Plan requirements

The Voice Library (and voice cloning) is part of the Voice Cloning feature, which is included from the Growth plan upwards.

• If your Voice AI plan includes it, the Voice Library item appears in the sidebar and the page opens normally.
• If your plan is below Growth, opening the page shows an upgrade screen titled "Voice Cloning requires Growth" with an Upgrade to Growth button (and the current monthly price). You can compare plans from the same screen.
• If Voice AI has not been switched on for your account at all, you will see a "Voice AI not enabled" message asking you to contact your administrator.

---

The Voice Library at a glance

When the page loads you see a header, a row of filters, and a table of voices.

Header

Account administrators also see two extra controls in the header — Sync Providers and an Import panel. These pull in voices from connected providers and are described under Managing your voices. Regular users will not see them.

---

Browsing voices

Each voice is shown as a row in the table. From left to right, a row contains:

How voices are ordered

The library arranges voices to put the most useful ones first:

1. Favourites are pinned to the top in their own Favourites section (with a star and a count).
2. Below that, under All Voices, voices that already have a playable preview are listed before voices that don't.
3. Within each group, voices are interleaved by provider so you see a healthy mix rather than one provider's voices all bunched together.

Duplicate voices (same name from the same provider) are automatically collapsed so the list stays clean — you only ever see the most recent copy.

Empty states

• No voices at all: You'll see "No voices yet." If you're an account admin, a Clone Your First Voice button appears. Regular users are prompted to "Ask your admin to add voices to the library."
• No matches for your filters: You'll see "No voices match your filters" with a Clear all filters button to reset.

---

Filtering and searching

The filter bar sits directly under the header. All filters work together — the voice count on the right updates instantly to show how many voices match.

The filters

Using a dropdown filter

1. Click a filter pill (e.g. Language). A panel opens beneath it.
2. If the panel has a search box at the top (Language does), start typing to narrow the options.
3. Click an option to apply it. The pill turns solid white to show it's active and now displays the chosen value.
4. To remove that single filter, click the ✕ on its pill, or open it again and click the same option.

Clearing filters

• A Clear button (with a count, e.g. Clear (3)) appears whenever one or more dropdown/toggle filters are active. Clicking it removes all of them at once while keeping whatever you typed in the search box.
• The empty-state Clear all filters button resets everything, including search.

Note: The search box matches the voice name only. To narrow by accent, gender, or use case, use the matching dropdown rather than typing those words into search.

---

Previewing a voice

The round button on the left of every row is the preview control. Listening before you commit is the best way to pick the right voice.

Playing a sample

1. Click the ▶ play button on a voice row.
2. The button turns white and shows a pause icon; a thin progress bar appears along the bottom of the row.
3. Click anywhere on that progress bar to scrub to a different point in the sample.
4. Click the button again (now a pause icon) to stop.

Only one preview plays at a time — starting a new one automatically stops the previous voice.

When there's no sample yet

Some voices don't have a pre-recorded sample. What happens depends on the provider:

• Generate-on-demand providers (xAI, OpenAI, ElevenLabs, Google, Deepgram): the play button is still active. Click it and the platform generates a fresh sample for you. The button shows a spinner while it works, then plays automatically and keeps the sample for next time.
• Other providers / cloned voices without a stored sample: the play button is dimmed and cannot be clicked because no preview is available.

Failed voices

If a clone failed, its play button shows a warning triangle instead of a play icon and cannot be played. See Voice statuses explained for how to retry.

---

Favourites

Favourites let you pin the voices you reach for most to the top of the list.

• To favourite a voice: hover over its row and click the star on the right (it turns gold). On a voice you've already favourited, the star is always visible.
• To unfavourite: click the gold star again.
• To see only favourites: turn on the Favourites toggle in the filter bar.

Favourited voices are grouped into a dedicated Favourites section at the very top of the table, with a count, so they're always one glance away.

Favourites are stored in your own browser. They're personal to you and to the device/browser you're using — they don't change what other people in your account see, and they won't carry over to a different computer.

---

Assigning a voice to an agent

The fastest way to put a voice to work is the Assign button on its row. This sets the voice as the speaking voice for one of your agents without leaving the library.

The Assign button only appears on voices that are Ready. Voices still processing, pending, or failed can't be assigned until they finish.

Steps

1. Find the voice you want and click its ⚡ Assign button.
2. The Assign to Agent dialog opens, showing the voice name at the top.
3. Use the Search agents… box to find the agent you want, or scroll the list. Active agents are listed first; archived ones appear lower, labelled with their status.
4. Click an agent to select it — a green check appears next to it.
5. Click Assign to "[agent name]" at the bottom.
6. A confirmation appears: "[voice] assigned to [agent]." The dialog closes.

What assigning actually does

Behind the scenes, assigning updates that agent's voice engine settings so it uses the selected voice (and the matching provider). If the agent was previously set to a different provider, the platform handles switching it over for you. The next time the agent speaks — on a test call or a live call — it will use the new voice.

Assigning replaces whatever voice the agent had before. If you want to keep the old voice as an option, note it down first, or use a separate agent.

---

Choosing a voice inside the agent editor

You can also set a voice from within an agent itself, which is handy when you're building or tweaking an agent. Open an agent and find the Voice Engine section.

How the voice control behaves depends on the provider you've chosen for the agent:

1. Pick a Voice Engine (provider/model) from the first dropdown in the section.
2. The voice control underneath adapts:
   • For providers with a fixed set of built-in voices, you get a simple Select voice dropdown.
   • For providers backed by your library, you get a rich voice picker with a search box, avatars, gender/accent/age details, and a play button on every row so you can preview each voice right there before choosing.
3. Pick a voice. Your choice is saved to the agent's configuration.

Cloned and library voices appear here automatically once they're Ready and match the agent's selected provider. If you don't see a voice you expect, check that the agent's Voice Engine provider matches the voice's source.

---

Voice cloning: the Clone Voice wizard

Cloning lets you create a custom voice from a real recording — your own voice, a colleague's, or any clean voice sample you have the rights to use. The whole process is handled by a guided four-step wizard.

Open it by clicking Clone Voice in the header (or by going straight to the /cloning address, which opens the wizard automatically). Press Esc or click outside the wizard to close it at any time.

Step 1 — Name your voice

Fill in the details for your new voice:

These details are what later power the filters, so filling them in makes the voice easier to find. When you've entered at least a name, click Continue.

Step 2 — Add your voice

Here you provide the actual audio. There are two modes, chosen with the cards at the top: Record or Upload File.

Record mode (guided teleprompter)

The recorder walks you through a script of short phrases that cover a range of emotional tones — neutral, happy, sad, firm, surprised, nervous, and professional — so the clone captures the full character of the voice.

1. Click Start Recording. Your browser will ask permission to use your microphone — click Allow. (If you deny it, you'll see a message asking you to enable microphone access.)
2. Read the large phrase shown in the centre out loud, in the tone shown by the small coloured label above it (e.g. Happy, Professional).
3. The recorder listens and, when it recognises that you've read the phrase, automatically advances to the next one. The faint lines above and below show the previous and upcoming phrases.
4. If a phrase isn't recognised or you'd rather move on, click Skip to jump to the next one.
5. A progress bar and an "X / 20" counter track how far through the script you are.
6. When you finish (or click Stop), you'll see a green "X phrases captured" badge. If you want a fresh take, click Re-record to start over.

The phrase-recognition that auto-advances uses your browser's built-in speech recognition. It works best in Chrome. Even if recognition isn't available, you can still record — just use Skip to move through the phrases manually, and the full audio is still captured.

Upload mode

Prefer to use an existing recording? Switch to Upload File.

1. Click the dashed drop area, or drag an audio file onto it.
2. Accepted formats are MP3, WAV, M4A, OGG, and WEBM, up to 50 MB.
3. Once a file is loaded, a small player appears so you can ▶ play it back and confirm it's the right clip. An ✕ removes it if you picked the wrong one.

The wizard shows tips for the best result:

• Use at least 30 seconds of clear, natural speech.
• Record in a quiet room with minimal background noise.
• Speak at a consistent volume and pace.
• Avoid music, echo, or heavy audio compression.

When you have a recording or an uploaded file, click Continue.

Step 3 — Review and clone

A summary screen lists everything you've entered — name, description, visibility, gender, age, language, use case — plus the audio source and its size.

1. Check the summary. To change anything, use Back.
2. Click Clone Voice to submit.
3. The wizard shows a "Processing your voice…" state. Cloning typically takes a short while as the platform builds your voice. The wizard polls automatically and moves on as soon as it's done.

Step 4 — Voice ready

When the clone succeeds you'll see a tick and "Voice ready — [name] has been cloned successfully." Click View in Library to return to the list, where your new voice now appears at the top, ready to preview and assign.

If cloning fails, you'll get a message explaining why (for example, the recording was too noisy, the format wasn't supported, or no voice provider is configured for your account). Try again with a cleaner sample in a quiet environment, or contact your administrator if the message mentions a missing provider.

---

Managing your voices

Editing a voice's details

If you're an account administrator, hovering over a voice row reveals a pencil (edit) icon. Clicking it opens a compact Edit Voice dialog where you can:

• Change the voice's Name.
• Switch its Visibility between Private and Public.

Click Save to apply, or Cancel to discard. The row updates immediately.

Deleting a voice

Account administrators can also delete voices. Hover over a row and click the trash icon, which appears for private voices (public/shared voices are protected from accidental deletion). The voice is removed from the library straight away.

Caution: Deleting a voice can't be undone. If any agents were using it, you'll want to assign them a replacement voice.

Retrying a failed clone

If a clone shows as Failed, a Retry link appears on its row.

• Click Retry to re-run the clone using the original recording. The status changes to Retrying… and updates automatically when it finishes.
• If the original recording is no longer on file, the platform clears the broken entry and re-opens the Clone Voice wizard so you can record a fresh sample without any manual cleanup.

Admin-only library tools

Two header controls are only shown to account administrators and are worth knowing about:

If you're not an admin and your library looks empty, this is usually why — ask your account administrator to sync or import voices for you.

---

Voice statuses explained

A voice's status controls what you can do with it.

---

Tips and common workflows

Find a voice fast Combine filters: choose a Language, then its Accent, then a Gender, and watch the voice count drop to a short list. Preview the survivors and favourite the winner.

Build a shortlist Star a handful of candidate voices, then flip on the Favourites toggle to compare just those side by side before deciding.

Clone, then assign in one sitting

1. Click Clone Voice and complete the wizard.
2. On the success screen, click View in Library.
3. Your new voice is at the top — click ▶ to confirm it sounds right.
4. Click ⚡ Assign and pick the agent that should use it.

Pick a voice while building an agent Open the agent, go to Voice Engine, set the provider, then use the inline voice picker — every row has a play button, so you can audition voices without leaving the editor.

Give every voice good metadata When cloning, fill in gender, age, language, and use case. These power the filters, so a well-described voice is far easier for you (and your teammates) to find later.

---

Troubleshooting

---

Quick reference

</content> </invoke>