Skip to main content
Open Voices (/dashboard/voices) to explore voices for your agents and, when enabled for your organization, create custom voices from a short audio sample.

​
Browse the voice library

Use search plus the Language and Gender filters to narrow the catalog. Each row shows the voice name, supported language information, a built-in audio sample, and any premium surcharge that applies to all or some languages. The agent builder filters this same catalog to voices compatible with the agent’s primary and additional languages. Selecting a voice in the library does not change an agent; open the agent’s Config tab to assign it.

​
Hear a voice say your text

Choose Hear voice say your words, enter 1–500 characters, select a supported language, and generate a sample.
  • Each new sample costs 3¢ from the organization balance.
  • The confirmation dialog shows the charge before generation.
  • An identical cached sample is served free and marked Served from cache.
  • Each organization can create up to 30 samples per hour.
  • A voice can be previewed only in languages available for that voice.
The Voice samples API exposes the same workflow programmatically.

​
Request custom voice access

Custom voices are enabled per organization. If access is not enabled, choose Request access, enter the number of voice slots you need, and optionally describe the use case in up to 2,000 characters. Only one request can be pending at a time. After access is enabled, the page shows the number of used and available custom-voice slots. Use Request more slots when you need a higher quota.

​
Create a custom voice

1

Confirm your rights

Confirm that you have the rights to use and clone the voice. A custom voice cannot be created without this confirmation.
2

Record or upload a clean sample

Record up to 15 seconds in the browser, or upload a WAV or MP3 file up to 4 MB. Use one speaker, a steady pace, and as little background noise as possible.
3

Name and classify it

Enter a name of 1–100 characters, choose its language, and optionally mark the voice as female or male. Select Create voice.
Creation is asynchronous. A voice moves through Creating, Ready, or Failed. Ready voices can be played and assigned to agents; a failed voice can be retried with a new sample.
The custom-voice language picker contains the languages supported for custom voice creation. It may be narrower than the complete voice-agent language catalog.

​
Rename, retry, and delete

  • Rename changes the label shown throughout the dashboard without changing the voice id.
  • Try again starts a new creation attempt after a failure.
  • Delete removes the voice from your organization’s picker and frees its quota slot. Agents already using it keep their saved configuration until you re-save them with another voice.
For the complete object, status, quota-request, preview, rename, and deletion contracts, see the Custom voices API.

​
Pricing

Static premium voice/language paths are marked +3¢/min in the library and builder. The effective speech surcharge is based on the actual voice-language combination; language and premium-voice labels do not stack into two separate 3¢ charges. Organization-specific custom-voice pricing, when applicable, is returned by the custom voice policy and reflected in the builder’s all-in rate.