Skip to main content
Use a custom voice when you have permission to clone the speaker’s recording. For provider selection and general voice configuration, start with Voice Studio.

​
Prepare the sample

Choose a clear recording of the intended speaker. The clone dialog accepts MP3 (audio/mpeg), WAV (audio/wav), MP4/M4A audio (audio/mp4 or audio/x-m4a), OGG (audio/ogg), and FLAC (audio/flac). The size limit is 10 MiB (10,485,760 bytes), shown as 10 MB in the dialog. The file must be recognized as a supported audio type; renaming an unsupported file extension does not convert it.

​
Create the voice

  1. Open the assistant’s voice selector and go to My Voices.
  2. Open Clone Your Voice.
  3. Enter a recognizable voice name.
  4. Upload the audio file.
  5. Confirm the rights and consent checkbox after reviewing it.
  6. Click Clone Voice.
  7. Wait for the upload and cloning operation to finish.
The current cloning path creates an ElevenLabs custom voice. Provider catalog selection and sample cloning are separate capabilities; selecting another provider does not make this upload dialog clone with that provider.

​
Apply and test

Find the new voice in your available voices and preview it. Apply it to the assistant, publish, and run a voice test using the names, numbers, and language your customers will use. Check pronunciation and consistency rather than relying only on the short preview.

​
Copying a custom voice

Copying by ID requires a shareable custom voice with copying allowed and source audio available. The requested ID is the Assistable voice record’s ID, not an arbitrary provider voice ID. A new clone is owned by the selected account and starts with copying disabled. A voice that disallows copying cannot be duplicated through this route.

​
Troubleshooting

  • Upload rejected: check the actual audio format and size.
  • Cloning fails: retry with a valid sample after checking whether the voice was created.
  • Voice is absent in another account: verify the account and the voice’s sharing/copy permissions.
  • Preview works but calls use another voice: check the selected assistant and publish the updated voice configuration.