Skip to main content

Cloned voices and pronunciation

Organization settings → Voice holds the organization's cloned voices and its pronunciation dictionary. Owners, admins and editors can open it.

Cloned voices​

A cloned voice reads narration in your own voice, or the voice of a colleague who agrees to it. Cloning needs ElevenLabs narration: the platform administrator sets it up, or on Self Hosted you add your own ElevenLabs key under Settings → AI text, translation & voiceovers. Cloned voices speak only when your voice model there is an ElevenLabs model.

To clone a voice:

  1. Open Voice in Organization settings and select Clone a voice.
  2. Consent: enter the Speaker's full name, confirm that they agree, and select Next. Clone only your own voice or a voice whose owner has agreed; the organization keeps the name and the date of consent.
  3. Name and language: enter the Voice name shown in the editor, choose the Language of the recording, then select Next.
  4. Recording: select Record and read aloud the text shown, about thirty seconds in a quiet room, or Upload a recording of the same speaker in that language (MP3, WAV, M4A, OGG or WebM, up to 5 MB). Trim the silence before using the take. Changing the language afterwards discards the recording.
  5. Select Clone voice.
  6. In the editor, open Settings → Voice settings, choose the voice under Your organization's voices and generate a voiceover to hear it.

The recording is sent to ElevenLabs, which creates the voice in the account whose key is used: the platform's shared account, or on Self Hosted the personal account of the member who cloned it. Each clone uses part of your AI allowance; a recording that ElevenLabs refuses costs nothing.

An organization keeps up to 10 cloned voices. Its creator, owners and admins can delete one; narration already generated with it keeps playing, but new narration cannot use it.

Pronunciation dictionary​

The dictionary tells every voice, cloned or not, how to say product names, abbreviations and jargon, for example SQL → sequel. Owners and admins edit it; editors can read it.

  1. Under Pronunciation dictionary, select Add a term.
  2. Write the term under Written and how to say it under Say it as. Choose a Language to limit the entry to narration in that language.
  3. Select Save dictionary.
  4. Remove a voiceover that uses the term, generate it again and listen to it.
  • A term matches whole words regardless of case, so Acme also changes "ACME" but not "Acmes". A longer term wins over the words inside it.
  • An entry for one Language wins over the same term for All languages.
  • The dictionary changes only what the voice hears. The text on screen and the subtitles keep the term as written.
  • Narration generated before a change keeps its audio. Remove the voiceover and generate it again to use the new pronunciation.

The dictionary holds up to 200 terms.