Create or clone a voice

On Standard plans and up you can make your own voices and use them in any Text to Voice speaker. Below Standard, the voice picker says Custom voices require a Standard plan.

Open it

  • Click + on the left toolbar, then Generate Voice > Create or Clone Voice.
  • Or, on an audio node, choose Audio > Clone Voice.
The New voice dialog: Upload, Select from canvas, Record and Generate from text, with the 10 second minimum below.
The New voice dialog: Upload, Select from canvas, Record and Generate from text, with the 10 second minimum below.

Clone a voice

  1. Add speech in one of three ways: Upload (audio or video files up to 10MB), Select from canvas (click audio nodes on your canvas), or Record with your microphone.
  2. Add at least 10 seconds of clear speech. 30 to 60 seconds gives better results.
  3. Tick the box confirming you have the rights to clone this voice, then click Next.
  4. Preview the voice, give it a Name and an optional Description, and click Save voice.

Describe a voice

Choose Generate from text, click Describe, and write what the voice should sound like. Pick one of the previews, name it and save it.

Use your voice

In the voice picker, open My voices. Voices you made are under Created. The tab shows how many voice slots you have used: Standard has 3, Pro 10 and Studio 25. Remove a voice to free a slot. A line whose voice was removed will not run until you choose another voice.

My voices in the voice picker: Created voices marked Cloned, Saved library voices with a heart, and the voice slots used at the top right.
My voices in the voice picker: Created voices marked Cloned, Saved library voices with a heart, and the voice slots used at the top right.

Save library voices

Click the heart on any voice in Explore to keep it under My voices > Saved. Saving works on every plan.

If something looks wrong

Next stays greyed out. The sample is shorter than 10 seconds of speech, or the rights box is not ticked.

The clone does not sound like the speaker. Use a longer, cleaner sample: one speaker, no music, little room echo.

Next