Beta
Clone an Arabic voice without losing its regional character
Keep the voice private
A clone belongs to the account that created it. Publishing a voice is a separate, explicit decision rather than the default.
Use one voice across Arabic text
Select your saved voice in the studio, or pass its voice ID to the speech API to generate new recordings.
Make consent part of the workflow
The cloning flow requires confirmation that the speaker owns the voice or gave permission for it to be cloned.
A short path from sample to speech
Create a voice in the dashboard by uploading a recording and confirming that you own the voice or have the speaker’s permission. Once encoding finishes, use the resulting voice ID with the speech endpoint.
- 01
Record a clean sample
Use 10–30 seconds of one speaker with little background noise and no music.
- 02
Confirm permission and create
Name the voice, confirm the speaker's consent, and let the platform encode the sample.
- 03
Generate with the new voice
Select the clone in the dashboard or pass its voice id to POST /v1/audio/speech.
Questions developers ask
How much audio is needed to clone an Arabic voice?
Use a clean sample of about 10–30 seconds with one speaker. Background noise, music, or multiple speakers reduce the usefulness of the sample.
Can I clone someone else's voice?
Only with their explicit permission. The person creating the clone must own the voice or have consent from the speaker.
Can a cloned voice be used through the API?
Yes. After creation, pass the clone's voice id to the same OpenAI-compatible speech endpoint used by the built-in voices.