Kurdish Voice Cloning

Kurdish voice cloning turns a short recording of a voice into a custom Kurdish text-to-speech voice. Upload or record 6–10 seconds of one person speaking Sorani or Kurmanji, give the voice a name, and Kurdish TTS enrols it on the same AI speech engine that powers the studio. Once enrolled, the cloned voice can read any Kurdish text you type, and on paid plans it sits in your voice list next to the built-in voices, both in the web app and through the API.

What is Kurdish voice cloning?

Classic voice cloning trains a new model for every speaker. This does not. The engine uses your clip as an in-context reference: it listens to how you sound and conditions every sentence it generates on that reference, so there is no training run to wait for. The clip is also transcribed to anchor the voice, which is why you choose the dialect when you upload: Sorani or Kurmanji decides which Kurdish speech recogniser writes that transcript. Right after enrolment you hear the new voice read a short preview, and the engine reports whether your clip qualified for its highest-fidelity path or fell back to a weaker one, so you know straight away whether to re-record.

How to clone a Kurdish voice

1. Sign in. Cloning needs an account; the free plan lets you try one clone in preview mode.
2. Upload a WAV, MP3, FLAC or OGG file of up to 50 MB, or record directly in the browser.
3. Choose the dialect the clip is spoken in, Sorani or Kurmanji, and name the voice.
4. Create the clone and listen to the preview. If the engine flags the clip, record again. Otherwise open Text to Speech, where the voice is listed among your own voices, or press “Use in TTS” on its card.

Recording a good reference clip

Longer is not better. The engine wants about 6–10 seconds of continuous speech, and pauses do not count, so a twelve-second clip with long gaps can still fall short. Clips must be at least 3 seconds and no more than 30. Use one speaker only, with no music or other voices behind them, in a quiet room with little echo, and say one complete sentence at a natural pace; the in-browser recorder finishes at the next pause after you press stop. Every clip is checked before it is sent: too short, too long or too little speech is rejected with a specific message, while a noisy recording gets a warning rather than a refusal.

Who can clone voices

Every signed-in account can clone. The free plan includes one cloned voice in preview mode: you can create it and hear it read a fixed sentence, but not use it on your own text. Paid plans unlock the voice in Text to Speech and through the API, with an allowance per plan: Creator 3 cloned voices, Developer 10, Pro 20 and Business 30. The page shows a running count, and when the allowance is used up you can delete a voice to make room or upgrade; Business customers who need more can contact us. Voices can be renamed or deleted at any time, and older subscription plans keep an equivalent allowance.

Cloned voices work everywhere the built-in voices do: in Kurdish text to speech, where they appear in your voice list, and in the TTS & STT API for developers. Compare the plan allowances on the plans page, see API pricing, or find answers to common questions in the FAQ.