Instant voice cloning is one of the most distinguished Speechify API features. Our system lets you create a voice cloned from a short sample of your own or another voice for which you have permission, and use this cloned voice for speech synthesis requests. We have no limit to the number of voices you can clone.

Voice Sample Recommendations

The ideal sample is several seconds of audio; normally, 10–30s is enough. However, keep it under one minute and below 5MB (if you’re using a pre-recorded sample file).

Try to avoid any background noise and speak clearly during the recording. The cloned voice will capture many aspects of the sample, including accent, speaking style, and tone of voice. Using a good microphone is recommended for optimal results. The speech itself should be in one of the supported languages.

If you need an idea, you can read this text when recording your sample:

Listening is like riding a storytelling rollercoaster, where you can lean back and enjoy the ride without having to steer.

The speaker's voice becomes your trusty guide, leading you through twists and turns. It's like having a personal audiobook adventure just for you!

So, buckle up, and let the fun begin!

API Support

To create a new cloned voice send the POST to https://api.sws.speechify.com/v1/voices: docs. Please note that alongside the params, such as the audio data and the voice name, you must also provide consent that the voice belongs to you, or to someone you represent.

Once created, the voice will appear in the voices list alongside the shared voices provided by Speechify. You can now use this voice's ID as the voice_id param to the speech generation endpoints (/v1/audio/speech and /v1/audio/stream).

Finally, you can delete the personal voice using the voices deletion API .

Speechify Console Support

You can clone the voice via the Speechify Console UI, and it will appear in the voices list, letting you test it right away. The console lets you both upload the sample from an audio file and recod the sample from your browser.