Voice Cloning from an Audio File

Already have a recording? Here is how to clone a voice from an existing audio file, which formats work, and how to prepare the clip for the best result.

Clone your voice free

You don't have to record live to clone a voice. If you already have a clean recording, such as a podcast intro, a voice memo or a narration take, you can upload it in the Studio's Upload file tab.

Supported formats

Any audio format your browser can decode works: WAV, MP3, M4A/AAC, OGG and usually WebM and FLAC. The file is converted automatically to the format the model expects: mono, 24 kHz.

How to prepare the clip

  1. Pick one speaker. The clip must contain only the voice you want to clone. Interviews and conversations need to be cut first.
  2. Choose 10–20 seconds of continuous speech. Avoid sections with laughter, music or long pauses.
  3. Remove background music. The model copies everything it hears. A vocal-isolation tool can help with podcast or video audio.
  4. Avoid heavy effects. Reverb, auto-tune and phone-line filters will be copied into every sentence.

The Studio trims silence at the start and end and normalises the volume. Clips longer than 30 seconds are cut to the first 30 seconds, so put your best material first.

Uploading someone else's recording does not give you the right to clone their voice. Only clone your own voice or a voice whose owner has explicitly agreed. See our privacy and consent page.

Frequently asked questions

What is the maximum file length?

There is no file-size limit for the upload itself, but only the first 30 seconds of speech are used. 10–20 seconds is ideal.

Why does my clone sound echoey?

The source recording probably contains room echo or reverb. Use a drier recording, made closer to the microphone in a softer room.

Ad space · inline