AI Voice Clone

Voice Cloning

Use a recording of your own voice or one you have permission to clone. Sign in and use credits to create the voice and generate speech in the cloud.

Upload a reference audio & clone the voice

Hear a voice read a new line

The reference is a synthetic voice made for this example. The second clip was generated by cloning that reference. Both clips are free to play.

Synthetic referenceWelcome to the riverside gallery. We opened the windows early this morning to let in the cool air. As you walk through the first room, listen for the quiet sound of water outside.
0:000:00
Cloned voiceTake a closer look at the blue glass bowl. It was made by hand.
0:000:00

Frequently Asked Questions

How does voice cloning work?

Creating a voice uploads your reference recording for cloud processing. When you generate speech, the service receives the new text and the saved voice identifier.

How long should the reference audio be?

A clear 10–20 second clip works best — long enough to capture natural rhythm and intonation, without needing more. It should be mono, at least 24kHz, and can be up to 60 seconds or 10 MB, but a short, clean sample beats a long, noisy one every time.

How is it priced?

Playing the examples is free. Creating a voice and generating speech with it each use account credits after sign-in.

What sample format and quality are needed?

WAV, MP3 or M4A files work, ideally mono and at least 24kHz, with 10–20 seconds of clear speech (up to 60 seconds and 10 MB). One person talking normally in a quiet room, with no music or background noise, gives the most natural clone.

Why does my clone sound off?

Almost always the sample is to blame rather than the tool. Background noise, room echo or clipping from recording too loud all carry straight into the clone. Re-record a cleaner clip in a small, soft space and the result usually improves dramatically.

Whose voice am I allowed to clone?

Only your own voice, or someone who has clearly agreed to it. Cloning a person's voice without their consent isn't permitted and can cause real problems — that's not what the tool is for.

How natural does the cloned voice sound?

With a clean 10–20 second sample the clone closely matches the original's tone and delivery. The real test is typing a sentence the sample never contained and listening — new words reveal the true quality, since any tool can echo back a phrase it was given.

Related Guides