Voice cloning without the creepy part

A cloned voice is a powerful thing. It can read a podcast in five languages or give a founder back a voice lost to illness. It can also be misused, so we built the guardrails first and the feature second.

Consent is recorded, not assumed

Before a voice is cloned, its owner reads a short, randomly generated sentence out loud. We match that recording against the sample. No match, no clone.

Every clone has an owner

Cloned voices live inside one workspace. They cannot be shared, exported, or used through another account, and the owner can delete them at any time.

Audio is marked

Everything we generate carries an inaudible watermark. If a clip turns up somewhere it should not, we can tell whether it came from us and which workspace made it.

What we refuse

• Voices of public figures without written permission
• Samples scraped from video or social media
• Anything meant to impersonate a real person to someone else

We would rather lose a customer than become the reason someone’s grandmother wires money to a stranger. That trade feels easy.