Voice Cloning Policy

Consent must be recorded, not asserted

Before any model is built, the person whose voice it is records a short spoken statement naming themselves, stating the date, and confirming that they consent to a voice model being made and used. It takes about a minute.

This applies to your own voice as well. The flow is identical, which keeps the provenance record consistent: every model in the system has a consent recording attached to it, and there are no exceptions that would let one slip through.

A written message, a screenshot, or a ticked checkbox is not a substitute. Those are assertions about consent. A recording is evidence of it.

What we will never clone

Public figures, including politicians, executives and celebrities. Performers, including voice actors and the actors behind specific characters. Neither is unlocked by a parody framing, a research framing, a disclaimer, or an enterprise agreement.

Anyone under eighteen. There is no consent process that unlocks this, including parental permission. A model of an identifiable child's voice is a tool for impersonating that child and no creative requirement justifies making one. The child character voices in our library are synthetic models built for the purpose, not recordings of children.

Deceased people. This is the request we receive most often, usually from someone who wants to hear a parent or grandparent read to their children. We understand it completely and the answer is still no: consent cannot be obtained after death, and in most jurisdictions an estate cannot grant it retroactively. What we can help with is casting a library voice in a similar register and using it as a character, which is a different thing from an impersonation and is the version that does not go wrong.

Deletion and withdrawal

The person whose voice was cloned can have the model deleted at any time by contacting us, whether or not they own the account that created it. We do not ask for a reason and we do not require the account holder's agreement. Withdrawn consent is consent withdrawn; any other policy would make the consent meaningless.

Deleting a model deletes the sample recording and the derived model. The consent recording is retained briefly as a record of the deletion request and is then removed.

If you are a voice performer

If you believe a model of your voice exists on this service without your consent, contact us and we will investigate and, where the claim holds up, delete it and close the account that made it. You do not need a lawyer to start this conversation.

If you want to license your voice deliberately — as a library voice with a revenue share — that is a programme we are building, and the partnership page is where to start.

Why we are strict about this

Not primarily because of legal exposure, although that is real. It is because a voice cloning tool with loose rules becomes, quite quickly, a tool whose main use is impersonation — and then the legitimate uses become impossible, because audiences stop believing any recording.

The rules above cost honest users about a minute each. They cost dishonest users the whole product. That is a trade we are happy to make.