A voice is part of a person’s identity. Cloning technology should begin with clear permission and a defined purpose. This guide is practical information, not legal advice; local rules and platform requirements may also apply.

Get specific permission

Permission should explain whose voice will be cloned, who may use it, the intended project and how long the permission lasts. A casual recording sent for another purpose is not automatically permission to build a reusable synthetic voice.

For paid work, keep written records. If the project changes—from an audiobook to advertising, for example—confirm the new use before publishing.

Use a clean, voluntary recording

Tell the speaker that the recording will train or create a synthetic voice. Use a quiet sample with one speaker and no copyrighted background music. Do not secretly extract a voice from calls, videos or social media.

Collect only the amount of audio needed for the stated purpose. More data creates more responsibility for storage and deletion.

Set boundaries before generation

Agree on prohibited subjects and whether every script needs approval. Sensitive areas can include political messages, financial claims, medical advice, adult content or statements presented as personal testimony.

A consented clone should still not be used to mislead the audience about live participation or to bypass identity checks.

Store and delete responsibly

Protect the account with a strong sign-in method and restrict access to source recordings. Remove unused voice profiles and project files instead of keeping them indefinitely.

Provide a clear way for the voice owner to withdraw permission when your agreement allows it. Record deletion requests and confirm completion.

Disclose when context requires it

Disclosure is especially important when listeners could believe a real person personally recorded the message. Use plain language such as ‘AI-generated narration’ or ‘synthetic voice used with permission’ where appropriate.

Never use cloned voices for fraud, harassment, false evidence, deceptive political content or unauthorised celebrity impersonation.

COMMON QUESTIONS

Frequently asked questions

Is a public video permission to clone the speaker?

No. Public availability is not the same as consent to create and reuse a synthetic voice.

Is verbal permission enough?

Requirements vary, but written, specific permission provides a clearer record for everyone involved.

Important: Always review AI-generated audio before publishing. Use only voices and material you are authorised to use.