Voice consent for AI dubbing: what you need from guests and interviewees

Published 8 October 2026

Before an AI dubbing tool reproduces someone's voice in another language, that person should know about it and agree to it, and you should be able to prove they did.

This article is general information, not legal advice. Rules differ by country and by situation. If you are unsure, talk to a lawyer who knows data protection law where you and your speakers are.

Why is a voice treated as personal data?

A voice identifies a person. People recognise each other by voice, and software can do the same. That alone makes a voice recording personal data under the EU General Data Protection Regulation (GDPR) and Turkey's Personal Data Protection Law (KVKK).

AI dubbing goes a step further. To make a speaker sound like themselves in another language, the system analyses their voice and builds a model of its characteristics. That model is designed to capture what makes the voice unique. GDPR Article 9 and KVKK Article 6 place biometric data in a special category with stricter rules. Whether a particular voice profile falls into that category depends on how it is built and used, and authorities have not settled every case.

The practical answer is to treat voice cloning as if the stricter rules apply. Under both laws, explicit consent is the clearest basis for this kind of processing. It is also what your guests would reasonably expect: most people would want to be asked before a machine speaks in their voice.

Who needs to give consent?

Anyone whose voice will be reproduced. In a typical video that includes:

• You and your team. If you are the creator, your own consent is easy, but write it down for any staff presenters too.
• Guests and interviewees. Being on your show does not mean agreeing to have their voice cloned in other languages. Ask separately.
• Co-hosts and regular contributors. One consent can cover future episodes, as long as it says so clearly.
• Background speakers. People who are clearly audible, such as someone answering a question at an event or a passer-by in a street interview, also count if their words will be dubbed.

If a voice is only ambient noise and will not be reproduced, you usually do not need voice cloning consent for it. The question is whether the dub will recreate that person's voice.

What should a voice consent include?

Consent works when it is specific, informed and freely given. A vague line buried in a release form is weak. Use a short, separate statement that covers:

1. Who. The person's name and how they appear in the video.
2. Which videos. A specific episode, a series, or all future appearances on a named channel.
3. Which languages. List them, or say "any language" if that is the intent.
4. Which platforms. For example, your own channel, social media clips, or licensing to broadcasters.
5. How long. A fixed period, or until the person withdraws.
6. The right to withdraw. Say that they can withdraw at any time and what will happen if they do.
7. Contact. An email address or person for questions and withdrawal requests.

It also helps to say that the output is AI-generated and will be labelled as such, and that their voice profile will not be used for anything else. Keep a signed copy, or a recorded verbal consent, with the project files.

What about minors?

If a speaker is a child, consent should come from a parent or legal guardian. The age at which someone can consent alone differs between countries, so do not assume. Where it is reasonable, explain the dub to the child as well and respect their wishes.

For content featuring children, it is worth asking whether cloning is needed at all. A neutral voice or subtitles may be the better choice.

What happens if consent is withdrawn?

Withdrawal must be as easy as giving consent. When someone withdraws:

• Stop using their voice profile for any new dubbing straight away.
• Ask your dubbing provider to delete the profile and any reference audio, and get confirmation.
• Decide what to do with published videos: replace their part with a neutral voice, add subtitles, or take the dubbed version down.

Withdrawal does not usually make past use unlawful if consent was valid at the time. Continuing to use the clone after withdrawal is a different matter. Write down what you did and when.

What if a speaker has not consented?

You still have options that do not involve cloning their voice:

• Neutral voice. Dub their lines with a synthetic voice that does not imitate them. The audience still understands what was said.
• Subtitles. Keep their original audio and add translated subtitles for that segment.
• Cut. Remove the segment from the dubbed version if it is not essential.

Which one fits depends on the content. For an interview, a neutral voice often works well. For a short background comment, a cut may be simplest.

How does Dublayer handle consent?

At Dublayer, consent is confirmed for every speaker before work starts. If a speaker has not given permission, their voice is not cloned: we use a neutral voice or remove their part, as you prefer. You are the data controller and we are the data processor, with a data processing agreement as part of the contract. Voice profiles are used only for your videos, are never shared or sold, and are deleted when your plan ends. Our acceptable use policy and security page explain the details. To talk about your own videos, request a quote and we'll set up a meeting.