Dubbing an ad in the original speaker’s voice is one of the most convincing ways to take it to another market. The founder, the customer in the testimonial or the presenter keeps their own voice, now speaking German or Spanish.
That is also why it needs care. A cloned voice is a real person’s voice saying words they never recorded. This article covers the consent to get, the disclosure the EU AI Act now expects, and how to keep a record. It is general information, not legal advice; for a specific campaign, ask your lawyer.
What counts as voice cloning in an ad
In ad localization, voice cloning usually means building a synthetic voice from recordings of a real person and using it to speak a translated script. Three situations are common:
- Keeping the original speaker. The person in your ad is dubbed into other languages in their own voice.
- Cloning from a separate recording. A voice is built from a sample recorded for that purpose, then used across ads.
- Lip-sync on top. The speaker’s mouth is matched to the new voice, so the face and the voice both change.
Stock voices and voices designed from a text description are different: they do not imitate a real, identifiable person. The consent questions below are about the first group.
Consent: what to get in writing
A voice is part of a person’s identity, and recordings of it are personal data under the GDPR. Before you clone one, get written consent that covers:
- Who. The speaker’s name, and that they are the person in the recording.
- What. Which recordings may be used, and that the voice may be cloned and used to speak new scripts.
- Where. The ads, channels, languages and markets the cloned voice may appear in.
- How long. The period of use, and what happens to ads already running when it ends.
- Payment. Any fee or usage payment, especially for actors and creators.
- Withdrawal. How the speaker can withdraw consent, and what you will do when they do.
Contracts with actors, presenters and creators often already cover usage rights for the original recording. Most were not written with voice cloning in mind, so check that the agreement allows it, or add it.
The EU AI Act: when you have to disclose
Article 50 of the EU AI Act sets transparency obligations that apply from 2 August 2026. Two are relevant to ads:
- Providers of AI systems that generate synthetic audio, images or video must mark the output in a machine-readable way, where that is technically feasible.
- Deployers, the people and companies who publish the content, must disclose when image, audio or video content is a deep fake.
The Act defines a deep fake broadly: AI-generated or manipulated image, audio or video that resembles existing persons, objects, places or events and would falsely appear authentic. An identifiable person speaking, in their own voice, a language they never recorded will usually meet that definition. There is a lighter regime for evidently artistic, creative, satirical or fictional work, but a product ad is rarely that.
The disclosure has to be clear and distinguishable, at the latest when the viewer first sees or hears the content. In practice that means a visible label from the first second, for example:
Voice translated with AI.
Breaches of Article 50 can be fined up to 15 million euros or 3 percent of worldwide annual turnover, whichever is higher. The European Commission has been preparing guidance on Article 50. Guidance is not legally binding, but it shows how regulators read the rules, so check the latest version for your case.
Platform rules
Ad platforms have their own policies for AI-generated and digitally altered content, and they change often. Several require advertisers to label realistic AI-generated content or to disclose digital alteration in certain ad categories, such as political and social issue ads. Check the current policy of each platform you run on, because a label that satisfies the AI Act may still need a platform setting as well.
How Ad Localizer handles consent
Before Ad Localizer clones the voice of a speaker in your video, it asks you to confirm a statement:
The person speaking in this video agreed to their voice being cloned and used to voice translated versions of it, and I have the right to use their voice this way.
The statement you confirmed is recorded with its version, the time and your account, and kept as the record that consent was given. You keep the speaker’s own consent document. If you would rather not use a real person’s voice at all, the voice options include stock voices you can audition and voices designed from a description.
A checklist for advertisers
- Written consent from every speaker whose voice you clone, covering use, languages, markets and duration.
- A check that existing talent contracts allow voice cloning.
- A clear label on ads that use a cloned voice, visible from the start.
- A check of each ad platform’s AI disclosure policy.
- A record of consent you can find quickly, and a plan to remove the voice if consent is withdrawn.
If you are still deciding between a dub and subtitles, read dubbing or subtitles for video ads. Subtitles keep the original voice and need no cloning at all.