
InsightsNovember 20, 2026
AI Voice Cloning for Ads: Founder and Brand Voices
AI voice cloning for ads: how brands clone a founder, team or creator voice, what consent and labels TikTok, Meta and the EU require, and how to set it up.
AI voice cloning for ads means recording one real person, usually a founder, a team member or a creator, and turning that recording into a voice that can read any new script in your ads. It works when the person has agreed in writing and the ad is labeled where the platform requires it: TikTok, for example, counts a cloned voice saying something its owner never said as AI-generated content that must be disclosed.
Done right, a cloned brand voice is one of the most useful AI tools in ad production: the same familiar voice on every new hook, every cut and every market, without booking a studio. This guide covers how voice cloning works, whose voice to clone, the rules on TikTok, Meta and in the EU, and a setup you can copy.
What is AI voice cloning, and how does it work for ads?
A voice clone is a model trained on voice samples of one speaker that turns text into speech in that speaker's voice. Most voice cloning tools offer two levels. The documentation of ElevenLabs, the voice engine many ad teams already use, describes them like this (checked November 20, 2026):
| Instant voice clone | Professional voice clone | |
|---|---|---|
| Audio needed | At least 1 minute recommended | Bare minimum 30 minutes, ideally 2 to 3 hours |
| Whose voice | Anyone, after you confirm you have the right and consent | Only your own voice, for now |
| Check | A consent confirmation when you save the voice | A verification recording before training |
| Best for | Brand voices, creator voices, fast tests | A founder or presenter who records hours of material anyway |
For most ad teams the instant clone is enough to clone your voice or a creator's for ads. One to three minutes of clean, consistent audio, recorded the way the person normally speaks, gives a voice that can read hooks, product claims and calls to action. The professional voice cloning docs also note that more audio means a more accurate clone, but for a 15-second ad the difference rarely decides performance; the script does. For the best results, record in the same room, mic and tone each time.
Why are brands cloning their own voice instead of using stock AI voices?
Because listeners judge the voice, not the technology. In a UCL study published in PLOS ONE on August 2, 2023, 529 participants identified synthetic speech only 73% of the time, in English and Mandarin alike, and training them to spot it helped only slightly:
Our findings confirm that humans are unable to reliably detect deepfake speech, whether or not they have received training to help them spot artificial content.
If people cannot reliably tell human voices from cloned ones, sound quality is no longer the question; voice cloning technology has crossed that line. Whose voice it is becomes the question. A generic stock voice from a voice AI library sounds like every other ad in the feed. A voice the audience already knows from your organic videos, your founder's story or a creator they follow sounds like your brand.
That is what we hear from the brands we work with, more often every month. Teams clone the voice of a founder, a team member or a creator who agrees to it, and lay it over their UGC and product footage. They tell us customers recognize the voice, and that a creator's own voice on new edits sounds far more authentic than a voice picked from a library. Several also train one voice on several languages, so a winning ad sounds like the same person in every market. Many had started in ElevenLabs on their own before the voice became part of their editing workflow, which is why we built ElevenLabs voices into the editor.
Whose voice should you clone?
| Voice | Why it works | What to watch |
|---|---|---|
| Founder | Personal, credible, the face of many DTC brands | Scripts must sound like things the founder would really say |
| Team member | Builds a recognizable brand voice, easy to record | Agree what happens when the person leaves the company |
| Creator | Brings the trust content creators have built to new edits and markets | Paid usage terms, duration, and whether the creator may veto scripts |
| Voice actor | Professional delivery, clean rights | Less personal; you pay for a voice, not a relationship |
Two practical rules from our onboarding calls. First, pick the person whose voice already appears in your best organic content, so the clone extends something people know. Second, write the consent before you record: which brands, channels and languages the voice may be used for, for how long, and how the person can withdraw.

What rules apply to cloned voices in ads?
Four sets of rules matter for most brands advertising in the US and Europe.
TikTok. The TikTok advertising policy on misleading and false content (updated November 2026) counts "making the primary subject say something they didn't actually say, using AI voice-cloning" as significant AI modification. The same rule covers audio that is completely AI-generated, which is what a cloned voiceover is. Such ads need the AIGC label or your own clear disclaimer, caption, watermark or sticker; undisclosed AI content is rejected or restricted. TikTok's advice for borderline cases is to label. The same page bans using the voice or likeness of public figures without permission.
Meta. Meta's GenAI transparency update explains that ads created or significantly edited with its own generative AI tools carry an "AI info" label, and that Meta also detects ads made with third-party AI tools through industry-standard signals and labels them in "About this ad." Our Content Credentials checker shows which signals a file carries before you upload it.
European Union. Since August 2, 2026, the transparency rules in Article 50 of the AI Act apply: providers of AI systems must mark synthetic audio in a machine-readable way, and anyone who publishes a deep fake, content that resembles a real person and could pass as authentic, must disclose that it is artificial. Our Article 50 guide for advertisers explains what falls on you and what on your AI vendor.
United States. There is no federal voice law for ads yet, but states are moving. Tennessee's ELVIS Act, signed by Governor Bill Lee in March 2024 and in force since July 1, 2024, added vocal likeness to the personal rights the state protects and created a civil action against publishing someone's voice without permission, PBS NewsHour reported. For brands the practical rule is the same everywhere: no clone without written permission from the person whose voice it is.

How do you set up a brand voice for ads?
- Get written consent. Brands, channels, languages, duration, payment, and how the person can withdraw. Keep it with the voice file.
- Record clean audio. One to three minutes in a quiet room, in the tone the person uses on camera. Read a mix of short hooks and longer explanations.
- Test three scripts. A hook, a product explanation and a call to action. Listen for brand and product names; many tools let you set pronunciation rules once.
- Let the person approve. Send the first ads to the speaker before launch, especially for creators.
- Label and track. Turn on the platform's AI label where it applies, and note in your ad naming which versions use the cloned voice, so you can compare them with the original recordings.
Then test it like any other variable, whichever voice cloning tool you use: the same footage and script with the cloned brand voice against a stock voice, and against the original recording where you have one.
How to use a cloned brand voice with Solid
The voice is only useful if it is part of the edit. In Solid, ElevenLabs voices are built into the editor: you clone a brand voice from a short sample, set pronunciation rules for product and brand names once in the Brand Hub, and Solid Agent writes the script and generates the voiceover for the whole script over your real footage. Change a line and the voice changes with it, and translated versions use the same voice style in every market, which is how teams expand to new countries without recasting. The footage stays real, the voice stays yours, and every version still goes through review before export.
For the bigger picture on where AI helps and where it backfires, read what brands learned from AI commercials, and for scripts your voice can read, our UGC script template.
FAQ
Is voice cloning legal for ads?
Cloning a voice with the person's written consent and using it within the agreed scope is generally allowed, but the ad still has to follow platform and local rules. TikTok requires a label when a cloned voice makes someone say something they never said, and the EU AI Act requires disclosure of deep fakes. Using anyone's voice without permission is the line not to cross.
Can you clone a creator's voice for your ads?
Yes, if the creator agrees in writing and the deal covers paid usage, channels, languages and duration. Expect to pay for it, as you would for usage rights on their videos. Let the creator approve the first scripts, since the ad will sound like their endorsement.
How much audio do you need to clone a voice?
For an instant clone, ElevenLabs recommends at least one minute of clean audio. A professional clone needs 30 minutes at minimum and ideally two to three hours, and ElevenLabs currently allows it only for your own voice.
Does a cloned voice work in other languages?
Multilingual voice models can speak other languages in the cloned voice, and brands we work with use one voice across several markets. Have a native speaker check each language before launch, because accents and product names are where clones still slip.
Ready to give every ad your own voice?
Clone your brand voice once and let it read every new hook, cut and market.
Start free trial or Book a demo
About the author
Christoph Schütte is Co-Founder and CTO of Solid Labs, the company behind Solid, the AI creative engine for brands and their agencies. A physicist and former Engineering Lead at Google, he worked on video intelligence and computer vision in Google's self-driving cars and robots unit and with brands at Google Shopping on the loop between ad performance and production. Connect on LinkedIn
Sources
- ElevenLabs Documentation: Instant Voice Cloning (checked November 20, 2026)
- ElevenLabs Documentation: Professional Voice Cloning (checked November 20, 2026)
- UCL News: Humans unable to detect over a quarter of deepfake speech samples, August 2, 2023 (study in PLOS ONE by Mai, Bray, Davies and Griffin)
- TikTok Advertising Policies: Misleading and false content, last updated November 2026
- Meta: Expanding GenAI Transparency for Meta's Ads Products, February 2025
- European Union: Regulation (EU) 2024/1689 (AI Act), Article 50
- PBS NewsHour (AP): Tennessee just became the first state to protect musicians and other artists against AI, March 2024