Using AI to generate speech that sounds like a specific real person from a short audio sample. For agencies, it changes voiceover production and raises real consent and licensing questions.
Also known as AI voice replication, synthetic voice
Voice cloning is the process of training or prompting an AI model to reproduce the sound, pacing and tone of a particular person’s voice. Modern tools need only seconds to minutes of clean audio to produce a usable copy that can read any script.
The output is a synthetic voice that can say lines the real person never recorded. That makes it useful for production and risky without clear permission, because the voice is tied to a real individual’s identity and livelihood.
Voiceover is one of the most repeated, schedule-sensitive jobs in agency production, so cloning lands there first.
Revisions get cheap. A last-minute script change no longer means rebooking talent and a studio, if the talent has agreed to a licensed clone.
Localization scales. The same voice can be rendered across languages, which helps multi-market campaigns keep one recognizable sound.
Consent and contracts are the exposure. Using a voice without a written, specific license can breach talent agreements, union rules and publicity rights, and the legal risk sits with whoever published the spot.
A regional retail client wants twelve radio cutdowns with weekly price changes. The agency has a signed license from the original voice actor covering AI-generated reads for this client, with a fee per use and an approval step for each script. Producers generate the weekly versions in an hour instead of rebooking the studio, and every file is logged with the license reference. The same team turned down a request to imitate a famous actor’s voice for a scratch track, because no license existed and a scratch track tends to become the final.
The automations and agents module of the workshop teaches you how to build AI workflows that compress the busywork without taking the craft out of the studio.