AI is now part of most video pipelines in some form, whether it is cleaning up audio, generating a background plate or cloning a voice for a pickup line. The tools are easy to use. The questions about who owns the result, who agreed to it and what the audience deserves to know are not. This article sets out the working rules we apply on brand projects so that marketing teams can make informed decisions before the edit starts.
Consent for voice and likeness
The clearest line is a person’s face and voice. Cloning a presenter’s voice to fix a mispronounced product name is a small, useful task, but it still requires that presenter’s written permission, with a description of what the clone may be used for and for how long. The same applies to digital doubles, de-aging, and any avatar built from a real employee or spokesperson.
Talent agreements written before generative tools became common often do not cover synthetic reuse. If you are working with a brand ambassador, a physician for a medical device video or a musician for a label, assume the contract needs an explicit clause rather than hoping the general usage terms stretch to cover it.
Ownership of generated material
Who owns an AI-generated image, music bed or voice track depends on the tool’s terms and on the rules in the jurisdiction where you operate, and both change frequently. The practical approach is to treat generated elements as licensed assets rather than owned ones. Keep a record of which tool produced what, under which account and plan, and confirm that the terms allow commercial use in advertising.
For anything that becomes a core brand asset, such as a logo animation, a mascot or a signature sound, it is worth having a human designer or composer create the final version. Generated material is useful for exploration, but a brand element you cannot clearly own is a liability.
Training data and third-party rights
Generative tools can reproduce recognizable styles, faces or products from their training data. A prompt that asks for a car ad in the style of a well-known campaign can produce something that looks uncomfortably close to another brand’s work. The same goes for celebrity lookalikes and distinctive architecture or artwork.
Our rule is simple: if a generated frame looks like it belongs to someone else, it does not go in the film. Reference other work for mood, but build the final images from your own footage, licensed stock or original design.
Disclosure to the audience
Advertising regulators in many markets are moving toward requiring disclosure when synthetic people or synthetic voices appear in commercial content, and social platforms increasingly ask for AI content to be labeled at upload. Requirements differ between markets, so check local advertising rules and platform policies for each country you publish in, including the UAE and Saudi Arabia, rather than assuming one standard applies across the region.
Beyond compliance, disclosure is a trust question. Audiences generally accept AI-assisted editing, color, cleanup and translation without comment. They react badly when they discover that a testimonial, a doctor or a customer was synthetic and nobody said so. A short caption or end card costs nothing and removes the risk.
Brand safety and factual accuracy
Generative tools invent details. A rendered pharmacy shelf may show products that do not exist, an AI translation may soften a safety warning, and an avatar will read a wrong dosage as confidently as a right one. For regulated categories such as healthcare, finance and food, every AI-touched element needs the same review as the original script.
A short checklist we use on brand projects:
- Written consent on file for every cloned voice and every likeness, with scope and duration.
- A log of which tools generated which assets and under what license.
- No generated element that imitates a named person, brand or artwork.
- A disclosure plan agreed with the client before the final export.
- A human review of all AI-generated text, translation and product imagery for accuracy.
Closing thoughts
None of this should slow you down. Most of it takes an extra hour at the briefing stage and saves weeks if a problem surfaces after launch. The brands that will use AI well are the ones that decide their rules early and apply them consistently, rather than the ones that use it most.
If you would like a second opinion on how AI fits into your next brand film, or a production partner who already has these checks built into the workflow, Gizmo Media Production is happy to talk it through. Get in touch and tell us about your next project.
Comments are closed