Ethics, Consent and AI Media

Likeness and Voice in Digital Media

Why a person's face and voice deserve special care in digital media, particularly when edited, cloned or generated, and how creators can respect them.

On this page
  1. Why a face and a voice are different
  2. What “likeness” includes
  3. Ordinary media versus altered media
  4. Consent for faces and voices
  5. Voices deserve their own thought
  6. The dead, the famous and the satirized
  7. Practical habits for creators
  8. When you are the one pictured

Why a face and a voice are different

Most material in a project can be treated as a resource: a stock background, a piece of music, a font. A face or a voice is not like that. It is tied to a particular person, it signals who they are, and other people use it to recognize, trust and judge them.

That is why personal likeness deserves care. Altering or reproducing someone’s face or voice can affect their reputation, safety, livelihood and sense of control over their own identity, even when no harm is intended.

What “likeness” includes

The term usually covers more than a photograph. It can include:

  • A face, body or distinctive look
  • A voice, including accent, tone and recognizable manner of speaking
  • A name, signature or catchphrase
  • A performance style that people strongly associate with a person
  • A convincing imitation, whether by an impressionist, a lookalike or software

Legal protection for these things varies considerably by country and region, and for public figures, private individuals and commercial use. This article covers general principles only. For anything commercial or contested, official guidance and qualified advice are the right sources.

Ordinary media versus altered media

A direct recording of a person, used in the context they expected, is the simplest case. Trouble grows as the distance increases between what the person did and what the audience sees.

Light edits. Cropping, color correction, trimming silence and audio editing to remove noise are usually ordinary. Even so, trims can change meaning, so keep the sense of what was said.

Heavy retouching. Reshaping a body or face, or smoothing features to an extreme, can mislead about the person and, when applied to a depiction of someone else, can feel disrespectful. Ask whether the subject would agree.

Compositing and re-contextualizing. Placing a person in a setting they were never in, or captioning a photo to suggest something untrue, can mislead without altering a single pixel.

Synthetic reproduction. Cloning a voice, generating a face or building a digital double goes further still, because the person never performed the new material. See the broader concerns around synthetic media.

Consent to be photographed or recorded is not the same as consent to be simulated. The two are separate permissions, and a person who is happy to appear in one video may be unhappy to become the template for a talking avatar.

Good practice includes:

  • Asking explicitly about any generated or heavily altered version, and describing it in plain terms
  • Limiting the scope. Specifying what the simulation may say or do, where it will appear and for how long
  • Writing it down. A clear written agreement protects both sides
  • Allowing withdrawal where practical, and handling it kindly
  • Paying attention to power imbalances. A performer, employee or student may feel pressured to agree

Details such as payment, ownership and duration are best handled through a proper agreement, with professional advice if the project is commercial. The principles in getting permission to feature people apply, but with higher stakes.

Voices deserve their own thought

A voice can feel less personal than a face, but it carries identity, emotion and trust. Cloned voices have been used in impersonation scams where the listener believes a relative or colleague is speaking. Even in benign projects, a synthetic voice that sounds like a specific, real person can imply that they said or endorsed things they did not.

If you want a narrator and a generated voice suits the project, choose one that does not imitate a real individual, and tell listeners what it is. If you work with a human voice performer, set expectations early about how recordings may be reused, particularly for training or generating other material. Skilled microphone technique and clean home recording make it easy to get usable audio without needing to imitate anyone.

The dead, the famous and the satirized

Questions arise about people who cannot consent or who are widely known.

  • Deceased people. Families and communities may still be affected by a simulated version of a loved one. Approach with humility and ask those close to the person, if they can be identified.
  • Public figures. Fame does not mean unlimited use. Commentary and satire have recognized places in many cultures, but a realistic fake that could be taken as real is a different matter. Clear labeling helps, but it is not a guarantee against harm.
  • Historical figures. Illustration is common, but presenting invented words or images as authentic is misleading.

Practical habits for creators

  1. Use real recordings for real claims. Do not use synthetic material as evidence of what a person did or said.
  2. Get specific permission for any altered or generated version of an identifiable person.
  3. Label clearly. The AI disclosure principles apply directly.
  4. Keep records. Note what was agreed, and store releases with project files and file versioning notes.
  5. Avoid sensitive contexts. Do not place real people in sexual, violent, political or commercial scenes they did not agree to.
  6. Listen to complaints. If someone objects to how they are shown, take it seriously and be ready to change or remove the material.

When you are the one pictured

If your own face or voice has been used in a way you did not agree to, save evidence, use the reporting tools of the platform involved and consult official consumer, privacy or legal resources in your region. Outcomes vary, and removal is not always fast, but documentation helps.

The underlying idea is straightforward: a person’s face and voice belong to them first. Treating them that way costs little and prevents some of the most serious harms that digital media can cause.

The complete guide to ethics, consent and ai media

Ethics in Content Creation: A Plain-Language Guide
Read the guide

Keep exploring

Guides from other topics

All topics →