Can You Make an AI Video of a Real Person? Consent, Labels, and What We Refuse to Make
You can make an AI video of a real person, but only with that person's written consent, given for the specific project. We do not make AI content showing public figures, politicians, celebrities or anyone who has not agreed, and we label what we deliver as AI-made. This post explains what our published AI content policy requires before we use a real face or voice, how we trace and label every piece, what we refuse to make, and what to do if something goes wrong. The full policy is on our site; this is the plain-language version for anyone commissioning, appearing in, or reporting an AI film.
Key takeaways
- A real person's face, body or voice is used only with their written consent for the specific project; a voice is cloned only at their written request.
- We do not make content showing public figures or anyone who has not consented, and a child's likeness needs a parent's written consent.
- Fictional characters come from prompts or our own fictional images, never from a photo of a real person.
- Every piece is traceable: prompts, model and version, logs, inputs and approvals are kept for at least three years.
- Deliveries are labelled as AI-made, provenance marks are never stripped, and reports get a 24-hour response and a three-working-day investigation.
The short answer
A real person's face, body or voice goes into an AI film only with that person's written consent, and the consent is for the specific project. For client work it sits inside the signed agreement or a separate signed consent. There are no exceptions for public figures; we simply do not make content showing public figures, politicians or celebrities, or anyone who has not consented.
Consent is not a formality for us. Generating a believable film of someone is a power that can be misused, and the only protection that works is a process that makes misuse hard and traceable. That process is the rest of this post.
What counts as consent
Consent is written, it is for this project, and it covers what is actually being made. A person's face or body is used only with that person's written consent; their voice is cloned only at their written request, never otherwise. If a child's likeness is involved, we need a parent's written consent. A real person's likeness is uploaded only to platforms that verify identity and do not train on customer data, and never to a library of virtual or fictional characters.
That last point is practical as well as ethical. A real face placed in a pool of fictional avatars can end up used for things its owner never agreed to. Keeping it to verifying platforms keeps the trail short and the use bounded.
Fictional characters are different, and held to their own rules
When a film needs a presenter who is not a real person, we create the character from text prompts or from our own earlier fictional images, never from a photo of a real person. The prompt states that the person is an original fictional character not resembling any celebrity, and we reject any output that looks like a known person. Third-party logos and brand marks are removed before use. This is how a presenter in a promo can be entirely invented, as in our work for an online tuition service, where the presenter was cast from a generated still and never existed.
The evidence we keep
For every character and every delivered video, we keep records for at least three years after delivery: the generation scripts and prompts, the model name and version and the date; the usage logs, including how many input images were used; the input images and every intermediate version; and the client's written approvals and consent documents. They live in the project's records and are backed up.
The reason is accountability in both directions. If someone says a film shows them without permission, we can show what was made, from what, when, and with whose consent. If a client is asked how a film was produced, they can be answered from evidence rather than memory.
How we label what we deliver
Every preview leaves with a visible AI-generated or company watermark. We keep the platform's own watermark or provenance metadata wherever the platform adds it, and we never strip a provenance mark in order to pass content off as real footage. Final deliveries carry a credit saying the content was made with AI, unless the client's contract says otherwise and the content cannot mislead anyone about a real event or a real person.
That last condition is the one that matters. A cartoon song for children carries little risk of misleading anyone; a realistic film of a person doing something they did not do carries a great deal, and the exception is not available for it.
What we will not make
There is a short list of work we decline regardless of who is asking or what they offer.
- Content that deceives people about real events.
- Content that impersonates a real person or organisation.
- Sexual content.
- Content targeting a private individual.
- Political persuasion using synthetic people.
- Anything unlawful in India, in the client's country or in the platform's country.
If something goes wrong
Anyone can report misuse by writing to admin@nexinfinitymeta.ai. We acknowledge within 24 hours. Within 24 hours of a report we pull the content from our channels, ask the client and platforms to take it down, and suspend the related assets. Within three working days we check the evidence, establish what happened and tell the affected person. We notify the AI platform concerned and, where the law requires it, the authorities, then fix the process, update the policy and log the incident with dates and actions.
| When | What we do |
|---|---|
| Within 24 hours | Acknowledge the report |
| Within 24 hours | Pull the content, request takedowns, suspend the related assets |
| Within 3 working days | Review the evidence, establish what happened, inform the affected person |
| After that | Notify the platform and, where required, the authorities; fix the process and log the incident |
What happens after a report
What we need from you
If your film involves real people, the checklist is short, and having it ready is the quickest way to start.
| Item | Why |
|---|---|
| Written consent from each person whose face, body or voice appears | No real person's likeness is used without it |
| A parent's written consent for any child who appears | A child's likeness is never used without it |
| A written request for any voice clone | We clone a voice only at that person's request |
| The purpose and where the film will be shown | Consent is for the specific project |
| Your approval of each version we send | Kept as part of the project record |
Before we start a film that includes real people
This is our process, not legal advice
Rules on synthetic media differ between countries and platforms and change over time. Our policy is reviewed every six months and whenever a platform or the law changes, but you should take your own legal advice for any film that could affect a person's rights.
Read the policy, then talk to us
The full text is at our AI-generated content and deepfake prevention policy, version 1.0, effective 15 October 2026. For how a film like this is made and priced, see how much an AI video costs and keeping a real face real in photo-to-video, and for the work itself, our AI video creation service. To start, send the brief, and tell us at the outset if real people are involved.
Generating a believable film of someone is a power that can be misused. The only protection that works is a process that makes misuse hard and traceable.
Frequently asked questions
Can I make an AI video of a real person?
Yes, but only with that person's written consent, given for the specific project. We do not make AI content showing public figures, politicians, celebrities, or anyone who has not consented, and we never use a child's likeness without a parent's written consent.
Can AI clone someone's voice for a video?
Only at that person's written request. We never clone a person's voice without it.
Do AI videos need to be labelled as AI-generated?
Our policy is that previews carry a visible AI-generated or company watermark, we keep platform watermarks and provenance marks, and final deliveries carry a credit saying the content was made with AI, unless the contract says otherwise and the content cannot mislead anyone about a real event or person.
What AI content do you refuse to make?
Content that deceives people about real events, impersonates a real person or organisation, is sexual, targets a private individual, or is political persuasion using synthetic people, and anything unlawful in India, the client's country or the platform's country.
What records do you keep for AI videos?
For at least three years after delivery we keep the scripts and prompts, the model name and version and date, usage logs, input images and intermediate versions, and the client's written approvals and consent documents.
How do I report misuse of an AI video?
Write to admin@nexinfinitymeta.ai. We acknowledge within 24 hours, pull the content and request takedowns within 24 hours, and investigate within three working days, informing the affected person.
Is your AI content policy a legal certification?
No. It is our own published process, version 1.0 effective 15 October 2026. We do not claim any certification or standard, and it is not legal advice.
Have a project in mind?
We design, build, and ship software end-to-end — with a fixed, written quote after a free scoping call.