Process servers: no generated targets
The person is private. Plates of documents you own, not faces.
The person is private. Plates of documents you own, not faces. A process-serving shop that generates the defendant, the witness, or “a typical recipient at the door” is not making a style of legal content. It is making a replica of a private figure. Calling it cinematic does not move a platform or a decent client.
A deepfake of a private person is not a style. That post is the rule. This page is the production split: the face is out. The paperwork you already hold is in.
A private face is not a production asset
Private means not a public figure. The person named on the papers. The person at the door. A staff server whose headshot you pulled off Slack “for consistency.” If a reasonable person who knows them would say “that is them,” you have a likeness problem, not a lighting problem.
TikTok’s AI-generated content help is explicit: young people under 18, and adult private figures used without their permission. Labelling the clip AI-generated does not create a grant. “It is an original character who happens to look like him” is not a defence if you used his face as the seed.
This is general information, not legal advice. The production version of the grant is a likeness release that covers generation. A standard model release authorises using footage you shot. It does not authorise making new footage of someone.
What is allowed, tightly: a person who does not exist, with no living referent; a person who signed a replica grant; your own face with consent written down. What is not allowed: “we changed 15% of the face,” “it is a vibe, not her,” a private headshot used as a “style reference.” Style reference of a private face is how you get their head. Minors are a hard stop.
Documents you own are the plate
The ad that does not need a face: a still of your affidavit, your proof of service form, your sealed envelope — documents you own, already in the file, no third-party identity on the plate. Crop names you do not have a right to show. Redact. Then, if the feed needs motion, animate the paper, not a person.
That still is frame one. The motion prompt is “papers settle, stamp lands, camera static.” It is not “a nervous man answers the door in Leeds.” The second prompt is a generated target. Do not.
If you need a human in frame, hire one, or film your own staff with a release that covers the use. Do not generate defendants. Background is not a likeness carve-out for a recognisable face.
Captions are speech, not a generated target
If the clip has spoken audio — a training take you actually recorded, a line you wrote and voiced yourself — burn captions on the file. Do not generate a mouth to carry the line.
Add captions to a video is the named job: speech in, styled subtitles out. Versely transcribes the spoken audio with VEED (add_veed_captions) and burns the result as positioned text. Twenty-one BASIC presets; nine DYNAMIC presets at 2×; 165 language codes. It transcribes. It does not invent a face.
If nobody said the line out loud, this is the wrong tool. A headline belongs on overlay, not on a transcription pass. Captions will not stand in for a generated defendant.
Invented extras still have a person in them
A labelled, obviously synthetic clip of an unnamed staff member or a photo you pulled off LinkedIn is already over the line on TikTok. The aesthetic is irrelevant. Separate “depicts a company” from “depicts a person” in the brief. The look of the work is paperwork and procedure. It is not someone else’s face.
FAQ
If I label the video AI-generated, can I use a private person’s face?
Not on TikTok, and not as a general rule anywhere else. The private-figure restriction is permission, not labelling. A label tells viewers the media is synthetic. It does not create a grant.
Can I generate a “generic” recipient so the face is not anyone real?
Only if there is no living referent and no reference photo of a real private figure. “Generic” that started as someone’s Instagram still is that someone. Use a document plate.
Will captions replace a talking actor?
No. Captions transcribe speech that is already on the file. They do not create a person. Record a line you have a right to use, or overlay text you wrote.
What if the person is an employee and we already have a model release?
Read it. If it authorises use of photographs from a shoot, it probably does not authorise generating new performances. Get a replica clause with a specific use, or shoot them.