FAQAvatarsRights & CompliancePolicy

Avatar Likeness and Consent FAQ: Actors, Clones, Deepfakes, Law

Whose face is in a Klip Kanvas ad — stock-actor consent, cloning yourself or a founder, prohibited uses, deepfake risk, revoking a clone, and why regional law still sits with you.

Updated 2026-03-1814 min read

Every face in the library is a licensed performer, not a scraped stranger. Custom clones need verified consent. This hub covers whose likeness you are using, what you may not clone, how to retire a face, and why we will not certify you against regional deepfake or advertising law.

01

Whose Likeness

1.Whose face is a stock avatar?

A paid, consenting actor who licensed their likeness for commercial advertising use in this product. They are not a random person pulled off the internet, not a composite of your customers, and not an employee of your brand. That licence is why you can run a generated talking-head as a paid ad without clearing each file with the actor yourself. It does not transfer the person to you. You hold a licence to use the generated video, not a property right in a human face, so you cannot present the avatar as a named individual, as your staff, or as a specific customer. If a retailer asks who is on screen, the accurate sentence is “a licensed AI avatar,” not a made-up biography.

#Whose Likeness

2.Are stock avatars real people or fully synthetic characters?

They start from real, consented performers and are built into persistent characters you can generate with. In a 9:16 feed, at arm’s length, they are meant to pass as filmed UGC. Frame by frame on a laptop, a trained eye still finds tells — hands, teeth, hair against a busy background. That realism limit matters for consent conversations too: the person agreed to a production method, not to you inventing a scandal in their likeness. Do not write scripts that would defame a real human just because the output is “AI.” The library now sits at 50+ faces and grows monthly. Casting still has to match the buyer, not your taste. Appearance gets you a shortlist; only a live test tells you who converts.

#Whose Likeness

3.Can I find out the actor’s real name?

No, and you should not try. The product relationship is with a licensed character in a workspace, not with a talent agent’s roster you can book for a dinner event. Do not reverse-search stills, do not DM people who look similar, and do not put a guessed name on a landing page. If a customer asks “who is that,” answer with the production method. Agencies that present the library as their own talent stable are misrepresenting the arrangement to the client and to the performer. Custom clones are the opposite case: you already know who it is because they sat on camera and read a consent script. Keep those two lanes separate in every brief.

#Whose Likeness
03

Cloning Yourself

8.How do I clone myself or a founder?

On higher-tier Klip Kanvas plans, record a short reference video — two to three minutes, well lit, phone at eye level — reading the consent script on camera, then upload it. A human consent check sits in the pipeline; budget one to two business days before the avatar is usable. Shoot in the framing you want. If the source performance is flat, every ad will be flat. Founder clones earn their keep where the person is the trust signal: coaching, agencies, formulator stories, local services. In commodity catalogues a stock avatar often performs the same, so do not gate a first campaign on a clone. Check /pricing for which tier includes cloning today. Time-to-first-ad on stock faces is still under 10 minutes.

#Cloning Yourself

9.Can I clone an employee as the brand face?

Only with a verified on-camera consent recording from that employee, plus an employment or contractor agreement that says what happens when they leave. A Slack “sure” is not enough. Do not clone a junior hire as a shortcut around the library; you are creating a likeness-and-HR problem that will outlive the campaign. If they are the trust signal, pay them like talent, bound the term, and plan a second face before you need one. If they are not, use the stock library. Employee clones are how brands accidentally build their entire ad account on a person who resigns. Concentration risk is operational as well as legal. We will not mediate that resignation.

#Cloning Yourself

10.Can several people in a company be cloned into one workspace?

Yes, each with their own consent recording and review. Keep them labelled, and do not mix a departed person’s face into new tests because the render is already in the folder. Agencies should clone a client founder inside that client’s workspace so access ends with the contract. Credits are driven by length and resolution, not by which face you pick, so testing a founder against a stock avatar costs the same as two stock faces. Use that: 3 hooks × 2 avatars, founder versus library, one body script. If the founder loses on hook rate, believe the number. Vanity casting is the most expensive clone you will ever approve.

#Cloning Yourself
04

Prohibited Uses

11.Can I clone a celebrity, influencer or customer?

No, unless that exact person records the verified consent script and you have a talent deal that matches the use. There is no route around the check for public figures, for a customer who emailed “you can use my photo,” or for a creator whose content you “just want to look like.” Lookalike filters in the 50+ library are the legitimate move; impersonation is the prohibited one. Uploads that try to bypass consent are rejected at review and should also be rejected by your own legal team. If you want a celebrity association, licence it through a normal talent route and film it. AI is not a cheaper version of a stolen endorsement. This is a hard no.

#Prohibited Uses

12.Can I make an avatar of a competitor’s founder or a private individual?

No. That is impersonation, and it is the fast way to a legal letter and an account restriction. Do not upload their keynote, their Instagram, or a still from a webinar. Do not write a parody that still uses their face. Competitive ads can attack an idea without wearing someone else’s skin. If your brief requires a specific private person, you are in talent-and-defamation territory the generator will not make safe. We will reject the clone attempt. If you generate a lookalike via a stock face and then name the competitor’s founder in captions, you have rebuilt the problem with extra steps. Stop at the brief.

#Prohibited Uses

13.What uses are prohibited even with a willing clone?

Illegal content, non-consensual sexual content, anything involving minors, scams, hate, and deceptive deepfakes — including political impersonation and fake news. A founder who consented to product ads did not automatically consent to a deepfake apology video, a fake endorsement of a medical claim, or a smear of a rival. Bound the use in the consent agreement: products, media, term, territory. Our terms already forbid misleading, fraudulent and infringing uses; your talent paper should be narrower still. The model will often technically speak a reckless line if you type it. That is a limitation of generation, not a green light. Human claims and ethics review still sit in front of publish.

#Prohibited Uses

14.What is the practical checklist before I use a face in paid media?

Stock avatar: intended advertising use, no fake biography, no prohibited vertical tricks, platform AI disclosure on. Clone: verified recording, written term, offboarding plan, deletion path, same disclosure. Then the creative bar: 15–60 seconds, 9:16 at 1080p, claims a human would stand behind, no uncleared trending music. Then the test bar: 3 hooks × 2 avatars, read at 48–72 hours or ~1,000 impressions. If any box is a shrug, do not render yet. Credits are cheap compared with a likeness complaint. The library exists so you can skip amateur face-scrape tools; use that advantage. Consent is the feature. Everything else is an ad.

#Prohibited Uses#Cloning Yourself
05

Deepfake Concerns

15.Is this a deepfake tool?

Klip Kanvas is a commercial UGC generator that uses licensed performers and consent-gated clones. That is a different job from a face-swap app built to put anyone’s head on anyone’s body. We still take deepfake risk seriously because the output is a realistic person. Guardrails: consent recording, no celebrity bypass, workspace-scoped clones, rejection of prohibited uploads. Guardrails are not a force field. A determined advertiser can still write a deceptive testimonial around a licensed face. Disclosure, honest framing, and your own refusal to impersonate are the rest of the defence. If your use case is political manipulation or humiliation, this is the wrong product and we do not want the work. Use a tool built for ads, for ads.

#Deepfake Concerns

16.How do I reduce the risk that viewers feel tricked?

Disclose when the platform asks, do not invent a biography, and do not lean on “real customer” chrome. Hybrid edits help: avatar for the hook, real product footage for the demo, so the proof layer is photographed. Keep talking-head shots short; past 8–10 seconds the tells show, and a viewer who spots the tell in second twelve will assume the claim is fake too. Target a 30%+ hook rate with a clean open rather than with intimacy theatre. If comments on YouTube start with “this is AI,” answer factually and fix the framing next batch. Trickery is both a ethics problem and a performance problem. The accounts that last treat synthetic as a production method, not as a con.

#Deepfake Concerns

17.Could a stock avatar be recognised by someone who knows the actor?

Yes, occasionally, especially as models get better. That is a known residual risk of a consented commercial likeness, not a bug you can patch with a filter. It is also why you do not write scripts that would humiliate a real person. If a campaign is sensitive — medical, political, scandal-adjacent — do not put a stock face in it. Use a founder clone whose contract matches, or film. Recognition risk is one reason we will not claim the output is indistinguishable from a camera. In a feed it often holds; in a friend group it might not. Plan for the more public of those two rooms. That honesty belongs in the client pitch as well as in this FAQ.

#Deepfake Concerns#Whose Likeness
06

Revoking a Clone

18.How do I revoke or delete a custom clone?

Request deletion so the private avatar is removed from the workspace and cannot be used in new generations. Then pause live ads that still use that face and replace them in the next 48–72 hour cycle. Deletion does not reach into Ad Manager, CDNs, or the MP4s you already downloaded — pulling those down is your job. Put revocation in the consent agreement: who can ask, how fast, what happens to in-flight campaigns. Founder and employee clones should be on the offboarding checklist next to email and laptop recovery. If you only remember to revoke during a dispute, you are already late. We will not automate away the consent check at create time, and we will not ignore a deletion request at end of life.

#Revoking a Clone

19.What if the person withdraws consent but the ads are winning?

You stop using the face in new generations and you wind down live ads on the timetable your agreement allows. Performance does not override consent. Swap in a stock avatar or a second clone on the same winning script — that recut is a 3–5 minute render and it is exactly why you should not have one human carrying the entire account. If the contract promised a wind-down window, use it; if it did not, treat withdrawal as immediate for new spend and get counsel on the residual files. This is a limitation of likeness-based creative, and it is the cost of using a real person at all. AI does not make a person easier to ignore. It makes them easier to over-use.

#Revoking a Clone

20.Can I keep running old ads after the clone is deleted?

Only if your consent agreement still covers those placements for that term. Deletion inside the product stops new renders; it is not a legal opinion about files on a DSP. Default operationally to pulling the face out of paid media when consent ends, even if last week’s ROAS was pretty. Keep a mapped list of ad IDs that use each clone so media can pause without a forensic hunt. If you cannot produce that list, you are already too concentrated. Stock-avatar ads you generated under the platform licence are a different stack and are not “revoked” because a founder clone was deleted. Do not mix the two in your head or in your folders.

#Revoking a Clone
07

Regional Law

21.Does GDPR or US state law change cloning?

Likeness, voice and a consent video are personal data and, in some places, a statutory publicity right as well. That means your legal basis, your retention, and your deletion process have to be real, not a checkbox in a slide. We process the recording to build the avatar; you still have obligations to the person, especially if they are an employee or an EU/UK resident. This article will not certify that a given clone is lawful in California, Illinois, the EU, India or anywhere else. Have counsel map it. The product contribution is the verified recording, the private workspace scope, and deletion on request. The rest is your employment paper and your regional advertising law. Hedge every sentence you put in a board deck.

#Regional Law

22.Do I have to disclose AI in every country I run the clone?

Many markets now expect a disclosure when a realistic synthetic person appears in an ad, and the exact label differs by platform and by region. Meta, TikTok and YouTube each have their own fields; some jurisdictions add advertising-law duties on top. We can recommend a disclosure; we cannot certify a global label. Default to disclosing, then ask local counsel where a market is strict. Skipping a label because “the founder really looks like that” is backwards — a clone of a real person is exactly when authenticity confusion is highest. Pair this with the Meta and TikTok policy FAQs. Regional law plus platform policy is a stack. Clearing one layer does not clear the other.

#Regional Law#Deepfake Concerns

23.Can I run a clone in political or issue ads?

Treat political, election and regulated issue advertising as out of scope unless your counsel and the relevant platforms have given you a current yes — and even then, impersonation and misleading synthetic media rules are at their sharpest. A founder speaking as themselves about their own company is different from a synthetic citizen giving a political testimonial. We are not a political-comms vendor, and we will not help you fabricate a voter. If your brand runs issue ads, film the humans who are accountable for the words. This is a limitation we are happy to keep. The cost of being wrong here is not a rejected ad; it is a scandal.

#Regional Law#Prohibited Uses

24.Will you testify or issue a legal opinion that my ad is cleared?

No. Klip Kanvas can describe our actor-licensing model, our consent check, and our policy intent around clones. We do not certify that a given file is lawful in a given market, that a disclosure is sufficient, or that a platform will approve it. Anyone selling “fully cleared AI talent worldwide” as a warranty is selling a future dispute. Your counsel owns the regional analysis; you own the script; the platforms own enforcement. Our limitation is also a boundary: we will not pretend a video generator is a law firm. If procurement needs artefacts, ask sales for the documents that actually exist. If they need a legal opinion, they need a lawyer.

#Regional Law

Ready to put this into practice?

Create your first AI UGC video ad in minutes — no filming, no actors, no editing.

Try Klip Kanvas free

More in this section

Ready to make ads like these?

Paste a product link and Klip Kanvas writes the script, casts the creator and renders the ad — no filming, no actors, no editing.