Quick Answer
Ethical AI video comes down to four duties you can actually check. Consent: clone your own face and voice, or hold written permission, because Tennessee's ELVIS Act turns on use "with knowledge that use of the voice or likeness was not authorized". Disclosure: the EU AI Act has applied since 2 August 2026 and requires deepfake content to be disclosed, while TikTok, YouTube and Meta each require their own label for realistic content. Provenance: keep the photo, the voice sample and the consent together. Refusal: do not make a real person appear to say something they did not say.
Four checkable duties for AI video: consent for a face or voice that is not yours, disclosure under the EU AI Act and platform rules, provenance records, and what never to make. Checked 16 September 2026.
"Ethical" is usually where a page stops being specific. This one lists four duties instead, each with a rule behind it and a date, read on 16 September 2026. It is information rather than legal advice.
1. Consent, and what it has to cover
The straightforward case is your own face and your own voice. On Percify, Clone Yourself makes both for 5 credits each, once, from one photo and 15 to 30 seconds of speech.
When the face or voice belongs to someone else, get permission in writing that names AI cloning and the uses you intend, and keep it. This is not lawyerly caution: Tennessee's ELVIS Act, in force since 1 July 2024, defines "voice" to include a "simulation of the individual's voice" and prohibits use "with knowledge that use of the voice or likeness was not authorized". The full legal picture, including why copyright is the wrong claim here, is in voice cloning and the law.
2. Disclosure, which is now a legal duty as well as a platform rule
- The EU: Article 50 of the AI Act ↗ requires deployers of deepfake content, meaning generated or manipulated image, audio or video resembling real people that would falsely appear authentic, to disclose that it was artificially generated or manipulated. The Act applies from 2 August 2026, so this is current law rather than a proposal.
- TikTok requires a label ↗ on realistic AI content including AI generated speech, and a label it applies itself cannot be removed.
- YouTube requires disclosure ↗ of AI used to meaningfully alter or generate photorealistic content, through the 'AI use' setting, and states that creators do not need to disclose non realistic content made with AI.
- Meta has required disclosure ↗ of photorealistic video or realistic sounding audio since 6 February 2024 and shows viewers an "AI info" label.
The per platform table is in each platform's label rule.
3. Provenance, which costs nothing until you need it
Keep the source photo, the voice sample, the consent and the date together, per avatar. If a video is ever questioned, the difference between an awkward week and a simple answer is whether you can show where the face and the voice came from. Treat it as ordinary bookkeeping rather than a legal exercise.
4. Refusal, which is the part no tool enforces for you
Do not make a real person appear to say something they did not say. That single line covers the public figure deepfake, the fake endorsement and the impersonated colleague, and it is the fact pattern behind both the ELVIS Act and the pending federal NO FAKES Act of 2026, which was reported to the Senate on 24 June 2026 and is not law. Separately, the FTC proposed a rule in February 2024 ↗ prohibiting impersonation of individuals, and asked whether platforms that know their tools are used for impersonation should be liable too.
What Percify does about this
Every plan, the free one included, carries commercial rights and adds no watermark, so nothing here depends on paying more. Read on 15 September 2026: the avatar is 5 credits for the face and 5 for the voice, once, and talking video is 2 credits a second, about $2.53 a minute on the Creator plan.
What it does not have, which matters for this topic: no live avatars, and only a small library of ready-made presenters. Its API refuses to clone a voice unless you attest that you have the rights to it, but it cannot verify that attestation. The consent, the disclosure and the refusal are yours. See plans and packs.
This page is information, not legal advice, and the law is moving: the federal bill is pending and state statutes continue to be written.
Ready to Create Your Own AI Avatar?
Make a talking video of yourself from one photo, in your own voice. No watermark on any plan.
Get startedCancel anytime, no contracts
Got questions?
Frequently asked
Four checkable duties rather than a general attitude: consent for any face or voice that is not yours, disclosure of realistic AI content, provenance records kept with each avatar, and refusing to make a real person appear to say something they did not say.
In the EU, yes for deepfake content: AI Act Article 50 requires deployers to disclose that such content was artificially generated or manipulated, and the Act applies from 2 August 2026. Platform rules apply separately: TikTok, YouTube and Meta each require a label or disclosure for realistic AI content. Read on 16 September 2026.
Permission is the minimum, not the whole answer. Get it in writing, naming AI cloning and the intended uses, keep it with the voice sample, and disclose the result where the platforms require it. Tennessee's ELVIS Act turns on whether the use was authorized, and it treats a simulation of a voice as the voice.
No. Every plan, including the free one, carries commercial rights and adds no watermark. Read on 15 September 2026.
A real person appearing to say or do something they did not. That covers fake endorsements, public figure deepfakes and impersonated colleagues, and it is the conduct both the ELVIS Act and the pending NO FAKES Act of 2026 are aimed at.
