Most creators want a digital twin. Most creators also want to make sure they do not become the cautionary tale.
That is a legitimate tension. AI avatar tools have built reputational debt for the entire creator-twin category by cutting corners on consent, ownership, and disclosure.
This post is where ENCORE draws the line. Not the marketing version. The actual policies that ship with the product, and the reasoning behind each one.
The four lines we will not cross
Line one: consent is mandatory and verified.
You cannot create a twin of anyone who has not personally enrolled. The enrollment requires a live recording from the account owner's device. The voice in the enrollment must match the face in the enrollment within a similarity threshold. Attempts to enroll using prerecorded video are rejected.
This rules out: twins of public figures, twins of celebrities, twins of friends or colleagues without their explicit enrollment, twins of deceased persons.
We rejected a $2.4M pre-seed offer in 2025 from a fund that wanted us to add "celebrity twin licensing" to the roadmap. The legal cover was thin. The trust cost was infinite. Pass.
Line two: the twin is yours.
The training data lives in your account. We encrypt at rest. We do not share it with third parties. We do not use it to train any general-purpose model. If you cancel your subscription, you can request full deletion of your twin and underlying training data within 7 days.
Your twin is also exportable in a sense: you can request a portability package (the trained model weights are proprietary to the vendors, but the metadata, enrollment footage, and render history are yours).
This rules out: "data licensing" business models where your twin trains models for someone else's benefit, even with anonymization.
Line three: platform compliance is enforced at render time.
Every Twin Mode render carries an AI-disclosure watermark and a metadata flag. TikTok, Meta, and YouTube require this in 2026. LinkedIn and Pinterest do not require it yet. We enforce it everywhere.
The watermark is small (8 pixels in the corner of the lower third). It does not affect retention or engagement based on our private beta data. It does affect platform trust scores, which means your reach stays stable even when the platform later adds stricter AI policies.
This rules out: "stealth mode" twin generation, removing disclosure for advertising purposes, hiding the watermark with creative cropping.
Line four: no political, medical, or financial misinformation.
Twin Mode cannot be used to generate content where you, as a non-credentialed source, make claims that look authoritative on politics, medical advice, or financial recommendations.
Specifically, our content classifier blocks twin renders that contain: specific medical claims (cancer cures, drug interactions, treatment protocols), specific financial advice (buy/sell signals, portfolio recommendations, tax advice), and political content with named candidates within 60 days of a US election.
Educational content on these topics is allowed. Personal opinions are allowed. Specific authoritative claims are not.
This rules out: pharma deepfakes, election interference content, get-rich-quick schemes using your face.
The Twin Mode primer covers what the twin can and cannot do.
Why creators sign off without hesitation
The four lines above are not consumer-friendly on paper. They restrict what you can do.
Every creator I have walked through them on a demo call has thanked me for them. Not one has pushed back.
Here is why.
The downside of getting this wrong is permanent. A creator whose twin gets used in a deepfake scandal loses the audience forever. Even if the creator was not at fault. The audience cannot un-see it.
Platform trust is the longest-running asset. TikTok and Meta are getting strict on AI content. Creators who use compliant tools get reach. Creators who use stealth tools get suppressed. Six months of better reach is worth more than any short-term advantage from skipping disclosure.
The legal landscape is shifting fast. State-level deepfake laws are passing weekly. Federal frameworks are coming. Creators who use compliant tools today will not have to redo their entire content library when the laws hit.
The bigger argument for why your twin is a switching cost (positively, in your favor) is here.
The consent flow, explained
Concretely, here is what happens during enrollment.
Step 1: identity confirmation. You verify the email and phone on your account. Two-factor required.
Step 2: live capture. You record the 3-minute enrollment script on a connected device camera. The system rejects pre-recorded uploads. Time-of-capture metadata is logged.
Step 3: voice-face match. The enrollment voice is compared against the enrollment face's lip movements. They must match within a similarity threshold or the enrollment is rejected.
Step 4: explicit consent confirmation. Two-tap confirmation. "I confirm I am the person in this recording. I authorize ENCORE to use this data to train my digital twin. I confirm the twin will carry AI-disclosure metadata on all renders."
Step 5: storage and encryption. The raw enrollment footage is encrypted at rest. It is accessible only to the training pipeline (which runs in isolated infrastructure) and to you (via your account).
Step 6: training and review. 4 to 6 hours of training. You preview before the twin goes live. You can reject and re-enroll if the likeness misses.
The whole process is auditable. We log every step. If you ever need to prove consent (for example, in a legal situation involving content scraped and misused), we can produce the chain.
What we do when something goes wrong
Three failure modes we have planned for.
Account compromise. If someone gains access to your account, they could theoretically render twin videos as you. We mitigate with mandatory 2FA for all Twin Mode renders, IP-anomaly detection on render requests, and a 24-hour cooldown for new render destinations.
Subpoena or law enforcement request. We respond to lawful requests under US law. We notify the user unless we are legally prohibited. We have not received any such request as of June 2026.
Content used in malicious ways. If you discover a twin video being used in a way that violates our terms (deepfake elsewhere, identity theft, etc.), we have a fast-track takedown request flow with the major platforms via established channels.
The bigger picture
Digital twins are going to be a default creator tool by 2028. The question is not whether they exist. The question is whether the category gets built on consent-first infrastructure or stealth-first.
We are betting on consent-first. The market signals so far (5,200 waitlist signups, 63% pricing preference, 4.7/5 concept test score) suggest the market agrees.
If you want a twin that you own, that the platforms trust, that the law respects, ENCORE is the option that does not skip any of those steps.