digital likeness

in English: AI twin / digital double

basicspreading

An AI-generated copy of a real person that speaks with their face, voice and style in video or live conversation.

In plain terms

A digital double is like a very lifelike wax figure that can also speak. It learns from recordings of you how you look, how you speak and which words you use, and then it can say text that you never said on camera.

It is useful so you do not have to go to a studio every time. An instructor can create dozens of training materials from a single recording, a CEO can greet employees in multiple languages, and an actor can 'appear' in dangerous scenes.

A likeness is created with the consent and permission of the person concerned, and this distinguishes it from abuse. If someone's face or voice is copied without their consent, it is the same technology, but now it is deception.

How it works

Recording: video and audio samples are taken of the person. Today, a few minutes of material are enough; more is needed for better quality.
Training: an AI model learns facial movements, expressions, tone of voice, and speech rhythm. Some systems also learn the individual's writing style from their texts.
Text to video: you type what the likeness should say. The system generates the voice and adjusts lip movement and gestures.
Live version: if the likeness is linked to a language model or AI agent, it can also answer questions in real time, for example in a video call or customer service.
Labeling: responsible providers check the permit and mark the final material as AI content with a visible or hidden watermark.

An everyday example

You open the workplace e-learning material, and the company's training manager explains it; the following week, the same person speaks flawless German to foreign colleagues. Neither video was recorded: their digital twin spoke the typed text.

Who is affected

marketerYou can create videos in many languages and variations without filming, but obtaining the face owner's consent and adding AI labels is also your responsibility. → Marketing

entrepreneur, executiveUsing your own likeness saves time, but a fake version made of you can be used to defraud in your name, such as requesting transfers from colleagues.

lawyer, citizenYour face and voice are personal data, and your rights to your likeness and voice recordings protect them, so as a general rule, a likeness can only be created with your consent. → Law

everyoneYou increasingly see videos where the person themselves is not speaking, so a familiar face alone no longer proves authenticity.

A common misconception

MythA digital twin knows and thinks everything the original person does.

In realityA likeness only imitates the appearance, voice, and at most the style. It says what is written into it or what the language model running behind it generates, meaning it does not convey the person's actual opinion.

Limits and risks

The same technique allows anyone to be copied without permission, so always verify videos and voice calls requesting money or data via another channel.
If you have your own likeness created, read who stores your recordings, how they can be used, and how you can withdraw consent.
A likeness can say things in your name that you would not say, because the live, language-model-connected version can make mistakes and fabricate things.
The depiction of movement, gaze, and emotions can still appear stiff or unnatural, especially in longer videos and live conversations, though quality is improving rapidly.

Similar, but not the same

deepfake – The technology is the same; the difference lies in consent and intent: the term "deepfake" is commonly used for unauthorized, deceptive forgeries, while a "digital likeness" is created with the subject's consent. Under EU labeling requirements, however, even consented likenesses count as deepfakes and must be labeled.
digital twin – A digital twin is a data-driven virtual model of a machine, building, or process, not a copy of a person's face and voice.
Virtual influencer – A virtual influencer is a fictional character, while a digital twin is based on a real person.

Related concepts

Broader

generative AI, synthetic media

Narrower

voice cloning, ajakszinkron

On the edge

AI avatar, deepfake, digital twin, Virtual influencer, watermarking, speech synthesis

A brief history

The film industry has long used computer-generated "digital doubles" to replace actors. The term deepfake became widespread in 2017 for face-swapping created via machine learning. In 2023, a major point of contention during the Hollywood actors' strike was the conditions under which studios could use digital replicas of actors.

Where things stand

At the end of 2026, several online services let you create a talking likeness of yourself in minutes, and companies mainly use them for training, marketing, and customer service videos. The transparency rules of the EU Artificial Intelligence Act, applicable since August 2, 2026, require such AI-generated content to be labeled.

Status: 2026 November

In the news

Appears in 5 Agent Age stories; the curve shows weekly occurrence in the news magazine.

AI clones lead indie developers to withhold their games from public viewOct 19, 2026 · Code
US startup's AI avatar passes as human in job interview test, fooling 48% of participantsOct 12, 2026 · Work & HR
Tokyo court extends legal protection to voice against AI clonesOct 1, 2026 · Jog
Singapore forces TikTok and Meta to verify ads after deepfake scam surgeAug 21, 2026 · Marketing
Experts warn against using AI clone of murdered woman's voice in courtAug 10, 2026 · Jog

Also known as

digital likeness · AI clone · digital clone · digital copy · AI twin · digital double · digital replica · AI clone

Sources

Wikipedia: Deepfake

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.