arccsere

in English: face swap

basicwidespread

Face swapping is a process that uses artificial intelligence to replace a person's face in an image or video with another person's face, producing a realistic result.

In plain terms

Imagine the photo booth at a fair, where there is a hole in the head of a painted figure, and you stick your face through it. Face swapping does the same digitally, only much more skillfully: the new face moves, speaks, and smiles along with the original body.

It is also used for good purposes. The film industry uses it to de-age actors or replace a stunt double's face with an actor's, while mobile apps offer it as an entertaining filter. It also appears in advertising if the participant has consented.

The problem begins when someone's face is placed on a recording they never appeared in, without their consent. This is how fake videos of politicians, fake intimate recordings, and fraudulent video calls are made.

What distinguishes it from traditional photo montage is that the face is not manually cut out and pasted in. The machine learns the facial features and redraws them on every frame from the correct angle with the appropriate lighting.

How it works

The program locates the face in the footage and marks key points: the eyes, nose, mouth, and face outline.
A neural network learns from images of the source face what makes the person recognizable; today's tools often need only a single photo for this.
The system reads head position, facial expressions, and lighting on every frame of the target footage, then draws the new face adjusted to these parameters.
It overlays the new face onto the original, smooths the edges, and adjusts skin tone and lighting to the environment.
In live operation, all this happens in a fraction of a second, so face-swapping works even during video calls.

An everyday example

On a social platform, a well-known doctor or TV personality recommends a "miracle cure" or investment in a video, even though they never said such a thing. Their face was superimposed onto another person's footage, and their voice was cloned separately.

Who is affected

everyoneAnyone's face can be used if a public photo exists, and anyone may encounter a fake video in their news feed.

parent, student, teacherA new form of school harassment involves replacing classmates' faces with humiliating or intimate footage. → Education & Family

bank customer, investorScammers use face-swapping to bypass video customer identification and use celebrities' faces to advertise fake investments. → Money

entrepreneur, executiveIt has happened that a fraudster impersonating a CEO in a live video call tricked employees into transferring money.

A common misconception

MythFace-swapping can always be spotted because the edge of the face blurs or the character blinks strangely.

In realityThese old errors have largely disappeared in better tools, and even weaker fakes can slip through on small phone screens. It is safer to verify the source than to examine individual pixels.

Limits and risks

Using someone's likeness without permission may infringe on their right to their image and data protection regulations, and in severe cases, could constitute a crime.
Recognition tools are useful but not infallible: they can flag real video as fake and let fakes pass.
If someone asks for money or data during a video call, call the person back on a number you know, because face and voice are no longer proof.
The fate of photos uploaded to free face-swapping apps is often unclear, so read first how the provider may use them.

Similar, but not the same

deepfake – Deepfake is the broader term that includes fake audio and entirely fabricated scenes; face swapping is just one type, albeit the most well-known.
ajakszinkron – In lip-syncing, the original face remains unchanged while only the mouth movements are adjusted to a new audio track.
facial recognition – Facial recognition determines whose face is in the image but does not alter it.

Related concepts

Broader

deepfake, synthetic media, generative AI

Narrower

live face swap, video face swap

On the edge

voice cloning, ajakszinkron, digital likeness, watermarking, deepfake detection

A brief history

Playful face-swapping filters appeared in mobile apps in the mid-2010s. The deep learning-based, realistic version became known in 2017, when a Reddit user named 'deepfakes' posted videos made with celebrities' faces; the word deepfake originates from this.

Where things stand

At the end of 2026, a single photo is already enough: a face swap takes a few seconds on a phone and also works in live video calls. Since August 2026, the EU Artificial Intelligence Act has required that anyone publishing such content must indicate that it is artificially created or manipulated. Nevertheless, the number of abuses is rising.

Status: 2026 November

Also known as

arccsere · face swapping · face-swapping app · face swap · faceswap · face swapping · deepfake arccsere · live face swap

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.