Synthetic faces: Inside the technology that generates fictional people

How computers construct human portraits from scratch

Every human face is a complex blend of symmetry and tiny imperfections: the curve of an eyebrow, fine lines around the eyes, skin pores, and soft shadows beneath the chin. For decades, digital artists spent weeks painting realistic faces by hand for video games and special effects. Today, artificial intelligence programs create convincing human portraits in a fraction of a second.

Unlike a camera taking a picture, the software does not capture light bouncing off a living person. It also does not simply paste existing eyes, noses, and mouths together like a collage. Instead, it creates an entirely new portrait by calculating what a typical human face looks like based on patterns learned from thousands of real photographs.

The analogy of the art student and the strict teacher

To understand how these images are born, imagine an art studio with a talented student and a demanding art teacher. The student's goal is to sketch human faces so convincingly that the teacher cannot distinguish them from real photographs. The teacher's job is to catch every flaw.

On the first day, the student produces messy, blurry scribbles. The teacher immediately rejects them: "Eyes are never shaped like that, and real skin has texture." The student returns to the easel, adjusting proportions, blending skin tones, and refining stray strands of hair. After thousands of critiques, the student becomes so skilled that the teacher can no longer tell whether a drawing was sketched in the studio or captured with a camera.

In computer science, this setup is known as a generative network. One part of the program generates new visual patterns from random numbers, while the other evaluates them against real portraits. When the two reach balance, the software can output an infinite gallery of realistic faces of people who never existed.

Where synthetic faces appear in everyday life

Artificial portraits have spread rapidly across commercial media and the internet, bringing practical conveniences alongside deceptive uses:

• Commercial stock photography: Companies use synthetic portraits in brochures and website mockups, saving the cost of hiring models and avoiding privacy or licensing disputes.
• Video game and virtual world characters: Game developers populate expansive digital cities with unique background citizens without designing every face individually.
• Anonymous digital identity: Online privacy advocates use artificial profile pictures on public forums to shield their personal identity from facial recognition scanners.
• Deceptive social media profiles: Fraud networks deploy synthetic faces on automated accounts to appear trustworthy when running marketing scams or political influence campaigns.

What this means for you and how to spot them

While synthetic faces look stunning at first glance, computers still struggle with the messy irregularities of the physical world. A few simple visual checks reveal the artificial origin:

• Inspect jewelry and eyeglasses: Software frequently creates mismatched earrings, glasses frames that melt strangely into ears, or lenses with inconsistent reflections.
• Look closely at the background: Real photos usually feature recognizable rooms or streets. Synthetic images often blur the background into surreal, swirling shapes or melted textures.
• Examine hair and collar boundaries: Watch for stray hairs that dissolve into thin air or collar edges that blur softly into the neck.

As synthetic imagery becomes an everyday part of our visual culture, keeping our curiosity active helps us appreciate the craft while staying grounded in reality.