There’s a reason professional headshots look better than selfies, and it’s not the camera.
It’s the light.
Professional photographers spend more time on lighting than on any other variable. Because light determines how skin renders. Harsh overhead lighting creates shadows under the eyes, accentuates pores, and makes texture look like topography. Soft, diffused light wraps around the face, minimises shadows, and makes skin look the way you see it in a mirror — which is to say, the way your brain expects it to look.
Most photos aren’t taken in professional lighting. They’re taken in offices with fluorescent tubes, in bars with a single overhead bulb, in bedrooms with a window on one side and a wall on the other. The camera captures what’s there. And what’s there is unforgiving.
This is the gap that retouching fills. Not vanity — correction. Bringing a photo closer to what the subject actually looks like under neutral conditions, rather than what they looked like under whatever accidental lighting was present.
Professional retouching in Photoshop is a skill that takes years to develop. Frequency separation, dodge and burn, colour grading — it’s a craft. And for commercial photography, it’s essential. But for the vast majority of skin-retouching needs — a LinkedIn headshot, a team page photo, a speaker bio image — the requirement isn’t “make me look like a magazine cover.” It’s “make the lighting look like it wasn’t terrible.”
AI-based skin correction handles this well. Modern models understand facial structure, can identify areas affected by harsh lighting, and apply corrections that smooth texture without eliminating it entirely. The result isn’t airbrushed. It’s corrected. The difference is important — airbrushing removes detail, correction redistributes it.
The privacy consideration is the same as it always is with face images. If you’re uploading a photo of your face to an online retouching tool, that image is being processed on someone else’s server. For a professional headshot, that might be acceptable. For anything more personal, it’s worth asking whether the tool needs your face on their infrastructure to do its job.
It doesn’t. The processing can happen locally.