A few years into freelancing, I had a client send back a batch of beauty headshots with a note that stopped me cold. “The eyes look like marbles,” she wrote. She wasn’t wrong. I had sharpened the irises, brightened the whites, and painted in a catch light I thought looked natural. What I’d actually done was strip out every subtle tonal variation that makes eyes read as human. They were technically clean and completely lifeless.

That batch sent me back to the drawing board, and what I figured out over the next several months changed how I approach every portrait that comes through my queue.

Why Eyes Are the Hardest Part of Any Portrait to Retouch

The eye is doing a lot of optical work at once. The sclera (the white) is almost never pure white. It picks up ambient color from the scene, and it has a subtle spherical shadow where the eyelid curves over it. The iris has radial texture, but it also has depth. Light enters the pupil and bounces around in ways that create a dark limbal ring at the outer edge and often a subtle lightening closer to the pupil. The catch light is a reflection of an actual light source, so its shape and position need to make physical sense.

When we retouch without accounting for these layers, we flatten the eye. We make it look like a sticker of an eye rather than an actual eye. The information was there, and we edited it out.

The 4-Layer Approach, From Base to Finish

I do all of this non-destructively on a duplicate of my merged working file. Every adjustment lives on its own layer, named and grouped so I can dial each one back independently.

Layer 1: Sclera correction. I use a Hue/Saturation adjustment layer clipped to a rough selection of the whites, feathered by about 4-6 pixels. I desaturate the reds and yellows slightly, usually around -15 to -25 depending on the original shot. Then I pull luminosity down just a touch, maybe -5 to -8. This keeps the whites clean without making them glow.

Layer 2: Iris depth. I create a new layer set to Multiply at around 20-30% opacity and paint into the outer ring of the iris with a soft brush loaded with a color sampled from the darkest part of the iris itself. This rebuilds the limbal ring if it got blown out in post-processing or was just naturally faint. Separately, I use a Curves adjustment clipped to the iris to add a gentle S-curve. I lift the midtones just slightly (input 128, output 138) and bring up the highlights a touch. This is where the eye starts to look like it has dimension again.

Layer 3: Catch light refinement. If there’s already a catch light in the shot, I almost never replace it. I might clean its edges with the Healing Brush, or use a tiny dodge to brighten its core. If the catch light is missing or poorly positioned, I paint one in on a new layer set to Screen, using a very small soft white brush (usually 8-12 pixels) at low opacity, building it up slowly. It should sit at roughly the 10 or 2 o’clock position on the iris, never dead center, and it should be slightly warm or slightly cool depending on the light source in the image.

Layer 4: Sharpening only what matters. I run a High Pass sharpen on a merged stamp (Ctrl+Alt+Shift+E), set to Overlay at 60-70% opacity, and I mask it entirely in black. Then I paint the sharpening back in with white only over the iris and pupil. I keep it completely off the whites and the skin around the eye. Sharpening the whole eye area is one of the fastest ways to get that plastic, over-processed look.

The Whites Trap That Gets Everyone

The single most common mistake I see in retouching students’ work is overcorrecting the sclera. I understand the impulse. It’s one of the most visible areas of the face in a close-up, and making it brighter seems like a quick win. But the eye sits inside a socket. Light wraps around it. The corner of the eye nearest the nose is almost always darker than the center, and the area closest to the upper eyelid carries a shadow.

If you paint the entire white area to the same brightness value, you lose that curvature entirely. A gradient painted at Luminosity blend mode, running from slightly darker at the edges to slightly lighter at the center, often does more for a portrait than any amount of brightening.

When to Leave the Eye Alone

Early in my career, before I figured out that restraint was a skill and not a failure of effort, I retouched everything I could see. I thought a well-edited image was a heavily edited one. Now I’ll sometimes spend twenty minutes on an eye and then pull my layers group down to 60% opacity overall because the original shot was doing most of the work already.

A sharp, well-lit eye from a good photographer needs cleanup, not reconstruction. Know the difference before you open your curves. If the catch light is compelling and the iris has tonal range, your job is to protect what’s there, not add to it.

One Final Check Before You Export

I keep a gray neutral (R128, G128, B128) swatch in the corner of my canvas while I work. Before I flatten anything, I zoom out to 50% and look at both eyes together, then glance at the gray swatch. It recalibrates my perception. After an hour of close work, I’ve usually lost objectivity about brightness levels, and eyes that read as natural up close can still look slightly lit from within at full view.

The goal isn’t technically perfect eyes. It’s eyes that make whoever sees the portrait feel like someone looked back at them.