The first time a client told me my retouching looked “plastic,” I was devastated. I’d spent three hours on that image, smoothing every pore, evening out every shadow. It looked clean to me. But she was right. The woman in the photo didn’t look like a person anymore. She looked like a rendering of a person, the kind of face you’d see on a video game character or a department store mannequin. I kept that image on my desktop for months as a reminder that more retouching is almost never the answer.

What I’d done was the most common mistake beginners make: I’d blurred the texture right along with the tone. Those two things, texture and tone, are completely different problems. And once I understood that, everything changed.

What Your Skin Is Actually Made Of (In Photoshop Terms)

When we look at a skin imperfection, we’re usually seeing two separate issues layered on top of each other. The first is a tonal problem: an area is too dark, too red, or unevenly lit. The second is a textural problem: a raised bump, an open pore, a scar with physical depth. Most retouchers smear both of these together with a healing brush and wonder why the result looks wrong.

Frequency separation solves this by literally splitting the image into two layers. A low-frequency layer carries the color and tonal information, things like blush, shadows, and uneven skin tone. A high-frequency layer carries the fine detail: pores, hair, texture, the tiny variations that make skin look like skin. When we work on these layers independently, we can smooth a ruddy patch of color on the low-frequency layer without touching the pore texture sitting above it. The skin still reads as real because that surface detail stays intact.

Setting Up Frequency Separation Without the Guesswork

The blur radius you choose determines how much information lives on each layer, and it’s the setting most tutorials gloss over. For beauty work at high resolution (typically 300 DPI, around 20 to 30 megapixels), I start with a Gaussian blur of 6 to 8 pixels on the low-frequency layer. For lower-resolution files or environmental portraits, I drop that to 3 to 4 pixels. Too low and you haven’t separated enough; too high and you start blending texture into the tonal layer, which defeats the whole point.

Here’s the setup, step by step:

  1. Duplicate your background layer twice. Name the bottom copy “Low Freq” and the top copy “High Freq.”
  2. Hide the High Freq layer. Select Low Freq and apply Gaussian Blur at 6 pixels (adjust for your file size).
  3. Make High Freq visible and select it. Go to Image > Apply Image. Set the layer to Low Freq, blending to Subtract, Scale 2, Offset 128. Click OK.
  4. Change the blending mode of High Freq to Linear Light.

That’s it. You now have a working frequency separation setup. I have a Photoshop action called “Eternal Sunshine” that runs all four steps in about two seconds, but doing it manually at least twice helps you understand what each layer is actually doing.

Working the Low-Frequency Layer Without Destroying It

This is where most people go wrong even after they’ve set up the layers correctly. On the Low Freq layer, we’re using the Lasso tool with a feather of 15 to 25 pixels (again, scale to your file) to select uneven tonal areas, then using Curves or the Clone Stamp at 10 to 15% opacity to blend them out. The key word is blend. We are not eliminating. A face with zero tonal variation looks as fake as a blurred one.

For redness, I’ll often add a Hue/Saturation adjustment clipped to the Low Freq layer, drop into the Reds channel, and desaturate just that problem area by 15 to 25 points. For dark circles or bruising, a Curves layer with a warm midtone lift does more in ten seconds than ten minutes of healing brush work.

On the High Freq layer, the Clone Stamp at 100% opacity with a very small, hard brush is your main tool. You’re copying texture from a nearby area of similar skin and stamping it over a blemish. Because there’s no color information on this layer, you can’t accidentally create color mismatches. It’s surgical in a way that healing brush work on a flat image never is.

The Year I Spent Getting This Wrong Before I Got It Right

I want to be honest about something. After I first saw a demonstration of frequency separation, I was convinced I understood it and spent the next several months applying it badly. My separations were inconsistent because I wasn’t accounting for file resolution, which meant my blur radius was too low on large files and the layers weren’t actually separated properly. I was doing the work on the wrong layer half the time. And I was over-retouching the High Freq layer, cloning texture until it looked repetitive and stamped, like a skin-texture wallpaper tile.

What fixed it wasn’t another tutorial. It was slowing down on a single image and asking, at each step, whether I was solving a tonal problem or a textural problem. That question alone, applied before every brush stroke, cut my revision rounds in half. Clients stopped commenting on the retouching, which, in this work, is the highest compliment.

The One Habit That Will Protect Your Work Every Time

Before I deliver any retouched file, I zoom to 50% and toggle my retouch layers on and off three times. Not at 100%, where every pore is a crisis, and not at 25%, where everything looks fine regardless. At 50%, which is close to how the image will actually be seen in print or on screen, the difference between real-looking skin and plastic skin becomes obvious immediately. If I can’t tell which version is retouched, I’m done. If I can, I’ve gone too far.

The goal was never a perfect face. It was always a believable one.