A few years into freelancing, a client sent me feedback that I still think about. She said the model’s skin in my delivery looked “like a wax figure.” She wasn’t wrong. I had been smoothing everything, healing over every pore, every shadow variation, every tiny imperfection, without understanding that those things were not flaws. They were texture. They were what made the face look like a face. I spent the next several months learning frequency separation properly, and it completely changed how I work.

What Frequency Separation Actually Does to Your Image

Most people hear “frequency separation” and think it just means working on skin without ruining texture. That’s the result, but it’s worth understanding the mechanism, because once you understand it, your decisions at every step become intentional rather than guesswork.

Every image is carrying two kinds of information at once: color and tone (think of this as the broad washes of light and shadow across the face), and texture (the fine detail sitting on top, the pores, fine lines, hair follicles, surface variation). Frequency separation splits those two things onto different layers. Low frequency holds the soft tonal information. High frequency holds the sharp detail. When they’re separated, you can even out a blotchy patch of color without touching the pores sitting on top of it. That’s the whole trick. You’re not removing texture. You’re correcting the layer underneath while leaving the texture layer untouched.

Setting Up the Layers in Photoshop (With Exact Settings)

Here’s how I build my setup on every portrait file. I work in Photoshop, and I’ve saved this as an action I call “The Godfather.” Once you run it a few times, it takes about 20 seconds to execute manually if you understand each step.

Start by duplicating your background layer twice. Name the bottom copy “Low Freq” and the top copy “High Freq.” Now, on the Low Freq layer, go to Filter > Blur > Gaussian Blur. For a file shot at 24 megapixels retouched at 100%, I use a radius between 4 and 8 pixels. The goal is to blur just enough that you can no longer read fine skin texture. If you can still see pores, go higher. If the image is a smaller web-ready file, you might drop to 2 or 3 pixels. The right setting depends on your file size and shooting distance, so test it.

Now select the High Freq layer. Go to Image > Apply Image. Set the layer to your Low Freq layer, blending to Subtract, Scale 2, Offset 128. If you’re working in 32-bit mode the math changes, but for most beauty work we’re in 16-bit and these values hold. Change the High Freq layer’s blend mode to Linear Light. Your image should now look identical to where you started. The two layers are mathematically reconstructing the original, but their information is now separated.

How to Actually Retouch on Each Layer

On the Low Freq layer, I use the Lasso tool with a feather of 15 to 25 pixels (again, scale this to your file resolution) to select areas of uneven tone. Then I run Filter > Blur > Gaussian Blur at a very low radius, around 3 to 5 pixels, to blend the selected region into its surroundings. This evens out redness, shadows from blemishes, and uneven skin tone without touching a single pore. You can also use the Clone Stamp or Healing Brush here set to Current Layer if you prefer a more targeted approach.

On the High Freq layer, texture correction lives. The Clone Stamp tool set to 100% opacity, Current Layer, and a small hard-edged brush is your best friend here. We’re sampling clean texture from nearby areas and painting it over problem spots. Because texture has no color information on this layer, you’ll see it rendered as a gray-neutral pattern. You’re not matching tone here. You’re only matching the surface.

The Mistake I See in Almost Every Beginner’s Work

When I teach workshops, the most common error I see is over-retouching the Low Freq layer. Students will paint over large sections with a soft brush trying to eliminate every tonal variation, and then wonder why the final result still looks plastic. Flat, uniform tone reads as artificial, even when the texture is intact. The skin on a real face has micro-variations in warmth and saturation. They shift across the forehead, around the nose, below the eyes. We’re not trying to make the face one color. We’re trying to make the transitions smooth and gradual.

A good rule: if you’re retouching a problem area on the Low Freq layer and the selection covers more than about 10 to 15% of the face, split it into smaller overlapping selections with slightly different blur values. You’ll maintain the natural variation while still cleaning up the issue.

Keeping Texture Believable When You Clone on High Freq

The High Freq layer is powerful and easy to overwork. When I clone texture, I always sample from the same facial plane, the forehead should source from forehead, the cheek from cheek, because the texture pattern and direction shifts across the face. Cloning cheek pores onto a forehead reads subtly wrong even when clients can’t name why. I also keep my brush size small, rarely larger than the blemish I’m covering plus a few pixels of margin.

One more thing we do: zoom out to 50% or 25% after every few strokes. It’s easy to lose your bearings at 200% and create a patch that looks clean up close but reads as a texture island when you step back.

The single most important thing frequency separation teaches you is that good skin retouching is about preserving variation, not eliminating it. The moment you start protecting texture and managing tone separately, your edits stop looking like corrections and start looking like the person just had a very good day.