Frequency separation is one of those techniques that works beautifully when you understand what it’s actually doing, and produces disaster the moment you don’t. The bad reputation it carries in retouching circles is almost entirely self-inflicted — not a flaw in the technique, but the predictable result of applying it without a solid grasp of the underlying logic. Pore-less alabaster skin, flat faces, the early-80s mall portrait look: all frequency separation, all applied wrong.
This guide covers how to set it up correctly, the two points where most people go wrong, how to use it for a specific and genuinely difficult problem (blown-out hot spots), and the philosophy that should govern when you reach for it at all.
A practical resource covering Photoshop retouching, Lightroom workflow, and post-processing technique for photographers.
What frequency separation actually does
The name describes the process exactly. You separate an image into two layers based on the frequency of detail: a low-frequency layer that holds color and tonal transitions, and a high-frequency layer that holds texture and fine detail — pores, individual hairs, blemishes, surface grain. Because these two types of information now live on separate layers, you can edit one without disturbing the other.
That’s the whole idea. Edit tonal transitions without smearing texture. Fix skin texture without altering the underlying color. Done with restraint, the result looks like natural skin that happens to be well-lit. Done without it, you get the plastic fantastic.
One thing worth stating clearly before going further: frequency separation is not an everyday tool. It takes time to do correctly, and most images simply don’t need it. A solid Lightroom workflow handles the majority of skin retouching. Frequency separation earns its place for specific, difficult problems — not as a default step applied to every portrait.
How to set up the layers
There are variations in how people build the frequency separation stack, but the core structure is the same across all of them.
- Make all your exposure corrections and remove any major blemishes first. Frequency separation works on what’s left — it isn’t designed to replace foundational corrections.
- Create a merged copy of all your existing layers. Mac shortcut: CMD + OPT + SHIFT + E. Windows: CTRL + ALT + SHIFT + E. Alternatively, go to Layer > Merge Visible.
- Duplicate that merged layer so you have two identical copies.
- Name the bottom copy Low and the top copy High. Putting both inside a Layer Group named “F/S” makes it easy to toggle the effect on and off to check your work.
Building the Low layer
Select the Low layer. You’re going to blur it until the texture disappears and only the color and tonal information remains. Two filters work well here:
- Gaussian Blur (Filter > Blur > Gaussian Blur) — the most common choice, straightforward to control.
- Median filter (Filter > Noise > Median) — preserves edge definition better than Gaussian, which helps maintain the lines that define facial structure. Worth trying on portraits where edge clarity matters.
Adjust the radius until the skin texture just falls away. Stop there. More on why this radius decision matters so much in the next section.
Building the High layer
Select the High layer, then go to Image > Apply Image. Set it up as follows:
- Layer: your Low layer
- Blending: Subtract
- Scale: 2
- Channel: RGB
- Offset: 128
The result should be a mostly gray layer with light and dark areas representing the texture and detail of the image — similar in appearance to a High Pass filter result. Set the blend mode of this layer to Linear Light. When combined with the Low layer, the two should produce an image that looks identical to what you started with. Toggle the Layer Group on and off to confirm.
Before working on either layer, make a copy of both the High and Low layers clipped to the originals. Work on the copies. If you push something too far, you can delete the copy and start again without losing the base layers.
The two mistakes that ruin most frequency separation work
Mistake 1: using actions without understanding the process
Frequency separation actions are everywhere and they’re genuinely useful once you understand what’s happening at each step. The problem is that actions skip the education entirely. You run them, layers appear, and you start painting — with no real understanding of what the blur radius is doing, why the Apply Image settings are what they are, or what you’re actually looking for as you work.
The recommendation here is straightforward: build the layers manually until the process becomes automatic. Once you’ve done it enough times that you could explain each step to someone else, actions become a legitimate time-saver. Before that point, they’re a shortcut to results you can’t diagnose or fix.
Mistake 2: setting the wrong blur radius on the Low layer
This is the one that produces most of the bad frequency separation work you see. The blur radius on the Low layer is not a fixed number — it varies with the resolution of your camera, the resolution of the image, and how large the subject is in the frame. A radius that’s right for a tight facial crop on a 45MP file will be completely wrong for a full-body swimwear shot on the same camera.
Images above by Kishore Sawh.
What you’re looking for is a specific visual threshold: blur until the texture disappears, but stop the moment the tonal transitions that define facial structure — the shadow under the cheekbone, the edge of the nose, the hairline — start to soften. Go past that point and you’ve blurred away the very information that gives a face its three-dimensional quality. The flat, unnatural look that most people associate with bad frequency separation almost always traces back to a radius that was pushed too far.
If the subject takes up a smaller portion of the frame, the skin area is proportionally smaller and you can generally get away with a lower radius. The larger and higher-resolution the subject in the frame, expect to use a higher value. Judge it visually every time. There is no correct number to copy from a tutorial.
One more thing: you don’t have to run frequency separation on an image only once. Running it multiple times with conservative settings is far safer than one aggressive pass. The less-is-more principle applies at every stage.
Working on the layers: how to retouch without creating plastic skin
Once the layers are set up, there are two main approaches to working on the Low layer. The first uses the Lasso tool to select areas and applies a Gaussian Blur to smooth tonal transitions. It works, but it offers limited control and makes it easy to overshoot into plastic-looking territory.
The more precise method uses the Mixer Brush directly on the Low layer copy. Set it up as follows:
- Enable “Clean brush after each stroke” and “Load brush after each stroke” — this prevents the brush from carrying over previously sampled color between strokes.
- Wet: start around 20%. This controls how much color the brush picks up at the start of a stroke, letting you build up the blend gradually.
- Flow: 10–20%. Keeps the effect subtle and controllable.
- Sample All Layers: off. This lets you paint on the Low layer copy while the High layer remains visible, so you can see your progress in context.
Work with light strokes from the outside of the problem area inward. Small circular strokes often help in tight areas. Because you’re on a copy of the Low layer, going too far costs nothing — delete the copy and start again from the original.
On the High layer copy, use the Healing Brush and Rubber Stamp to sample and transplant texture from surrounding areas. Set both tools to sample “Current Layer” only. Work carefully about where you’re sampling from to avoid obvious repetition in the texture pattern.
Leave smile lines, laugh lines, and other natural facial features alone. Removing them is what produces the ageless-but-wrong look that clients notice even if they can’t articulate why.
A practical use case: fixing blown-out hot spots
Hot spots — areas where specular highlights have blown out both the tone and the skin texture — are one of the most genuinely difficult retouching problems. Cloning or healing alone tends to produce smudgy artifacts. Frequency separation’s divide-and-conquer approach handles it cleanly because you can address the missing tone and the missing texture as separate problems.
Images in this section by Dennis Dunbar.
Step 1: blend out the tone on the Low layer
Using the Mixer Brush on your Low layer copy, work from the outside of the hot spot inward, pulling the surrounding skin tone into the blown-out area. The gradual build-up approach works better here than trying to correct it in one or two strokes.
Step 2: rebuild the texture on the High layer
Switch to the High layer copy. Use the Healing Brush and Rubber Stamp set to “Current Layer” to sample skin texture from the surrounding area and paint it into the hot spot region. Work from multiple source points to avoid tiled repetition in the result.
The philosophy: when to use it and when not to
Frequency separation has been oversold as a comprehensive skin retouching solution. It isn’t. It’s a precision tool for specific problems, and the decision about when to reach for it matters as much as knowing how to use it.
Going in to any retouch, the goal is natural-looking skin — not perfect skin. Even in beauty retouching, where a higher level of refinement is expected, the aim is to retain the detail that makes skin look real. The moment texture disappears entirely, the image starts to look wrong in ways that viewers feel even if they can’t identify the cause.
A useful framework for when frequency separation is actually warranted:
- Major blemishes and exposure corrections have already been handled elsewhere in the workflow.
- The remaining skin issues are too subtle for healing or cloning to address cleanly.
- The image is going to be scrutinized at a level that justifies the time investment — beauty campaigns, editorial work, key hero shots.
For the vast majority of portrait and wedding work, a well-built Lightroom workflow handles retouching without touching Photoshop at all. Frequency separation earns its place at the end of that process, on the images that genuinely need it, applied with a light touch.
Frequently asked questions about frequency separation
What is frequency separation in Photoshop?
Frequency separation is a retouching technique that splits an image into two layers: a Low layer containing color and tonal information, and a High layer containing texture and fine detail. Editing the layers independently lets you smooth skin tones without affecting texture, or fix texture without altering color.
What blur radius should I use for frequency separation?
There is no fixed number. The correct radius is the one that blurs the Low layer just enough to remove texture while preserving the tonal transitions that define facial structure. It varies with image resolution, camera resolution, and how large the subject is in the frame. Judge it visually every time.
What’s the difference between Gaussian Blur and Median filter for the Low layer?
Both work. Gaussian Blur is more common and straightforward to control. The Median filter does a better job of preserving edge definition — the lines that separate a hairline from skin, or define the contour of a nose — which can help maintain facial structure in portrait work.
How do I avoid the plastic skin look with frequency separation?
Set a conservative blur radius on the Low layer, work with low Mixer Brush wetness and flow settings, and err toward multiple light passes rather than one aggressive correction. Leave natural facial features like smile lines intact. The most common cause of plastic skin is a blur radius that was pushed too far on the Low layer.
When should I use frequency separation?
After exposure corrections and major blemish removal, on images that genuinely need refined skin work — typically beauty, editorial, or hero shots where the retouching will be scrutinized. It’s not a default step for every portrait. Most images don’t need it, and the time investment is only justified when the image warrants that level of finish.
