Frequency Separation Skin Retouching

Frequency separation is the technique that separates a portrait into two layers: one carrying tone and color, one carrying texture and detail. Once split, you can smooth uneven skin tones on the bottom layer without erasing pores, hair, or fine detail on the top layer. The result is skin that looks retouched instead of plastic.

This tutorial sets up frequency separation manually so you understand what each step is doing. Once you get it, save the steps as a Photoshop action and you will never set it up by hand again.

Setting up the layers

Step 1: Duplicate the base layer twice

Press Ctrl/Cmd + J twice to duplicate your base layer. Rename the bottom copy Low Frequency and the top copy High Frequency. Hide the High Frequency layer for now.

Step 2: Blur the Low Frequency layer

Select Low Frequency. Go to Filter > Blur > Gaussian Blur. Set the radius to a value where the skin tones blur out but the major facial features (eyes, mouth) are still recognizable. For a typical portrait shot at 4000 pixels wide, this is around 8-12 pixels. Click OK.

Step 3: Show High Frequency and run Apply Image

Show the High Frequency layer and select it. Go to Image > Apply Image. In the dialog: Layer = Low Frequency, Blending = Subtract, Scale = 2, Offset = 128. Click OK.

The High Frequency layer now looks gray with faint outlines of pores and texture. That gray sheet IS the high-frequency detail of the photo, isolated.

Step 4: Set High Frequency blend mode to Linear Light

With High Frequency selected, change the blend mode at the top of the Layers panel from Normal to Linear Light. The image snaps back to its original appearance. The texture from High is now sitting on top of the smooth tones from Low.

Retouching the layers

Step 5: Smooth uneven tones on the Low Frequency layer

Select the Low Frequency layer. Grab the Mixer Brush (B, then cycle to Mixer Brush). Set Wet to about 30%, Load to 30%, Mix to 30%. Paint over uneven skin areas. The Mixer Brush samples nearby tones and blends them together. Because you are working on Low Frequency, only the underlying color smooths; pores stay sharp because they live on the High Frequency layer above.

Step 6: Spot-clean the High Frequency layer

Switch to High Frequency. Use the Healing Brush (J) on individual blemishes, stray hairs, and skin imperfections. Sample a clean area of similar texture and click on the blemish. Because the layer is texture-only, your healing fixes only that detail and does not muddy the skin tone underneath.

What frequency separation is not

Frequency separation is for fine retouching. Do not use it to fundamentally reshape a face or change someone's appearance. Used correctly, viewers cannot tell a photo was retouched. Used heavily, it produces the over-smoothed plastic look that ruined a decade of magazine covers.

When frequency separation is the right tool

Frequency separation pays off on portraits and beauty work where you need to even out skin tone without losing skin texture. That is its specific job. If the goal is a magazine cover, a professional headshot, an actor's portfolio, or a product launch where a face is the hero, this is the technique.

It is wrong for almost everything else. Casual portraits do not need it (the skin texture is part of the realism). Documentary work does not need it (smoothing skin in photojournalism is an ethical problem). Landscapes, products, food, and architecture do not need it at all. Reach for frequency separation when the assignment specifically calls for retouched skin, and not before.

Common mistakes that produce the plastic look

The first mistake is too-aggressive Gaussian Blur on the Low Frequency layer. If you blur enough that the eyes, mouth, and nose lose their position cues, the entire face becomes mush when you bring the High Frequency layer back on top. Stay in the 8-12 pixel range for a 4000px image. Smaller files need less, larger files need more, but the major facial features should still be readable in the blurred preview.

The second mistake is over-painting the Low Frequency layer with the Mixer Brush. The technique smooths color and tone underneath the texture, but if you paint too aggressively, you wipe out shadows that give the face its three-dimensional shape. Drop the brush strength to around 30% and work in passes rather than trying to flatten everything in one stroke.

The third mistake is treating the High Frequency layer like a healing layer for everything. Use it for individual blemishes, stray hairs, and small skin imperfections. Do not use it to remove laugh lines, freckles, or other features that define how someone looks. The line between "retouched" and "erased" is whether the person still looks like themselves.

Adapting the technique to different sources

For a smaller image (web resolution, around 1500-2000 pixels wide): drop the Gaussian Blur to 4-6 pixels. The principle is identical, just scaled down.

For a larger image (print resolution, 6000+ pixels wide): increase the Gaussian Blur to 14-18 pixels, and consider running the Apply Image step at Scale 2, Offset 128 as written, then refining at higher resolution.

For an action that handles this automatically: record yourself doing Steps 1 through 4 as a Photoshop action. Future portraits get the layer structure in one button press. The retouching on Steps 5 and 6 still has to be done by hand, because that is the part that requires judgment.

For team workflows: save the action set as an ATN file and share it with collaborators. Everyone hitting the same starting setup means the team's portraits are consistent.

Frequently asked questions

Why Linear Light instead of just Overlay?

Linear Light is the blend mode that, combined with Subtract at Scale 2 Offset 128 in Apply Image, mathematically reconstructs the original image exactly. Overlay almost works but introduces a slight contrast shift. If you have ever done frequency separation and felt like the result is contrastier than the original, you used Overlay instead of Linear Light.

Can I do this on video?

Not directly in Photoshop. DaVinci Resolve has a Beauty effect that approximates frequency separation on footage. The principle is identical, the implementation is different.

Should I work on RAW or JPEG?

RAW is significantly better. Frequency separation depends on smooth gradients in the Low Frequency layer, and JPEG compression artifacts (8x8 block boundaries, color subsampling) show up as visible patches when you smooth across them. If you have to work on JPEG, do not push the technique as hard.

Is this technique going to be replaced by AI tools?

For casual users, probably yes. AI skin smoothers built into mobile apps and editing suites can produce a similar-looking result in one click. For professional retouchers, frequency separation will stay relevant because it offers surgical control that AI tools cannot match. The control matters when the client is paying for a specific look and not "something that resembles smoothed skin."