Black Friday Sale | Up to 45% Off

Post Production Tips

Frequency Separation | What Is It Really And When Should You Use It?

By Wendell Weithers on November 5th 2017

Frequency Separation is a powerful retouching technique used to soften a subject’s skin without degrading his or her skin texture. When executed well, it produces are a pristine photograph that gives your portfolio a higher level of professionalism. When executed poorly, it can suck the soul out of even the most attractive model.

Perhaps such retouching transgressions are committed because many don’t know why they are doing what they are doing. This tutorial from PiXimperfect can change that. If you want to add this skill to your retouching toolkit, know why you need it, and how best to use it, this video is for you.

[REWIND: Frequency Separation Pitfalls To Avoid | Beware Of Actions & Cautious With Radius]

In Frequency Separation you are addressing your subject’s skin tone/color and their skin texture separately in order to preserve both. Regarding texture, you’re addressing frequently occurring facial features such as pores, pimples, and other blemishes. With skin tone, you are dealing with the transitions of tones/colors on your subject’s face.

A good way to truly understand this tool is by answering a question. Which facial features change with a higher frequency and which features change with a lower frequency?

Survey says…
High-Frequency Changes – Skin Texture (pores, pimples, and other blemishes)

Low-Frequency Changes – (Transitions of tones/colors)

When should you use it?

Frequency Separation isn’t intended to remove major blemishes; rather, it should be used for less pronounced skin characteristics. Therefore, it should find a place in your workflow after exposure adjustments and major blemish removal but, before more stylistic edits such as dodging and burning. This is a part of your post-processing that is best used as a something to build earlier and not so late in the workflow that it becomes counterproductive.

PiXimperfect’s method.

1 – After you’ve made Exposure corrects and remove major blemishes create a merged layer and make a copy of it.

Merge Layer Shortcut
: CTRL + OPT + SHIFT + ‘E’
Windows: CTRL + ALT + SHIFT + ‘E’

One copied layer will be for Texture (High Frequency) and one will be for color/tone (Low Frequency).

2 – Blur the color/tone (Low Frequency) layer

Filter > Blur > Gaussian Blur. Then adjust the radius of the blur effect until the texture of your subject’s skin just goes away but, not beyond that. By blurring the image, you have effectively removed all of the high-frequency changes in the layer

3 – The next step is to “Subtract” the details from the (Low-Frequency Layer) leaving only the high frequencies.

Image > Apply Image > Select the “Color” Layer > Select Blending Mode “Subtract”.

Make sure the “Scale” is set to “2”, Channel is set to RGB, and the “Offset” is set to “128”

(Scale) = the amount of “Subtract” applied.

4- Set the Blend Mode of the Color Layer to Linear Light

5- Select the “Mixer Brush”

Make sure the brush is set to only mix existing colors.

Set Brush Wetness: 1-5%

Flow: 10-20%

Parting Tip: Be sure to leave the smile lines untouched, it keeps your image natural.

Additional Articles on Frequency Separation

Three Different Blend Modes to Use In Frequency Separation For Ease & Speed

Removing Tricky Color Casts On Skin + Free Frequency Separation Action

The Many Faces Of Frequency Separation. Are You Doing It Wrong?

Wendell is a business owner and contractor at Chick-fil-A coporate in Atlanta. When he isn’t shooting portraits and documenting important moments, he is shooting his wife’s work in their home cake studio in East Point, GA.

Instagram: Wendellwphoto

3 Comments

Please or register to post a comment.

  1. Riehan Bakkes

    Fantastic tut! Thanks Umesh for finally showing how its done properly!

    | |
  2. hugh mobley

    Umesh is great! I prefer using dust and scratches Umesh has action for D & S also, however since all skin is different surface blur works well sometimes, almost never use Gaussian blur any more

    | |