While I’m by NO means an expert retoucher, I’ve been using Frequency Separation for several years now. I might have come across it initially while on a CreativeLive binge, probably watching Pratik Nayak or Lindsay Adler. I used to do a good bit of high school senior photography, and it really came in handy! But I’ve found that for certain images of families and even children, Frequency Separation is just too effective not to use.
Did I mention it’s free? Yup – totally free. You don’t have to buy an action pack or spend big bucks on Portraiture, you just need to learn the technique and create your own action OR I’ll share the one that I use (that I created) at the bottom of this post (along with a PDF printable step by step guide.)
Look – I’m not gonna get all technical in this blog post. If you want the technical stuff, please do some googling or youtube searching. I’m just literally going to tell you what I know, off the top of my head. I’m gonna keep it easy.
Skin issues come in several varieties. You can have problems with skin texture (wrinkles, blemishes, etc) and skin coloring in the form of blotchiness. You can also have color issues when it comes to having skin that is too warm or cool, too saturated, etc. In this blog post, I won’t get too far into that. I personally use the CMYK Method taught by Sarah Wilkerson. It’s weird to wrap your brain around initially, but once you start using it, and if you use it often, it’ll become second nature. For me, CMYK is super easy and effective.
Let’s get back to Frequency Separation. We’re going to deal with texture and blotchiness. Pretty much everyone has texture and blotchiness issues. For SURE if you’re shooting newborns and even kiddos, you will run into skin issues. And mamas love the glamour skin treatment too, so don’t forget about them.
For this tutorial, I’m going to use images from a recent senior shoot. I used FS for almost every image in this session. It’s super easy to see where the texture and blotchiness issues were and how FS resolved them.
I must put out a disclaimer that this girl was GORGEOUS and her skin was nearly perfect. In fact, her mom expressed to me that she was feeling a little self-conscious about her skin before the session. When I looked at her, with my own eyes, I honestly didn’t see a single blemish. But the camera picks up all the little details, and it wasn’t until editing and zooming in that I saw areas that I could smooth out for her.
This technique will take longer than other methods you might be using. However, the goal of FS and the result is that you can smooth and retouch skin without losing texture. Overly smooth skin looks weird and fake and plastic. You know this…I’ve seen my Facebook friends post selfies using some app that smooths over their skin and it just looks awful. Then, they’ll use that app on pictures of their kids and I just want to yell “WHY??? It looks terrible. If you don’t know if your skin is too smoothed in your images, just ask me. I’ll shoot ya straight. Also inquire as to whether or not your eyes are too sharp. Another massive pet peeve of mine.
Ok – in a nutshell, the frequency separation technique creates two layers – one that contains skin texture data (the top layer) and one that contains skin color data (the lower layer).
To combat texture issues, I use a hard spot healing brush just a tiny bit larger than the blemish I’m removing. I change my brush size often. When I say hard, I’m talking in the neighborhood of 50-60% hardness.
To deal with blotchiness, we’ll do an initial gaussian blur (pronounced Gowssian, to rhyme with House-ian or Cow-sian. The S is pronounced as an S, not a zh. This is a mathematical term, not a photoshop term.)
Then, I’ll usually use my lasso tool, feathered to about 20-25 pixels, to select big chunks of skin (not detailed features) and do additional gaussian blurs until it looks right. For some images in this session, I ran a couple rounds of FS. Probably overkill, but the skin just didn’t look the way I wanted…yet.
Because you’re separating out the color and the texture, you can maintain good texture.
It’s important to note that this girl had amazing skin. If you are shooting a teen that really has some serious acne, I think it’s probably inevitable that the skin will look too soft/smooth. If there’s just not enough good texture, what can you do? You can’t create good skin texture out of thin air. (Although as I type this, you could add noise, which might give the appearance of a bit more texture. But that’s beyond the scope of where I was going with this article.)
I also want to note that there are many, many different variations on the FS technique. This is the way that I’ve found works best for me. And I’ve been doing it this way for awhile, and if it ain’t broke…
Frequency Separation Step by Step
- For the action to work properly, you’ll want to start with a fresh background layer.
I also recommend doing your FS BEFORE you do any further editing. This should be a first step, along with any distraction removal and cloning. Don’t do this last. This is first.
If you did some work prior to FS and you just cannot flatten, you could 1 of 2 things: Save your PSD layers and call it “Edit 1,” then flatten, and then finish your edit and save your layers as a PSD “Final Edit.” Or, you could unlock your background layer, group your previous layers, and create a Stamp Visible Layer by hitting Shift-Option-Command-E. Turn off your grouped layers. Click on your Stamp Visible pixel layer. Go up to Layer – New – New Background from Layer. (Or you could command J to duplicate your group, then right click and select Merge Group.)
Ok, so we should just have ONE background layer, yes? Ok, let’s move on.
- Run your action and then skip down to Step 14,
OR… To record an action, open up your Actions Panel. On my toolbar shown here, it’s the white triangle to the left of the histogram.
- Select “Create a New Action” which is the icon between the folder and the trash can (the square with an edget folded up.)
- A dialog box will appear, and you can fill the info. Name it what you like, you can add it to a set (if you want to create a new set first, you’d click the icon above that looks like a white folder), assign a function key which will act as a shortcut (note to self, need to do that) and color code it if you like. Then, hit Record.When you hit record, you’re on! You’ll now start recording your action in Step 5.
- Hit command J twice, which will duplicate your background layer twice.
- Rename the layers. The top layer can be called High or Texture and the bottom layer can be called Low or Color. You’ll want to do this step, even if you don’t often rename your layers – you’ll need to use it in step 11.
- Turn off the top layer.
- Select the bottom layer.
- Go to Filter – Blur – Gaussian Blur. For the radius, don’t go nuts here. You’re looking to just go high enough that the skin colors start to blur into each other. For this session, I was around 4.1 for most of them. I think a good rule of thumb is between 2-4. For a close-up portrait, you might need to go higher.
- Select top Layer and make it active (hit the eyeball).
- Now, with that layer selected, click Image – Apply Image. It’s vital at this point that you check what bit depth you’re working with. Is it 8bit or 16bit? Here’s how to know for sure:Look at the top tab, and you’ll see a 16* there.Or, on the bottom left corner of your screen, you’ll, see “16bpc” which is for 16 bits. If you see an 8, you’re in 8bit.
Change the settings to the following FOR A 8 BIT IMAGE:
Change the settings to the following FOR A 16 BIT IMAGE:
Be sure to change: LAYER to the bottom layer, BLENDING to add or subtract (depending on your bit depth), and the scale and offset numbers. Just trust me on those numbers. That’s just what they need to be. This is the case for every image, so you’ll never have to change these or know what they mean.
- Group your layers by selecting both (hitting command when clicking the second layer) and hitting Command G.
- Now, stop your action recording, by hitting the white square to the left of the red circle recording button.
- Now, this is where the work begins. You’ve set up your Frequency Separation, now we need to correct texture and color. I start with the High/Texture layer first. I use the spot healing brush tool (bandaid icon with dashed lines, shortcut J, hit shift-J to toggle to the spot healing brush tool, as shortcut J has several tools that it points to, like the patch tool.)
Below is my brush settings for the Spot Healing Brush Tool.
Make sure your brush is about 50-60% hardness, and Sample All Layers UNchecked. Dab around to remove any imperfections in skin texture.
- Now, select the Low/Color layer. Hit L for Lasso, or click on your lasso tool. Feather it to 20-25 pixels.Use your lasso to draw/select big patches of skin, staying away from details, like eyes, eyebrows, lips, the rims of the nose, etc. Only big patches, like the cheeks, chin, forehead, hose, etc.
When you have your selection, click on Filter and then select Gaussian Blur, which is the last filter you used. Just use that same one again. It’ll remember your pixel radius settings.
So, it goes like this: Lasso chunks of skin, Filter, Gaussian blur, repeat. Repeat this as often as you need to, hitting all the skin areas. You can go over the same sections more than once, as needed. If it’s just not strong enough, go down to Blur – Gaussian Blur, and select a higher pixel radius and see if that does the trick.As I mentioned above, I sometimes run FS a couple times (not more than twice). I’d get to the end of editing the image and still was seeing too much blotchiness in the skin tone.
(OPTIONAL:You can also create a blank/transparent layer between the low and high frequency layers. Here, you can use the clone stamp tool (set to Sample All Layers) and fix any color issues that the Gaussian blur is missing.
On this layer, you can also use your brush tool to even out tone. Hit B for Brush, then hold down Option (mac) and click on the image to sample a color. I’d suggest double tapping on your color swatch/color picker and pushing the color up a smidge brighter/lighter. I’d lower the brush opacity to 50% and flow to 5%. Brush on skin, making sure to sample again if you’re in a new area where the skin tone is different. For example, the cheeks will be pinker, and you want them to stay that way. Don’t use a peachier color sample on that area. Also, the shadowed side of the face is a bit darker, and you probably want to keep that dimension. Don’t use the same color sample there. I’d suggest sampling on the forehead, each cheek separately, the chin separately and maybe the nose separately. The tops of the cheeks/under eye area will be lighter than the bottoms, and the middle cheek should be pinker. It only takes a couple seconds to take new samples, so do it wherever the color is slightly different.
Now, on that transparent layer that you’re brushing/cloning on, in between the High and Low layers, go ahead and add a Gaussian blur to that bad boy too. I just leave the pixel radius the same as what I was using previously, around 4 is good. That should be enough to blur out any brush strokes. Turn this layer on and off several times and lower the opacity of this layer as needed.)
(OPTIONAL #2: If the tone is still too blotchy after some gaussian blurring on the low layer, you can duplicate the low layer (select the Low Layer and hit Command-J). Now, select Filter – Blur – Surface Blur. Your radius and threshold numbers here will really depend on the image! I’d start with a radius of anywhere between 10-30. Slide your threshold all the way to 255, and then slide it down slowly until the colors don’t bleed into each other. On this image, I was happy with a radius of 12 and a threshold of 31.)
I included these two optional layers in the action. Remember, you don’t have to use them, but they’re there if you need them.
- Lastly, I’d simply create a stamp visible layer (to do this, you’ll hit shift-option-command-E on a mac) and then group the previous layers and turn them off. Then, make that stamp visible layer the new background layer. (Layer – New – Background from Layer.) That way, I had everything all together and could go back if needed.
Now, you’re ready to continue on with your editing!
BONUS: My very last step is one that I just LOVE and highly recommend…
Oh – let me backtrack for just a moment and let you know that I do adjust skin tones using the CMYK method and curves layers.
I also dodge and burn to contour the face.
But the very last step, the pièce de résistance, is… Creating another stamp visible layer (doesn’t need to be a background layer). Remember, to do this, you’ll hit shift-option-command-E on a mac.
Then, click on Filter, Camera Raw Filter. Go to the Details panel (two triangles) and move the luminance up to around 20 and slide the Luminance Detail to zero.
Click OK, and then add a black mask (option – mask, or invert a white mask by hitting command I) and brush on the areas of skin you want softened using a very soft white brush at 100% opacity and 50% flow.
Turn that layer on and off and reduce the layer opacity if needed.
There ya have it! Frequency Separation!
If you’d like to download my action, feel free! Just fill out the form below and you’ll be taken to a page with a dropbox link for you.
I’ve also included a Step by Step Printable PDF which includes instructions for recording your own action.
Let me know if you have any questions and please tag me on any images on Instagram that you use it on so I can see. (Please note, the action I use is for 16bit images. I will create one for 8bit images. I do believe that there are a couple settings that will be different between the two. I don’t believe you can use the 16bit action on an 8bit image. You can try…but I think you’ll run into problems. Let me know if you do and how it goes!