Tthe healing brush. It behaves strangely if you move it too close to an area of significently different lightness or color, sucking the nearby pixels into the area you're trying to fix.
I found a way for that, and it's making a sleection before healing. It works perfect for remove the hair from the face of the subject. After healing at will, it will look odd in color and lightiness, but you can duplicate the layer where you were healing and set it to screen. By using the opacity slider you can match the skin brightness (specially in the borders which are between the hair and the face shape). After that, create a top layer, deselect and use this new layer to heal the rough borders that the selection have left and smooth the transition between.
It requires a bit of time to understand what's happens, but it really works and it's quick
The technique I use for skin smoothing is similar to what you guys do here. Healing brush first, for the obvious imperfections. Then a blur layer. The only thing I do different is I go in great detail to retain the hair. Some skin imperfections, like the one of the young girl below, are part of the character and shouldn't be removed.
But, You may have the need to remove all skin imperfections. This takes a lot of time.
The botom line, you have to make it look real. That last picture was with this effect overdone. Looks nice, but it's not real.
This is a better example of 'close to real perfect skin'
The more detail you try to retain, the longer it will take. It all comes down to time vs. detail. You'll want to mask the eyes, hair lips, nostrils, etc. Things that are no supposed to be smooth should be masked to original.
Isn't there software/ plugins that automate this process? Some of us are lazy!
Yes, Portrait Professional is another. But... to be honest, you still need photoshop depending on how fancy you want to go. The RAW rendering in PP is terrible, so you'll want to open up your raw file some other way.
I usually do a quick healing brush on a tif in PS, then send the file to PP. I generally run the skin slider to zero on PP (it still does some skin smoothing) but what is GREAT about PP is it's facial structure enhancements and how easy they are to use.
Then it's back to PS where I may do some additional retouching (mostly based on techniques from "Skin" by Lee Varis).
For samples, look at my "beauty" port on my website.
I have some personal reflections and tips regarding more subtle retouching of portraits, if someone could perhaps point a newbie to a guide on how to include pictures in my posts... Is the "include" picture option unlimited in size?
[edit] - DUH!.... Found it, in the most obvious place - of course. Sorry, post upcoming.
I've been lurking a while, and learned some great stuff around here, but this is a personal twist that I haven't seen - so I thought I'd share... This is mostly about a way to apply a more discrete PP, not all subjects like to be reshaped like commercial make-up dolls. I try to retain original skin structure and finer hair while smoothing overall shapes. This fits some types of work and clients better.
The basics are: 1 - Divide the picture into its different detail frequencies. 2 - "even out" the lowest layer which only contains very rough surfaces and no detail. 3 - Work on the remaining (higher) frequencies by masking their layers.
This may be kind of counterintuitive, but it works really well, and can be made semi-automatic. But the "Healing Brush" before doing anything else is mandatory!
First, I will show what the technique can do when on fully automatic. All the steps are customizable, and their effect can be manually adjusted in the final stage. I'll write a step-by-step instruction later tonight.
1. Make sure you have the original in a single bottom layer
2. Make two layer copies... They are going to be used for higher frequency overlays
Now, we're going to prepare the topmost layer which will contain the finest detail (highest frequencies)
3. Apply a Gaussian R2 on the two lower layers. This is important for next stage, be sure to do BOTH.
4. Select the topmost layer and "Apply Image" with the settings in [frame 3]
5. Set the blending type to "Linear Light" [Frame 4]
As you can see now; If you toggle the top layer visibility on/off, the layer beneath is the original picture - but blurred with a Gauss R2 - and in the topmost layer you have the detail that the Gauss "blurred away" as an overlay that modifies the blurred image back to its original shape. This can be useful!
Now turn the visibility ON for the two lowest layers and OFF for the topmost layer. This isn't important, but will make the next step more visible.
6. Select the bottom layer, and do a Gauss R5 on it. [Frame 5]
7. Select the middle layer, and "Apply Image" with the same settings as in [Frame 3]. Remember to turn the blending mode to "Linear Light" when you're done...
In the middle layer we now have all the detail in the picture that can be contained between the frequencies that a Gauss R2 and a Gauss R5 contains - i.e. the medium frequency contrasts.
Try toggling the modifying layers on/off and study what the layers contain... When all layers are visible, your picture should look exactly like the original picture - otherwise something's wrong.
Now we have prepared the picture for some real work...
With only the bottom layer visible and selected, make a new copy of it - the new copy should place itself second from the bottom. We're now going to work on this copy. Your layers should now look like [Frame 6]
8. Select second layer, do a "Surface Blur" with R12, Threshold 6. Then a Gauss R3.
The reason for the Gaussian blur is that the Surface Blur actually sharpens the edges in the very blurred layer, as it tries to blend pixels with pixels that are roughly the same value as itself. Edges between large areas of colour then get "compressed" as the areas tend to spread out towards eachother... We don't want this.
The Surface Blurred/Gaussed copy of the bottom layer now needs to be blended back into the bottom layer, and we'll do this with a mask to restrict the general flattening to the skintones.
9. With only the bottom and the second layer visible, and the second layer selected, do a "Select Colour Range" from the menu "Select". Click somehwere in the normal skintones of the picture and adjust "Fuzziness" until you can see that most of the face is selected (around 80?). Do try to use a lower Fuzzines and hold shift-key pressed while clicking on several spots in the face, this might give better results. The selection mask should look something like [Frame 7] Click "ok"
10. You now have a selection, and you can convert this selection into a mask just by pressing the "Mask" button in the Layers sidebar. The mask button is indicated in [Frame 5].
This mask is reusable in the top two modifying layers, so:
11. Copy it into the third layer from the bottom by pressing [alt] and drag it into the right place next to the layer Icon. Now turn the visibilty of the layer second from the top ON. As you can see the layer mask is masking the wrong sections of the image - we want it to mask the parts now visible... So invert it:
12. Invert the mask by selecting it and press [ctrl-I] or choose the menu Image/Adjustments/Invert
Your layers should now look like [Frame 8]. The masking effect is probably way to strong, we want to keep lots more of the original image!
13. With the mask selected, apply a "curves" to the mask and lift the left end of the curve about 2/3 up - like [Frame 9].
Now that's more like it. Now copy the mask from this layer up onto the top layer and turn on that layer's visibility too....
Finished.
The masks are of course customizable, from the coverage (the initial selection in [Select/Colour Range] to the intensities in the masks, which you can vary with applying curves on the masks themselves. You can also paint Black or White over parts in the mask that you want to show/hide more...
You select the scale of the detail you work on by varying the initial two Gaussian Blur Radiuses, R2 and R5 suits a smooth relatively small crop (face is ~1000 pixels high) like this fine. You can experiment by changing the SECOND value away from "5" - Less than 4 is meaningless, and more than 8 will probably be to much...
Hope anyone stayed until the end, and that someone has some use for some of the techniques applied in this, my first online tutorial... Happy PShop'ing to y'all.
And no - Picasa is not a good place to publish pictures into other websites from... They refuse to keep stable addresses to the individual pictures! Drives me nuts.
And an example of a finished result, no additional manual work added, except for the standard values I've added in my Action. The picture has been sharpened for web, otherwise nothing.
Of course the actions workmode can be inverted so that the modifying layers subtract from an untouched/unblurred BackGround layer, but that's another story...
For Adobe Lightroom 2 users, it is much simpler. The new adjustment brush can apply local adjustments, including skin softening (reduced sharpening and clarity). The brush creates a mask, and the adjustments are applied to the masked area. You can creates multiple adjustment masks (they can overlap), and adjust each one's parameters independendently, at any any time. You can remove any adjustment mask at any time. All non-destructively. LR2 uses the same raw processing engine as Adobe Camera Raw (PS's raw engine). The interface is a little different, but the results are identical. The local brush adjustments are just added into the mix when the image is rendered from the original file.
The new local adjustment brush in Lightroom 2 rocks! Don't overlook it.
Portrait Professional 6 Results
Wanted to share and get feedback on the results I got from Portrait Professional 6. This is a shot of my sister which I snapped of her at a recent party. I did some post to image in LR2... the usual, adjusted the curves a bit for more pop and stuff. I then ran the image through Potrait Professional. I went through the wizard that has you select landmarks on the persons face. After selecting the landmarks, it processes the image through some algorithms and adjusts the facial structure and texture of the skin. This is the before and after with the default automatic settings sans the facial structure adjustments.
Before:
After:
I personally think the results are pretty damn good considering it took me all of 2 minutes to select the landmarks and Portrait Professional to make the skin adjustments automatically. I did absolutely nothing to the image after PP was done making the auto enhancements.
Comments
I found a way for that, and it's making a sleection before healing. It works perfect for remove the hair from the face of the subject. After healing at will, it will look odd in color and lightiness, but you can duplicate the layer where you were healing and set it to screen. By using the opacity slider you can match the skin brightness (specially in the borders which are between the hair and the face shape). After that, create a top layer, deselect and use this new layer to heal the rough borders that the selection have left and smooth the transition between.
It requires a bit of time to understand what's happens, but it really works and it's quick
But, You may have the need to remove all skin imperfections. This takes a lot of time.
The botom line, you have to make it look real. That last picture was with this effect overdone. Looks nice, but it's not real.
This is a better example of 'close to real perfect skin'
The more detail you try to retain, the longer it will take. It all comes down to time vs. detail. You'll want to mask the eyes, hair lips, nostrils, etc. Things that are no supposed to be smooth should be masked to original.
Nikon Shooter
It's all about the moment...
Las Cruces Photographer / Las Cruces Wedding Photographer
Other site
www.fastcatstudio.com
www.fastcatstudio.net - blog
Yes, Portrait Professional is another. But... to be honest, you still need photoshop depending on how fancy you want to go. The RAW rendering in PP is terrible, so you'll want to open up your raw file some other way.
I usually do a quick healing brush on a tif in PS, then send the file to PP. I generally run the skin slider to zero on PP (it still does some skin smoothing) but what is GREAT about PP is it's facial structure enhancements and how easy they are to use.
Then it's back to PS where I may do some additional retouching (mostly based on techniques from "Skin" by Lee Varis).
For samples, look at my "beauty" port on my website.
Portland, Oregon Photographer Pete Springer
website blog instagram facebook g+
[edit] - DUH!.... Found it, in the most obvious place - of course. Sorry, post upcoming.
The basics are:
1 - Divide the picture into its different detail frequencies.
2 - "even out" the lowest layer which only contains very rough surfaces and no detail.
3 - Work on the remaining (higher) frequencies by masking their layers.
This may be kind of counterintuitive, but it works really well, and can be made semi-automatic. But the "Healing Brush" before doing anything else is mandatory!
First, I will show what the technique can do when on fully automatic. All the steps are customizable, and their effect can be manually adjusted in the final stage. I'll write a step-by-step instruction later tonight.
1. Make sure you have the original in a single bottom layer
2. Make two layer copies... They are going to be used for higher frequency overlays
Now, we're going to prepare the topmost layer which will contain the finest detail (highest frequencies)
3. Apply a Gaussian R2 on the two lower layers. This is important for next stage, be sure to do BOTH.
4. Select the topmost layer and "Apply Image" with the settings in [frame 3]
5. Set the blending type to "Linear Light" [Frame 4]
<table style="width: auto;"><tbody><tr><td>
As you can see now; If you toggle the top layer visibility on/off, the layer beneath is the original picture - but blurred with a Gauss R2 - and in the topmost layer you have the detail that the Gauss "blurred away" as an overlay that modifies the blurred image back to its original shape. This can be useful!
Now turn the visibility ON for the two lowest layers and OFF for the topmost layer. This isn't important, but will make the next step more visible.
6. Select the bottom layer, and do a Gauss R5 on it. [Frame 5]
7. Select the middle layer, and "Apply Image" with the same settings as in [Frame 3]. Remember to turn the blending mode to "Linear Light" when you're done...
In the middle layer we now have all the detail in the picture that can be contained between the frequencies that a Gauss R2 and a Gauss R5 contains - i.e. the medium frequency contrasts.
Try toggling the modifying layers on/off and study what the layers contain... When all layers are visible, your picture should look exactly like the original picture - otherwise something's wrong.
Now we have prepared the picture for some real work...
With only the bottom layer visible and selected, make a new copy of it - the new copy should place itself second from the bottom. We're now going to work on this copy. Your layers should now look like [Frame 6]
8. Select second layer, do a "Surface Blur" with R12, Threshold 6. Then a Gauss R3.
The reason for the Gaussian blur is that the Surface Blur actually sharpens the edges in the very blurred layer, as it tries to blend pixels with pixels that are roughly the same value as itself. Edges between large areas of colour then get "compressed" as the areas tend to spread out towards eachother... We don't want this.
The Surface Blurred/Gaussed copy of the bottom layer now needs to be blended back into the bottom layer, and we'll do this with a mask to restrict the general flattening to the skintones.
9. With only the bottom and the second layer visible, and the second layer selected, do a "Select Colour Range" from the menu "Select". Click somehwere in the normal skintones of the picture and adjust "Fuzziness" until you can see that most of the face is selected (around 80?). Do try to use a lower Fuzzines and hold shift-key pressed while clicking on several spots in the face, this might give better results. The selection mask should look something like [Frame 7] Click "ok"
<table style="width: auto;"><tbody><tr><td>
10. You now have a selection, and you can convert this selection into a mask just by pressing the "Mask" button in the Layers sidebar. The mask button is indicated in [Frame 5].
This mask is reusable in the top two modifying layers, so:
11. Copy it into the third layer from the bottom by pressing [alt] and drag it into the right place next to the layer Icon. Now turn the visibilty of the layer second from the top ON. As you can see the layer mask is masking the wrong sections of the image - we want it to mask the parts now visible... So invert it:
12. Invert the mask by selecting it and press [ctrl-I] or choose the menu Image/Adjustments/Invert
Your layers should now look like [Frame 8]. The masking effect is probably way to strong, we want to keep lots more of the original image!
13. With the mask selected, apply a "curves" to the mask and lift the left end of the curve about 2/3 up - like [Frame 9].
<table style="width: auto;"><tbody><tr><td>
Now that's more like it. Now copy the mask from this layer up onto the top layer and turn on that layer's visibility too....
Finished.
The masks are of course customizable, from the coverage (the initial selection in [Select/Colour Range] to the intensities in the masks, which you can vary with applying curves on the masks themselves. You can also paint Black or White over parts in the mask that you want to show/hide more...
You select the scale of the detail you work on by varying the initial two Gaussian Blur Radiuses, R2 and R5 suits a smooth relatively small crop (face is ~1000 pixels high) like this fine. You can experiment by changing the SECOND value away from "5" - Less than 4 is meaningless, and more than 8 will probably be to much...
Hope anyone stayed until the end, and that someone has some use for some of the techniques applied in this, my first online tutorial... Happy PShop'ing to y'all.
And no - Picasa is not a good place to publish pictures into other websites from... They refuse to keep stable addresses to the individual pictures! Drives me nuts.
Of course the actions workmode can be inverted so that the modifying layers subtract from an untouched/unblurred BackGround layer, but that's another story...
New Portfolio
For Adobe Lightroom 2 users, it is much simpler. The new adjustment brush can apply local adjustments, including skin softening (reduced sharpening and clarity). The brush creates a mask, and the adjustments are applied to the masked area. You can creates multiple adjustment masks (they can overlap), and adjust each one's parameters independendently, at any any time. You can remove any adjustment mask at any time. All non-destructively. LR2 uses the same raw processing engine as Adobe Camera Raw (PS's raw engine). The interface is a little different, but the results are identical. The local brush adjustments are just added into the mix when the image is rendered from the original file.
The new local adjustment brush in Lightroom 2 rocks! Don't overlook it.
Peace,
Thunder Rabbit
www.thunderrabbitgrfx.com
Lee
Thunder Rabbit GRFX
www.thunderrabbitgrfx.com
Wanted to share and get feedback on the results I got from Portrait Professional 6. This is a shot of my sister which I snapped of her at a recent party. I did some post to image in LR2... the usual, adjusted the curves a bit for more pop and stuff. I then ran the image through Potrait Professional. I went through the wizard that has you select landmarks on the persons face. After selecting the landmarks, it processes the image through some algorithms and adjusts the facial structure and texture of the skin. This is the before and after with the default automatic settings sans the facial structure adjustments.
Before:
After:
I personally think the results are pretty damn good considering it took me all of 2 minutes to select the landmarks and Portrait Professional to make the skin adjustments automatically. I did absolutely nothing to the image after PP was done making the auto enhancements.
What do you think?
Moderator of the People and Go Figure forums
My Smug Site