Critique - 1st attempt

[PeterWright]

Hi all

This is my first serious attempt at a beauty retouch using several of the tutorials on this site (Thanks..)

Primarily

Clean up
D&B - pore and sculpt (Twin curve layers)

I would really welcome any comments on what I have missed and how it might be improved - particularly with a focus on the skin.

Thanks in advance..

http://farm3.static.flickr.com/2700/...2f782345_o.jpg

[slash-5]

What technique did you use for D&B? The transition from light to dark on the cheeks looks a little orange and /or gray in parts. If you used the D&B tools, they are destructive and leave this effect. It's better to use a 50% Gray layer with a paint brush.

[PeterWright]

Originally Posted by slash-5

What technique did you use for D&B? The transition from light to dark on the cheeks looks a little orange and /or gray in parts. If you used the D&B tools, they are destructive and leave this effect. It's better to use a 50% Gray layer with a paint brush.

Thanks - I used twin curves layers with masks, so non destructive.. not sure what hascaused the colouration..

[secretagents]

I have worked 10 minutes on your image and here is the result. I don't think it's too good but perhaps it'll help you see what is not quite there in yours (depending on what you are aiming at of course).

[flexmanta]

Well, looks like you thinks D&B takes less time what what it really does.

Detailed D&B on skin is the first part of any retouching. Most people's workflow consist on getting a d&b'd, liquified, cleaned copy to be named something like "working version". From then on, you do your all the makeup and what not. Now, the d&b for that layer on such a closeup photo will take you no less than a couple days when working for yourself (working for a client is a whole different thing).

So. Not the best example for what i'm going to say: You might know everything about the technique already, so no need to study more. Just take more time, a lot more. Different zoom levels are necessary too. Some db will be done on a pixel level, some other is done on large areas. It's not a bad idea to separate frequencies first. DB works great for high freq detail, but for the low freq flaws (bigger spots and color imperfections), other techniques can be used.

[PeterWright]

Originally Posted by secretagents

I have worked 10 minutes on your image and here is the result. I don't think it's too good but perhaps it'll help you see what is not quite there in yours (depending on what you are aiming at of course).

Thanks very much - it looks smoother than mine..

Would you mind telling me what you did ?

[PeterWright]

Originally Posted by flexmanta

Well, looks like you thinks D&B takes less time what what it really does.

Detailed D&B on skin is the first part of any retouching. Most people's workflow consist on getting a d&b'd, liquified, cleaned copy to be named something like "working version". From then on, you do your all the makeup and what not. Now, the d&b for that layer on such a closeup photo will take you no less than a couple days when working for yourself (working for a client is a whole different thing).

So. Not the best example for what i'm going to say: You might know everything about the technique already, so no need to study more. Just take more time, a lot more. Different zoom levels are necessary too. Some db will be done on a pixel level, some other is done on large areas. It's not a bad idea to separate frequencies first. DB works great for high freq detail, but for the low freq flaws (bigger spots and color imperfections), other techniques can be used.

You are killing me - I spent about 1.5 days on the skin D&B.. my wife went nuts...

I think that I just need more practice because I got to the point where I just could not see where to work and started to make the image worse..

In your opinion does the image need to be smoother ? I was very conscious of not losing texture with this one.

[mistermonday]

Peter. inputs below on marked up image.
Regards, Murray

[PeterWright]

Originally Posted by mistermonday

Peter. inputs below on marked up image.
Regards, Murray

Wow - thanks very much indeed..

Food for thought..

[flexmanta]

Ok, ill try to make examples of where you need to d&b. As you say it does get hard to know where a stroke is needed. If you spent that much time on that face, you must be doing something wrong and have probably dodged and burnt parts that didnt even need it. Your eye will get used to see what needs work. Untrained people can only see "before afters", but can't point out what was done where.

For that reason, it's not a bad idea to help your untrained eye by temporarily boosting the flaws so you can spot them more easily. Some people create a desaturation adjustment layer as well as a very steep s-curve set to luminosity to boost the contrast of the flaws to be evened out.

Apart from that, know that you will need to work on 2 main levels. Pore level (pixels), and blemish level (large chunks). Here are 2 examples that illustrate what you see initialy, what your retouchers eye needs to be trained to see (s-curved), and how it needs to end up looking. These examples were done in 5 minutes or so, so the actual d&b is not of the best quality. I hope it helps you.

[PeterWright]

Originally Posted by flexmanta

Ok, ill try to make examples of where you need to d&b. As you say it does get hard to know where a stroke is needed. If you spent that much time on that face, you must be doing something wrong and have probably dodged and burnt parts that didnt even need it. Your eye will get used to see what needs work. Untrained people can only see "before afters", but can't point out what was done where.

For that reason, it's not a bad idea to help your untrained eye by temporarily boosting the flaws so you can spot them more easily. Some people create a desaturation adjustment layer as well as a very steep s-curve set to luminosity to boost the contrast of the flaws to be evened out.

Apart from that, know that you will need to work on 2 main levels. Pore level (pixels), and blemish level (large chunks). Here are 2 examples that illustrate what you see initialy, what your retouchers eye needs to be trained to see (s-curved), and how it needs to end up looking. These examples were done in 5 minutes or so, so the actual d&b is not of the best quality. I hope it helps you.

Thanks Flex - makes sense... work to do..

[PeterWright]

After the various comments and help - thanks very much..

Here is another version - does this look better ?

http://farm5.static.flickr.com/4024/...7d34bac6_o.jpg

[flexmanta]

Do you mind i use a part of the copy you just uploaded to touch it a bit?

[secretagents]

Originally Posted by PeterWright

Thanks very much - it looks smoother than mine..

Would you mind telling me what you did ?

Well smoothing on a specific frequency range as you would guess but I don't use the cumbersome high pass layer technique, I use -big scoop- the Corel KPT Equalizer plugin for that purpose, very flexible and a real time saver.

[PeterWright]

Originally Posted by flexmanta

Do you mind i use a part of the copy you just uploaded to touch it a bit?

Not at all - that would be great..

[flexmanta]

Ok, i have done some stuff. Please know that the examples i made here are only for ilustration purposes and in no way should the quality level taken as a final result ready to deliver. I also apologise for uploading so much stuff. I have separated frequencies to show you where the main problem with this photo is. She is a good Caucasian with pink skin... pink skin is actually not pink, its more like color-less. That is why, in the thinner areas, more blood is visible underneath, and you got a red spot. That is the problem with caucasians, blood. The problem with other types of skin might be something like pores, wrinkles... not with most of the caucasians. So, by separating frequencies you get to fix the redder spots in an isolated way, without messing with the perfectly fine texture.

1st: An s-curved version of your low frequency so you can see how bumpy it was.
2nd: An evened out low frequency makes for a big improvement already.
3rd: An s-curved high frequency soloed.
4th: What your d&b layer should start looking like.
5th: Not nearly finished D&B. Carved it a little... one of the disadvantages of fixing low frequencies is that you lose volume. You get the idea.

Again, sorry for uploading so much stuff.
Another piece of advice. Go to model mayhem and use some of the photos that they have posted there. Some are big uncompressed RAW files straight out of the camera back. All this time that you have used this photo for learning you could have spent with an amazingly detailed photograph.

[PeterWright]

Originally Posted by flexmanta

Ok, i have done some stuff. Please know that the examples i made here are only for ilustration purposes and in no way should the quality level taken as a final result ready to deliver. I also apologise for uploading so much stuff. I have separated frequencies to show you where the main problem with this photo is. She is a good Caucasian with pink skin... pink skin is actually not pink, its more like color-less. That is why, in the thinner areas, more blood is visible underneath, and you got a red spot. That is the problem with caucasians, blood. The problem with other types of skin might be something like pores, wrinkles... not with most of the caucasians. So, by separating frequencies you get to fix the redder spots in an isolated way, without messing with the perfectly fine texture.

1st: An s-curved version of your low frequency so you can see how bumpy it was.
2nd: An evened out low frequency makes for a big improvement already.
3rd: An s-curved high frequency soloed.
4th: What your d&b layer should start looking like.
5th: Not nearly finished D&B. Carved it a little... one of the disadvantages of fixing low frequencies is that you lose volume. You get the idea.

Again, sorry for uploading so much stuff.
Another piece of advice. Go to model mayhem and use some of the photos that they have posted there. Some are big uncompressed RAW files straight out of the camera back. All this time that you have used this photo for learning you could have spent with an amazingly detailed photograph.

Thanks a bunch - now some stuff is beginning to click..

May I ask a question or two;

1. Low frequency - all good here and I get this part
2. High freqency - my "apply image" layer does not look grey (Linear light blend) - do you work on yours for D&B in the grey mode or in the blended mode ?
3 D&B layers - mine also do not look grey but are regular curves layers with black mask - is this correct ? How can I get mine to look like yours so I can compare ?

Once again - thanks so much for your help...

[flexmanta]

2. Your apply image doesn't look grey?? It should. Linear light is a contrast mode, 50% grey has no effect. Review your apply image parameters and work in 16bits. Apply image according to 16bit. (Add, invert, scale 2, offset 0, linear light 100%). I work on my d&b layer viewing the whole image, not just the high freq, i still haven't got bionic eyes

3. There are many ways to dodge and burn an image. Working with curves is ok and needed when you work at 8 bits, but the increased amount of grey values available in 16bits makes it enough so that you can use half of the values for burning and half for dodging when using a 50% grey layer set to soft light and paint white or black on it. Most people use overlay, but i prefer soft light as it doesn't shift color so much. Also, i prefer a single layer so that i can switch from dodging to burning just by pressing the x key to switch from foreground to background color.

[PeterWright]

Originally Posted by flexmanta

2. Your apply image doesn't look grey?? It should. Linear light is a contrast mode, 50% grey has no effect. Review your apply image parameters and work in 16bits. Apply image according to 16bit. (Add, invert, scale 2, offset 0, linear light 100%). I work on my d&b layer viewing the whole image, not just the high freq, i still haven't got bionic eyes

It looks grey with these paramters until I select linear light - then it turns to looking exactly as the original image, which I thought was correct ?

[PeterWright]

Originally Posted by flexmanta

2. Your apply image doesn't look grey?? It should. Linear light is a contrast mode, 50% grey has no effect. Review your apply image parameters and work in 16bits. Apply image according to 16bit. (Add, invert, scale 2, offset 0, linear light 100%). I work on my d&b layer viewing the whole image, not just the high freq, i still haven't got bionic eyes

3. There are many ways to dodge and burn an image. Working with curves is ok and needed when you work at 8 bits, but the increased amount of grey values available in 16bits makes it enough so that you can use half of the values for burning and half for dodging when using a 50% grey layer set to soft light and paint white or black on it. Most people use overlay, but i prefer soft light as it doesn't shift color so much. Also, i prefer a single layer so that i can switch from dodging to burning just by pressing the x key to switch from foreground to background color.

I am also struggling now with where to remove my blemishes - I have been trying to add a healing layer over the low frequency and below the apply image layer - but any clone/healing work looks horrible and very strange results. I understood that blemsishes should be dealt with on the low freq layer but this does not seem to work ?

[amica999]

I do the healing on the apply image layer and set the option to "Current layer" not current and below. Should work

[flexmanta]

Ok, the low freq layer can be worked with in many ways. Since you are working mainly with color differences (where there shouldn't be any), try median. I don't like to use it too much and i prefer D&B, BUT, it doesn't look too bad when used responsibly, and it allows you to avoid correcting the color shifts that happen specifically when dodging and burning large areas of color (in RGB, most colors change hue and sat with d&b).

When working with median on the low freq, duplicate it. Then select the lasso tool, and DO NOT MEDIAN THE WHOLE FACE AT ONCE. Now, with the lasso tool and a small feather for the selection (8px or so), circle around a blemish area. Then apply a veeery slight median. Continue lasso'ing and median'ing, and since the last filter u used was median, you can just press control F skip the median dialog. Lasso, ctrl-f, lasso, ctrl-f, and so on. On this particular example i have taken it too far with the median only so that you can see. In fact, the best plastic skin look can be achieved by properly medianing the low freq and b&b the high freq for flawless texture.

THE IMPORTANT THING IS THAT, SKIN SHOULD BE RETOUCHED DIFFERENTLY FOR EACH FREQ. Color shifts on the low, texture on the high.

[PeterWright]

Originally Posted by flexmanta

Ok, the low freq layer can be worked with in many ways. Since you are working mainly with color differences (where there shouldn't be any), try median. I don't like to use it too much and i prefer D&B, BUT, it doesn't look too bad when used responsibly, and it allows you to avoid correcting the color shifts that happen specifically when dodging and burning large areas of color (in RGB, most colors change hue and sat with d&b).

When working with median on the low freq, duplicate it. Then select the lasso tool, and DO NOT MEDIAN THE WHOLE FACE AT ONCE. Now, with the lasso tool and a small feather for the selection (8px or so), circle around a blemish area. Then apply a veeery slight median. Continue lasso'ing and median'ing, and since the last filter u used was median, you can just press control F skip the median dialog. Lasso, ctrl-f, lasso, ctrl-f, and so on. Example:

Ok, cool - but can I also D & B the LF layer first, and then when done and it looks like yours (:-) I add another D&B layer for the HF and work at pore level and on spots etc ?

[flexmanta]

Oh yeah yeah, of course. In fact you can create a d&b layer and clip it to the low freq, and then another d&b layer and clip it to the high freq. But again, if you are having trouble finding the spots that need d&b, create a steep s-curve on top of everything and set it to luminosity so that you boost the flaws making them easier to find.

Remember to zoom in and out a lot, and to constantly check before and afters. Ctrl-z is a very used command when d&b large areas, just to check what you just changed.

Before you continue, i would recommend you to start from scratch with a better quality image. Im sending u an ok one. There are also available RAW files here and in Model Mayhem. Try to use face photos for learning this stuff.

http://img688.imageshack.us/img688/7...141412289l.jpg

[PeterWright]

Originally Posted by flexmanta

Oh yeah yeah, of course. In fact you can create a d&b layer and clip it to the low freq, and then another d&b layer and clip it to the high freq. But again, if you are having trouble finding the spots that need d&b, create a steep s-curve on top of everything and set it to luminosity so that you boost the flaws making them easier to find.

Remember to zoom in and out a lot, and to constantly check before and afters. Ctrl-z is a very used command when d&b large areas, just to check what you just changed.

Before you continue, i would recommend you to start from scratch with a better quality image. Im sending u an ok one. There are also available RAW files here and in Model Mayhem. Try to use face photos for learning this stuff.

http://img688.imageshack.us/img688/7...141412289l.jpg

Cool - got it..thanks..

Out of interest, why is my image not good enough - it was well lit and a D700 Raw file so surely enough res etc ?

Re my Apply layer - am I correct that it will only be grey looking until I apply the linear light mode, at which point it turns "regular" looking ?

[flexmanta]

Oh, my bad. I didnt know you had the raw file.

Yeah, when frequency separated, low freq + high freq set to linear light should look EXACTLY the same as the unseparated version. What might have confused you is that I soloed and s-cruved the high freq before so you could see that your problem wasn't in the texture, it was in the low freq.

[PeterWright]

Originally Posted by flexmanta

Oh, my bad. I didnt know you had the raw file.

Yeah, when frequency separated, low freq + high freq set to linear light should look EXACTLY the same as the unseparated version. What might have confused you is that I soloed and s-cruved the high freq before so you could see that your problem wasn't in the texture, it was in the low freq.

Got it - now for some work on another new image..

You are patient man.. thanks..

[flexmanta]

Originally Posted by PeterWright

You are patient man.. thanks..

Teaching is the best way of learning.

[PeterWright]

Originally Posted by flexmanta

Teaching is the best way of learning.

Sure - but it takes time and effort, which I really appreciate.

I am enjoying the journey - a bit like learning to light..

[PeterWright]

Ok - I am a sucker for punishment but here is another one from scratch..

Any improvement ..

http://farm3.static.flickr.com/2692/...15b2c700_o.jpg