Super-clean beauty retouch

[mikedimples]

That's great advice Tommy0, I'm going to create a new thread dedicated to frequency experimentation. I'll gather the resources and prepare some comparisons (original, dodge and burn, and frequency healing). Thanks again for your wisdom and guidance.

[Markzebra]

Ok despite your rudeness - …

So your saying that separating the low frequency and high frequency part of the image will not work for many technical reasons. Mind sharing three?

1 texture is something that you should be looking to preserve, not destroy - the badly entitled 'frequency' method that you linked to is actually precisely the same as a high pass technique. Its a simple High Pass, lower contrast - not more precise, not that it matters.

2 Localized tonal variation (texture eliminated or editied by methods such as High Pass, and blur) can not replace proper retouching. Because, that's about judgement, deciding this area of texture needs lifting, this pore needs sinking, this needs cloning. SOME texture needs nothing at all. I suppose its very vaguely possible that in the future some kind of analysis technology (not pure filtering) in the future might take us some of the way there - this would require something pretty different from whats available at the moment - I'm only saying this because I've learned in the past that you never know what on earth they will come up with, and wouldn't put anything past them.

3 A FILTER is a unmodulated mathematical process applied across an area- direct filtering on skin will always look like filtering on skin. Sure there are better ways to reduce texture than this gobbledegook, (if you want to) but nothing replaces hard work. If your going to use filtration, then might as well go the whole way, and blur and re-texturize - it would do a better job

The entire point of what I've been doing is to unify skin texture. I don't understand.

Well why are you blurring it then?- ironing out texture. By the word 'unifying' I meant making a rhythm.

How about saying what you find objectionable about it.

Its blurred, quite obviously visibly filtered, and you suggested in your intitial post - "MANY dodge and burn curves layers", and obviously the image doesn't reflect that -

Ok this time only …

[markc78]

yesterday I attended the webinar by Gry Garness which was overall very intersting not least because she used a refined degrunge technique as a final finishing effect to ehance the skin with a sheen , a tightness & reduction of skin tone blotchiness.
Personally I think blurring techniques when used sensibly can have an important place in high-end retouching & sometimes feel that there's an unfair snobbery against their use.
Let's face it does anyone really enjoy total D&B ?

[markc78]

Thankfully I don't live in a B&W world

[Godmother]

Originally Posted by mikedimples

So your saying that separating the low frequency and high frequency part of the image will not work for many technical reason. Mind sharing three?

Originally Posted by mikedimples

I'm starting to think you use frequency separation and are trying to prevent it from spreading. Stop acting so suspicious.

It doesn't work like that, he is ACTUALLY trying to help you, open your eyes.

1- It doesn't work because u can't use JUST ONE RANGE - I use high pass for several not so high end projects BUT never one radius in the whole face, that WILL NEVER WORK.

2- What he says, about making a decision with every "pore" it works exactly like that... maybe not EVERY SINGLE PORE but with every AREA of an image u make a decision when retouching... "this shadow will be darker to enhance the depth" or "I'll lighten this shadow up to make the transition between shadow-highlight smoother" --- This stray of hair needs blending in, I'l clone it in darken mode - I'll paste some other bit of hair over it - I'll make it darker and alter the tone to make it fit the rest of the hair - etc -etc
With a filter u don't make ANY CHOICES.

3- It's really about how u want your image to look... Flawlessly natural or "default glamour look"

He is not attacking u personally... but sometimes... when u read the same things over and over and see different ppl making the same mistakes... asking about the mistake, getting the "don't use blur" answer and complaining about the critics or harsh answers (Not u, this happens EVERY TIME) it gets a bit annoying and Mark is not a very patient guy

Have a nice day and i'f u're going to keep using that technique, then try this

DON'T do just 2

Do several in layer groups, masked out over the original image.

Work in each group with the level of texture u want to achieve in that particular area (but EVERY TIME, walk away from the computer or ZOOM OUT to see howthis affects the whole image)

Mask every one in where u want it applied.

hope that helps.

xxx

[snook305]

OR Like Mark mentioned you could use the Byro technique which works pretty well as Gry Garness showed in her last Retouch Live video tutorial..
Work very nice in keeping the details..
Worthe a try.
Snook

[Flashtones]

Originally Posted by Markzebra

Here's a quick attempt to show clearly the difference between manual technique and filtered, using the before and after image Mike posted earlier. On the right is the posted image, on the left the orginal

Mark. "Let's face it does anyone really enjoy total D&B ?' - if you don't enjoy it, don't do it. Up for a limited time…

That picture was worth more than 1000 words. Thank you.

[TommyO]

Mark,
That is an excellent image, a very good example of what the human touch adds to retouch, as well as demonstrating the advantage of pixel-by-pixel/pore-by-pore d&b.

Mike,
That is your target. If you can get close using a faster method, then you are on to something.

[Quantum3Studio]

Originally Posted by mikedimples

What do you mean by the right side? Did you mean the picture on the left (the 'before'), or did you mean the right side of her face?

The right side of the image, left side of her face.

[mikedimples]

Apparently I've been unclear on several points. Please allow me the chance to clarify. I don't apply this affect over the entire image, I've always painted the (low frequency) blur on the areas that need it. While I was researching it, I found this great post that explains it clearly.
http://www.modelmayhem.com/po.php?th...=5#post9709595

Godmother brings up some GREAT tips. I need to run several different frequency's depending on the needs of the area -- I'm excited to try this new method.

I never intended this technique to be a retouch-machine, but as an alternative step in the retouch process that creates attractive skin with realistic skin texture without the hours required by dodge and burn, nothing more.

I had every right to defend myself against the false information provided by Markzebra. I get irritated when people pretend to know about a subject they don't, and it makes me wonder about his intent (just because what he said was so BLATANTLY false). If I don't know about a particular technique or theory, I don't pretend that I do and spread misinformation to my peers. Sorry I didn't "open my eyes" to his "help."

1 texture is something that you should be looking to preserve, not destroy - the badly entitled 'frequency' method that you linked to is actually precisely the same as a high pass technique. Its a simple High Pass, lower contrast - not more precise, not that it matters. Separating it into 'high' and 'low' has no additional value.

This is completely false. You must have not read very far in this link - it's full of proof: http://www.modelmayhem.com/po.php?thread_id=439098

2 Localized tonal variation (texture eliminated up my methods such as High Pass, and blur) can not replace proper retouching. Because, that's about judgement, deciding this area of texture needs lifting, this pore needs sinking, this needs cloning. SOME texture needs nothing at all.

My fault for not explaining that I mask in the blur (just replaced this step with a selective bandstop filter) to the low frequency layer, and then clone what I need to on the high frequency layer. Sinking and raising comes with painting grey on the high frequency layer. It's not some kind of one-click effect.

A FILTER is a unmodulated mathematical process applied across an area- filtering on skin will always look like filtering on skin. Sure there are better ways to reduce texture than this gobbledegook, (if you want to) but nothing replaces hard work. If your going to use filtration, then might as well go the whole way, and blur and re-texturize - it would do a better job

You may think this technique is gobbledegook, but other high-end retouchers and photographers disagree with you. For example: Robert Randell. http://www.robert-randall.com/

Nice job on the dodge and burn -- I'll aim to recreate that. How long did it take?

[Markzebra]

without the hours required by dodge and burn,

The image in the center took me 45 minutes work. The image on the left is the original, center is my quick retouch, right is your blur. Apologies the image lost its color profile, on uploading and looks a little washed out.

This is completely false. You must have not read very far in this link - it's full of proof: http://www.modelmayhem.com/po.php?thread_id=439098

I did read that. Sigh - in detail: He's used Linear light BECAUSE the particular Apply image in 16 bit, that he has used, is only delivering a low contrast High Pass. Take his technique (run it through in the Action) - and compare it with an 8 bit high pass (same radius) at about 50%. This means make a solid gray layer underneath, and take the high pass down to 50% - its precisely the same result. The Apply Image is doing nothing original, in fact its doing a worse job than a straight High Pass. NOW there is one genuine advantage revealed by all this, used by many smarter retouchers already - that of using high pass techniques (if you need them for texture and sharpening) in 16 BIT!

Taking areas producing different filtrations based on tone (actioning them) and masking them, what Natalia is talking about, is the best achievable way to get these techniques work. In the end it will still look filtered though, and I'm sure she will agree with me.

You may think this technique is gobbledegook, but other high-end retouchers and photographers disagree with you. For example…

Ahem, I better make no comment there specifically. I think that there retouchers who do a specific type of work, forced to work to deadlines, I've recently have been asked to consider this myself actually. It all depends on your client what he (or she!!!) is willing to pay. I never said otherwise to this. The most successful of these FAKE techniques is actually a airbrush(blur) and re-texturise approach, this says "screw real skin texture and light completely".

Just don't ever use that crap in your folio. There are retouchers out there who's PDF's I've seen myself - you just immediately recognize - yep plastic skin. The experienced eye can immediately see the difference, never underestimate this.

[Godmother]

Originally Posted by Markzebra

In the end it will still look filtered though, and I'm sure she will agree with me.

Of course

I want to do a video tutorial (proper one) and I felt forced to include a "FAST" way to get things done - this is copy pasted from the "theme" notes I have:

High Pass Thing: The quick and dirty way of D&B
*How to
*When to use it
*When not to use it
*What does it do
*IMPORTANT: Not abusing the High pass thing
*How to chose %

Sadly - because I believe every image should be perfect - The client doesn't always have the budget to pay for "Perfection" or the time to achieve it.

I did Dove last week - got 6 images monday evening to be delivered Thurs morning. I explained to the client, I need a working day per image at least - she explained it to the client... the client said: Do the best u can we need them for a meeting. So u do the math.

x

[mikedimples]

I'm sorry I was rude to you Mark. I can see you know quite a bit about retouching, and most of this is my own fault for not explaining my process properly.

I generally do my work in 16 bit, so I didn't even consider the 8 bit angle, although the genius' on that site have already developed a workaround technique for 8 bit images as well as lab. Here's an example of a 30 minute skin-clearing session I did today that would have taken hours and hours with dodge and burn, although dodge and burn would give much more precise results. It's family, so there was no need for high-end d&b. Mind taking a look at the full version and telling me what's lacking? I really do value your opinion. Everything in this retouch was done with frequency separation healing, plus I did high frequency sharpening before I uploaded it. Now I have plenty of time for the fun stuff -- dodge and burn carving.

http://a4.vox.com/6a0110184cd071860f011016333a94860b-pi

I found out one of the major problems I'd been having. The frequency separation is done by using 'add mode' apply image (invert checked) on a gaussian blur of the original. This subtracts the blur, amazingly leaving the details. Instead of using a high value for the gaussian blur, I assumed a lower value would preserve more detail in the high frequency layer, which is the opposite of the truth.

The more I work with this technique, the more I like it. It feels vastly more precise from high pass detail preserving -- you guys should try it! Here's the action to separate the high and low frequencies. Use a gaussian blur between 5 and 10 for normal (for us) size images. Try cloning on the high layer, and sampling colors on normal brush blending mode on the low frequency layer to even out tone. Make sure to use it on a 16 bit image.

http://www.nunuvyer.biz/Photoshop/Frequency.zip


EDIT: Actually, I don't think there is a workaround for 8bit yet.

[mikedimples]

Oops, cloned out some important moles. I'll put those back in.

[Flashtones]

Mike, I don't have a horse in this race, I'm learning from both of you, but for comparison purposes I think it would help if you'd demonstrate what the best of your technique can do on the original image again so we can compare with Mark's samples. Apples to apples, etc.

[mikedimples]

Ah, gotcha. I'm about to leave for the day, but I'll put it up later tonight. Thanks to everyone still keeping an eye on this thread.

[Markzebra]

http://a4.vox.com/6a0110184cd071860f011016333a94860b-pi

AS THOSE IMAGES STAND AT THE MOMENT - it looks like you were trying to catch me out.

A lot of that image looks like relatively crude L&D. There are areas that are untouched, and have the original texture sharpened slightly. The highest res I can 2000 pix high, and that's not high enough to say for certain, there is no real pore detail in the original image. It has had detail work, someone has gone in, and done some detailed cloning or healing. Dropping back of tone is not consistent, and obviously brushed, suggesting dodge and burn.

Having said all that - Yep to me, even at that res, it still looks slightly filtered in some way sorry to say. Maybe its just the way you have approached the L&D.

You would have to work a higher res image to prove your point.

As for the rest of what you have written, I NOW think its best to ignore that. You haven't tackled any of the points I made.

[mikedimples]

You're right. I'll properly respond to your points later tonight.

[mikedimples]

Sorry it took me much longer than one night to reply. Better late than never!

I've been working quite a bit with frequency separation since my last post, and I think I'll be able to do a better job of explaining it this time. First let me respond to a post by Mark I neglected.

Originally Posted by Markzebra

Take his technique (run it through in the Action) - and compare it with an 8 bit high pass (same radius) at about 50%. This means make a solid gray layer underneath, and take the high pass down to 50% - its precisely the same result. The Apply Image is doing nothing original, in fact its doing a worse job than a straight High Pass. NOW there is one genuine advantage revealed by all this, used by many smarter retouchers already - that of using high pass techniques (if you need them for texture and sharpening) in 16 BIT!

I still don't think you fully understand what was said in the thread. Here's proof that's it's more accurate than high pass(7th post in the thread). "The image below is the difference between a HP + GB solution for frequency separation and the original image. Processing was done in 16bit mode with a 5px radius when the image was 800x527." -Sean Baker

As long as you convert the image to 16bit first and split the frequencies, your end results will be more accurate than high pass.


In the next post I'll try to clearly explain what frequency separation does because it's NOT a filter of any kind. I'm about to go do that lip retouch now that I have more experience with frequency separation.

Originally Posted by Markzebra

AS THOSE IMAGES STAND AT THE MOMENT - it looks like you were trying to catch me out.

Sorry, I'm not familiar with that expression. I even tried googling it!

Originally Posted by Markzebra

A lot of that image looks like relatively crude L&D. There are areas that are untouched, and have the original texture sharpened slightly. The highest res I can 2000 pix high, and that's not high enough to say for certain, there is no real pore detail in the original image. It has had detail work, someone has gone in, and done some detailed cloning or healing. Dropping back of tone is not consistent, and obviously brushed, suggesting dodge and burn.

I'm also not familiar with L&D, but I have a feeling I'm going to feel stupid once I know what it stands for. I went into that image, separated the frequencies (so far image is unchanged), and did quick, extremely sloppy cloning on the high frequency layer. The beauty of working only on the HF layers is that cloning doesn't pick up tone. It's extremely forgiving, but looks precise.

Any tonal changes were painted in on the low frequency layer. I just used lighten or darken mode with the standard brush, color pick what I want, and paint away. As long as the frequencies have been separated properly, the painting is done under the texture, looking very convincing.

I'm about to redo that lip retouch now that I've got a better understanding of the technique.

[Quantum3]

Originally Posted by mikedimples

What do you mean by the right side? Did you mean the picture on the left (the 'before'), or did you mean the right side of her face?

Meaning you've chosen the left eye for the symmetry. I would chose the right eye because it looks with more character.

[Quantum3]

Originally Posted by mikedimples

I've done the skin over again, mind telling me what's wrong with it?

I'm still have to tackle the hair.

LARGE SIZE: http://a3.vox.com/6a0110184cd071860f011018509c8b860f-pi

Much better as I'm only looking the skin. However, you have some dead/flat texture in some areas in the left of the face. There are some kind of textureless "patches" on her cheekbone and nose.

[mikedimples]

Thanks for the critique, and nice catch on the eye symmetry. I should have definitely gone with the other one.

[Quantum3]

I'm looking at the Robert Randal website. It has very good pics, but somehow the pic #7 has a very harsh blown highlight in the knee of the ciclyst. I'm totally wondering why. Not to mention other parts of the body and the flat red in the down part of the t-shirt. Too much obviusly overprocessed in those areas. But why!?!?!?!?

These things I always try to avoid, and even more when doing grayscaled images. Sometimes I push shadows and highlights too much out of the boundaires, but never so much like this (of course, using direct point flash can produce these things, but it looks like more due the contrast in the post):

[Markzebra]

You image hasn't come through Mike. If its supposed to be completely bkack, then good job. Ok Straigh talk - reading through this again I realise I was a bit of an asshole.

As long as you convert the image to 16bit first and split the frequencies, your end results will be more accurate than high pass.

Of course anything you do will be more accurate in 16 bit, do the high pass in 16 bit, that was my point. Looking at this again, the Apply image DOES not seem to create the blown highlights that high pass does I will agree, but in 8 bit High pass may be better I think because it creates a higher contrast result, and as a result is more flexible.

In the next post I'll try to clearly explain what frequency separation does because it's NOT a filter of any kind.

Now ahem, Mike Mike. - I can assure you I know what separation of textural areas does, and how to use two or more layers to divide what people are now seem to be calling 'frequencies', although I don't like the word. This is what I was calling "high Pass techniques". These things ARE NOT a replacement for learning from the ground up, believe me. And I'm somewhat disappointed that someone is posting this stuff on a forum at all. The main reason I say this is because, like I said in another thread, its like giving a four year old a razor blade - open to abuse. There is a school of retouching that does use these methods very extensively, but its so easy to produce very artificial results. Craig.

so far image is unchanged

Look at the image Histogram - any filtering process does some kind of damage. The secret to this is analysing differences and using them as repairs only.

L&D means dodge and burn. Using your image as an example the problem with painting on what you are calling low frequency is its - NOT AS GOOD as dodge and burn really for pure tonal work. High 'frequency' that's another matter.

[Godmother]

Originally Posted by Markzebra

And I'm somewhat disappointed that someone is posting this stuff on a forum at all.

The thing is Mark, in that particular forum, they DON'T BELIEVE high end is actually done using D&B for more than "Painting highlights"


Let me show u so u can understand.

Originally Posted by Andrew Thomas Designs

Dodge and Burn is really only for highlights and shadows, as Stephen said.
To get texture it's a lot of clone stamping, rubber stamp, etc. Then some of us use airbrushing, as well as applying hipass filters and stuff to get hte texture back.
Andrew Evans
www.mplsFASHION.com

Originally Posted by NYPHOTOGRAPHICS

can you name an high end ad level retoucher who uses that? I cannot and i know a lot of them.
Stephen Eastwood
http://www.StephenEastwood.com

Some of the answers in THAT forum

They really don't see the difference between manual work and "Frequency separation" (cof high pass cof)

And I don't think they ever will

Mike on the other hand is experimenting with this and he actually sees the difference (He states that in the other forum)

Have a nice day

x

[Markzebra]

Ok don't worry I'll shut up then Godmother.

[Cuervo79]

Hmm I come late into the game. but the MM thread originally was supposed to be a technique to sharpen with less noise than using highpass. I did think of using this instead of the "highpass technique" for skin. I guess the thread evolved into a skin smoothing technique than what it originally was.

There are people in MM like godmother said, who know that there's no substitute for D&B, Ronald N Tan comes to mind but most of the people there when talking about retouching are looking for the quick fix but still want realistic results. There haven't been threads lately on the subject in the MM forum but each time one comes out it becomes a who's who of blured vs realistic...

[Visna]

Hi all!

It's good debate!!! Any shortcut to bypass D&B? Yes it's possible!!! But's it's still a D&B technic after all. I busted Portrature plugin and I could achieve that with simply Photoshop. I have the same level of control and no need to spend money.

But, the goal it's to have more precise control to speed up the workflow. D&B it's one of part of the tool to retouch photo, not the only one solution for all purpose. First, it's to develop your eyes about the Highend trend and how to balance the retouch.

I made a script on Photoshop that's give me a control on many layers of the skin for shadow and highlight separatly to gain time. But, skin it's the not whole image. A lot a job to do on other side.

And for your Frequency... No matter 8 bit or 16 bit... You could achieve simply by mixt the same image with one is inverted. I use smart object to have playback control on the effect.

[mikedimples]

Welcome to Visna, I had a couple of questions about your post if you don't mind.

Originally Posted by Visna

Any shortcut to bypass D&B? Yes it's possible!!! But's it's still a D&B technic after all.

---
What do you mean by this? This is not a dodge and burn technique.

And for your Frequency... No matter 8 bit or 16 bit... You could achieve simply by mixt the same image with one is inverted. I use smart object to have playback control on the effect.

---
Wouldn't mixing an inverted image with itself leave you with grey? I don't understand this technique, mind elaborating?

[mikedimples]

Originally Posted by Cuervo79

There haven't been threads lately on the subject in the MM forum but each time one comes out it becomes a who's who of blured vs realistic...

---
Are you saying that whenever the topic of frequency separation comes up in the Model Mayhem forums, it becomes a blurred vs. realistic war? I thought I'd read the two topics dedicated to frequency separation, mind linking the other ones? Even if it's a flame war, there might be stuff I haven't read about it yet.

Here's the newest low frequency topic on Model Mayhem: http://www.modelmayhem.com/po.php?thread_id=487490

Here's the original one:
http://www.modelmayhem.com/po.php?thread_id=439098

Originally Posted by Markzebra

You image hasn't come through Mike. If its supposed to be completely bkack, then good job.

---
Wrong.

Originally Posted by Markzebra

but in 8 bit High pass may be better I think because it creates a higher contrast result, and as a result is more flexible.

---
The grey that comes as the neutral color in the high pass layer is limiting. With the separated high-frequency data, you can run a curves adjustment on only the fine detail in your image. The result of the contrast is decided from the sharpening you do to the fine detail (high frequency) layer. Remember, the sharpening process in this technique involves separating the fine detail, then smart-sharpening it separately from the rest! The blurry, low frequency data where the halos usually form are left untouched. I don't understand the logic of why you think high-pass is more flexible.

Originally Posted by Markzebra

Now ahem, Mike Mike. - I can assure you I know what separation of textural areas does, and how to use two or more layers to divide what people are now seem to be calling 'frequencies', although I don't like the word.

---
How do you use more than two layers to divide it? I really don't know and it sounds interesting.

People area calling it separating the frequencies because that's what it is. Here's how Sean Baker explained it, "The thread referenced deals with separating a 2D image into its component waveforms (or at least as well as we're able with the tools we have) and using that information for various purposes (emphasizing the high frequencies to sharpen, de-emphasizing intermediate frequencies for skin smoothing, etc.). The term 'spatial frequency' was taken out of writing from Dr. Livingstone at Harvard."

Here online encyclopedia article on it: http://www.encyclopedia.com/doc/1O13...ncyfilter.html

(More to come... I'm paranoid about accidentally hitting back and losing the whole thing.)