Variety strikes again...

[Zach Gray]

Look at the Milk Girls - their skin was great, even with our very earliest raw development software.

Graeme

When are we going to see Milk Girls - EpicHDRx?

 

[Steven Swigart]

Thanks for posting. This is a very well written post and argument.

Here is a google cache of the article from Variety: http://webcache.googleusercontent.c...ariety.com/article/VR1118029071?refCatId=1009

 

[David Mullen ASC]

From my experimentation, skin tone is more about lighting than camera. Look at the Milk Girls - their skin was great, even with our very earliest raw development software.

Well, to be fair, those girls had great skintones in front of the camera, pale & creamy.

It's always been tanned faces in warm light that has been a challenge for digital -- and even film going through a D.I. -- because skin tone has such a subtle mix of colors (even blue) that as soon as you go towards warmth, you run into problems with color space limitations wanting to make them go brown or yellow or magenta, i.e. simplifying the complex color into something the system can create or display. We all know instinctually what skin tone is supposed to look like, but spend hours in color-correction suites trying to "recreate" it in the digital realm using red, green, blue, yellow, cyan, and magenta adjustments.

Like I said, it happens even with film scanned and going through a D.I., any of us who have done this can remember the occasional shot where the skin tone just looks "wrong" and we've spent much time twiddling knobs trying to get something we can live with. It's hard to explain but you find yourself trying to essentially describe to a colorist what skin tone is, maybe it needs a hint of magenta but with the red chroma brought down, etc. and it all gets even harder in a scene where the light is supposed to have a color bias, like a hint of warmth or coldness. Now you are trying to recreate the color of skin but with a color bias on top of that. It can drive you up the wall talking to the colorist -- "I want the light to feel cold, but when you put some blue in the face, now it just looks purple, but now when you pull the chroma down, it just looks greyish... let's start all over again..."

3-strip Technicolor was designed around getting great skintones, but was helped by the fact that stars back in the 1930's and 40's wore a ton of make-up, so essentially faces were painted to look correct for the process. Nowadays, we have to work with all sorts of skintones in front of the camera, and some are tricky for digital systems to reproduce, such as olive skin.

The other problem is creative in essence, we have a lot more control over the colors of the image with digital color-correction, so creative choices come into play, in the old days of photochemical printing, you couldn't easily just pull up or down the chroma in a fleshtone. But now a director and DP can opt to make someone's skin look very monochromatic at times for dramatic/creative reasons, so it looks "wrong" but it's deliberate (that, or maybe the director or DP just have poor taste or inexperience in timing, I don't know.) It can be hard to argue with that, but it was less of an issue in the old days when the color response of film stocks, neg and print, was more or less standardized.

But I'm convinced that the color science behind skin tone reproduction is still in its infancy, the subject (and the color) is a lot more complex than people realize.

 

[David Mullen ASC]

I think resolution in capture and display may be a factor -- I recall seeing a 70mm print of "Far and Away" (shot in 65mm). Of course, it was pretty sharp, but because of the modern use of soft lighting and even some lens diffusion, it didn't look as crisp as some old 65mm movies such as "Lawrence of Arabia"... but the one thing that I did notice in the theater was that I was seeing subtle shades of color that I didn't recall seeing in 35mm movies shot on the same stock, shades of pink inside of an orange color, etc. and richer skintones.

I formed a theory that when you have more grains devoted to resolving detail, you were also resolving small amounts of color detail that were blurred essentially on smaller film formats, so a tiny area of the frame that was a blend of specks of color were being more accurately reproduced with the larger negative, so a brown patch of dirt, for example, in 35mm, now in 65mm showed hints of other colors mixed in. A patch of skintone now on the larger negative could show hints of blue from blood vessels, reddish variations, blue undertones, etc.

So I believe it's the same with digital, you want to be able to resolve more color detail, and more detail overall, in order to capture extremely subtle and complex colors -- like in skintones. So a 1080P camera recording 4:2:2, for example, is going to tend to reduce skintone to a simpler shade than a 4K RAW camera, at least in theory. But even a Bayer pattern, a compression scheme, and definitely the conversion algorithms, are going to have an effect on how much of that subtle color detail is retained. Hence why I think it's a complicated subject that still needs exploring.

 

[Graeme Nattress]

I now see things in a very different light than I did before I got involved in digital cinema. One of the things with skin tone, is that different people have different views, and what one thinks looks great, another will possibly not. Perhaps it's human perception that is the more complex aspect, than the scientific colorimetry aspect?

Another possible aspect is the complaints I read or hear when people know in advance it's digitally originated material. Or when I hear that after seeing some RED footage - those M-X skintones don't look as good as the M, when we were viewing M footage, or other contradictory examples.

I think you're really getting to the issue when we start talking about the creative choices.

Graeme

 

[David Mullen ASC]

It's definitely a mix of creative and technical issues, but it's also an issue of the chain of events from what you put in front of the camera all the way to how it's presented, and thus it can be hard to break it down to find out where something is starting to drift in the wrong direction. Was it the make-up? Did the HMI light have a color spike problem? Does the lens have a color tint? Was there a stray color coming from some bounce source? Was the transcoding, processing done correctly? How was it viewed during correction? Was the master made correctly? The copies? Is the display device calibrated correctly?

But beyond all of that, I do think there is an overall "issue" with skintone reproduction in digital systems, cameras, recorders, and codecs seem to want to push certain pastel and complex/mixed colors in certain directions, which affect color-correction decisions pro and con later, etc. But I also think we are talking subtleties here for the most part, but subtleties are what skintones are all about. Despite all the creative issues involved, there has been a tendency for Red-shot movies of the past to have what could be called an Agfa-look, sort of the impression of brown and cyan, and I don't know why. It can't entirely be due to creative choices. I'm waiting for more Red-shot movies to work in a more Technicolor/Kodachrome palette, not the Autumn-Winter colors of so many Red movies -- maybe "Spider-Man 4" with that red and blue outfit...

 

[Graeme Nattress]

David, one of these days we'll have to sit down together when we have the time and play around with all this stuff. Your first paragraph is utterly spot on - the amount of variables at play is immense.

I know, or at least I think I know, what you're getting at with your 2nd para, there's a sort of homogeneity of colour that you get, indeed on more pastel tones where instead of being the subtle variety of colours you get in reality, you get everything converging on the dominant colour. But I don't see this as a particularly digital thing (other with highly compressed stuff, no matter what the source, like on a DVD) as I see a similar effect in film source material too.

Graeme

 

[Tim Whitcomb]

Thanks all. Nice to know that many out there do "get it"

Graeme, David, Mark and the many others that have chimed in here with additional info to make the points so clearly and elegantly about why we do what we do is very beneficial.

Wonder if anyone from Variety has noticed this group groundswell from all of you, and that they can't just put such a skewed story out there without being called out on it.

+ Ted

Hi Ted. I actually think this article is pretty much Hollywood Journalism 101. Find a hot topic about a product, actor, film, etc and then find someone to CONTRADICT that hotness.

Its a super LAZY but very easy way to get LOTS of attention drawn to yourself.. Its a Magazine, and crap like this sells... His directly ignoring all the positives, guarantees he had no interest in anything but telling a CONTRARIAN view.

 

[Ruairi Robinson]

David, one of these days we'll have to sit down together when we have the time and play around with all this stuff. Your first paragraph is utterly spot on - the amount of variables at play is immense.

I know, or at least I think I know, what you're getting at with your 2nd para, there's a sort of homogeneity of colour that you get, indeed on more pastel tones where instead of being the subtle variety of colours you get in reality, you get everything converging on the dominant colour. But I don't see this as a particularly digital thing (other with highly compressed stuff, no matter what the source, like on a DVD) as I see a similar effect in film source material too.

Graeme

Yes, when it goes through a 2k DI. Not when you see chemically processed images projected on film. I think in 20 years the "early 2k DI look" will be seen as a bad phase.

The discussion of lighting comes down to - yeah, great, you do awesome lighting and it will look good. But there are occasions where bad lighting or no lighting at all will look acceptable on film, but the same thing shot digitally will look ugly. This goes beyond just matters of taste.

Having said that, finally the latest color science seems to have taken a big step in the right direction for the first time. Even Social Network had some... purple lips. On one of the actors.

R.

 

[David Mullen ASC]

I wonder if grain is a small factor in the perception of color complexity -- we've talked before about how grain gives the impression of detail and resolution because it gives the eye something to focus on as being sharp even if the actual image is not sharp... I wonder if the mix of color dye clouds in a film image to some small degree helps fight against that feeling of monochromatic blandness that skin can fall into with some digital photos. Not that it would contribute to some of the issues I am talking about, but it may be a factor.

 

[Graeme Nattress]

Grain could be a factor - it certainly influences the perception of resolution.

Graeme

 

[Adam C Lubkin]

David, one of these days we'll have to sit down together when we have the time and play around with all this stuff.

Graeme

May I suggest Graeme, David, and Michael Cioni running a Reducation Color Grading Module? That way you can get paid to sit down and play with this stuff and we can all learn from it.

 

[Steve Sherrick]

David and Graeme, this is a great conversation. I think what you guys are saying makes a lot of sense, and David you have wonderfully articulated some of the challenges, especially the challenge of evaluating a final image without knowing the exact chain of events that went into making it. There are so many variables, and when I see people go off about RED looking terrible in a particular film they saw, I take it with a grain of salt because they are commenting on the final look and not on the source material. And even then the source material has so many variables that went into making it.

The thing I admire about Graeme is that he continues to tweak the color science, striving to improve it. I think it has. And I expect it to get even better as he makes more discoveries and gets continuous feedback from the field. The more data he has to wade through that shows images under various conditions, the better the science will get.

I'd like to see this conversation continue. It's fascinating stuff.

 

[Elsie N]

I'd like to see this conversation continue. It's fascinating stuff.

Agreed. There are two other hot threads going on (Offhollywood getting his camera and Ridley Scott and Dariusz Wolski) but each time I check for new posts I look for this thread first.

 

[M Most]

May I suggest Graeme, David, and Michael Cioni running a Reducation Color Grading Module? That way you can get paid to sit down and play with this stuff and we can all learn from it.

Uhhhh.... nothing against any of those three people, all of whom I consider friends, but you might actually want to have a colorist involved in that sort of thing.

 

[Steve Sherrick]

Mike, stay tuned.....Not out of the realm of possibility......

 

[David Mullen ASC]

Maybe a good test would be to shoot five faces of varying skintone color and quality, in daylight, tungsten, and Cool White fluorescent, on the M-X sensor and on 35mm film stock, at different exposures, and see if there is a skintone reproduction issue or not. If the "problem" can't be replicated, then maybe it doesn't exist...

 

[Sanjin Jukic]

My opinion about RED colors, skin tones,...

My opinion about RED colors, skin tones,...

DIGITAL ACQUISITION MAIN CHAIN: LENS + SENSOR + RAW DEVELOPER APP = RENDERING OF IMAGE COLORS.

If RED want to have appropriate balanced colors straight out-of-the-box from the new cameras EPIC, Scarlet, R1MX,...

that means the new RED Color Science colorimetry should follow direction of color matrix from Olympus digital photo cameras.

To illustrate exactly what I'm talking about just follow the two quotes below.

The comparison Nikon vs. Canon vs. Olympus:

Portraits & Skin Tones>>>

Color Accuracy>>>

When colors are balanced from the beginning in a right way and direction than it's easy

for any colorist (professional or even amateur) to get out of image what ever they want.

An ideal situation is to get appropriate balanced colors straight out-of-the-box

that you don't need to tweak so much in post or even don't need CC at all.

Then you can get the most pleasing option of pale, natural or vibrant skin tones.

You could adjust that in camera if you have appropriate calibrated monitor or evf.

Also you can saturate or de-saturate you footage look in any direction you want before or after recorded image.

WYSIWYG could still rule digital acquisition if all the line of monitoring image is properly adjusted (in camera, on-set and in post).

 

[Tim Whitcomb]

Maybe a good test would be to shoot five faces of varying skintone color and quality, in daylight, tungsten, and Cool White fluorescent, on the M-X sensor and on 35mm film stock, at different exposures, and see if there is a skintone reproduction issue or not. If the "problem" can't be replicated, then maybe it doesn't exist...

Now this would be a fantastic test.

 

[Antonio Forjaz]

Ted, many people just have no idea what RED is....I have met many people in the film business that still argue that the 5D is better....I try my best to explain the differences... But you know, some people just don't want to learn, or know....but we know....and one day they will wake up, I hope, and see...feel the difference