Skin balance (rather than just white balance) picker

Hi Folks
I know "real colorists" don't use white and black point pickers but... for the less experienced...
wouldn't it be nice to have a skin picker?
This could be as basic as the regular eye dropper that says this pixel should be skin tone

It could eventually be as fancy and AI-based as identifying a skin area or face and letting Resolve party to make the skin look lifelike.

Skin and faces are what human eyes are most sensitive to - so making it easy to ensure those colors is a great road to better images for newbie Color Page users.

Well, when I use the qualifier (eye dropper) tool, I'd say a good chunk of the time is on skin color. I especially run into this if I have 3 actors that look great and one who has some makeup issues, so I'll try to work on the latter, and use a qualifier and a tracking window. Knowing how to use the qualifier is pretty basic. If they want to go even more basic and not use a qualifier, just click the red vector and make a little change there (hue or sat), and that'll do it a lot of the time.

White Balance is fraught with problems, and I've never liked the results on the Primary pallet. I do think that you can get a pretty reasonable facsimile of a white balance control on the Chromatic Adaptation plug-in, but I'd use that more as an overall for the actual color of white. You'd be surprised how often the overall color of the scene affects skin tone, and my choice is often to let it go on the assumption that (for example), if somebody is in a blue room, they're going to be a little on the blue side, or if they're close to a fire, they're going to be a bit on the warm side. So we don't necessarily use a qualifier to fix skin tones all the time.

Fair points all. Maybe what I’m thinking about isn’t feasible. Any idea what Canon’s color science does that tends to make skin tones look so good right out of camera? Might there be a way to do something like this in post?


Sent from my iPhone using Tapatalk

Fair points all. Maybe what I’m thinking about isn’t feasible. Any idea what Canon’s color science does that tends to make skin tones look so good right out of camera? Might there be a way to do something like this in post?


Sent from my iPhone using Tapatalk

is not the camera, but what you do with the data after you debayer that count.....

Any idea what Canon’s color science does that tends to make skin tones look so good right out of camera?

is not the camera, but what you do with the data after you debayer that count.....

Hi Walter. I don’t think I understand what you mean. Are you saying that Canon doesn’t have color science that makes skin tone look particularly good right out of camera?


Sent from my iPhone using Tapatalk

A friend down the road a bit does mostly local commercial video. Some for TV ad usage, much for company website use. Explainers, "meet our people", "here's our ethos system", that sort of thing.

So largely talking heads and b-roll of facilities or The Environment or whatever. A ton of skin tones.

He was using pro Canon camcorders. Tried a BMPCC4K last spring, and ditched his Canons

Because of skin tones in the BRAW. Much easier to get "pretty" skin he said.

Everyone's Mileage Always Varies.

Sent from my SM-S908U using Tapatalk

I’m sure you can find people who don’t like Cannon skin tones. Are we really going to say that all cameras are equal? It’s frustrating when people just disagree here. Is it really helpful somehow to just negate what someone says?


Sent from my iPhone using Tapatalk

Any idea what Canon’s color science does that tends to make skin tones look so good right out of camera?

is not the camera, but what you do with the data after you debayer that count.....

Hi Walter. I don’t think I understand what you mean. Are you saying that Canon doesn’t have color science that makes skin tone look particularly good right out of camera?


Sent from my iPhone using Tapatalk

correct. same for Arri, Sony, or Red: the cameras capture light, the electronic in the camera measure the light in real time and save it in a convenient file (RAW) for further manipulation.

how you manipolate that data after is what make the skin look good. For AGES I heard that Red cameras have a "bad skin tones" while the problem was not the camera per se but a combination of two factors:
1) Red cameras should be rated 800 or 400 ISO. stop using 1600, is too effing noisy and the sensor doesn't get enough light in the first place.
2) the LUTs/color science after you debayer teh data was plain crap. (This is the 800 pound gorilla)

get a movie like "Hustlers", that is done on red cameras, the skintones are where we want it to be.....

I think we’re talking past each other.
I have a decent understanding of how cameras work. I’m not talking about that. I’m talking about the color science the cameras do after they acquire the raw data from the sensor.

So I’m mostly talking about number 2 that you mention above.

Cannon - and Arri even more so - tend to look great right out of the camera because of their color science. Many other often-used brands less so.

It’s great that, with modern tools like Resolve, one can if they have the skill get amazing results. But the skill required is far greater than back in the day when we used film. The emulsion was really smart - a colorist built in. It just looked good.

Arri took the time with the Alexa to make things look great right out of the camera - largely I think because they were trying to make a system that felt a lot like film. And, though they were late to the party, everyone ate it up.

Oodles of companies have made LUT packs and more promising to mimic what Arri has done using other cameras. They work with varying degrees of success, complexity, and expense.

Is there a Resolve workflow that gives people an Arri level of simplicity and good initial results? I don’t think so. Sure would be nice if it existed.


Sent from my iPhone using Tapatalk

Kind of a side note, but I was blown away by the well-exposed skin tones coming from the original Pocket in an ACES work flow. Just blew away any camera I'd used previously.

All the colors, actually, nut just skin. BMD color science..I'm a fan.

I’ve got 3 of those. Haven’t tried them recently - or ever with ACES. I’ll have a look!


Sent from my iPhone using Tapatalk

Wasn't any attempt to be disagreeable. Simply pointing out (to me) the obvious thing. Note the everyone's mileage always varies bit?

I know of many who love Canon lenses in video. And others who give long lists of the horrible issues of Canon lenses. That isn't a bad thing. In normal human existence we all see things differently.

It's a great source of width and depth of ways of seeing "the same thing".

You like the look of those files ... when displayed where? How? And what specifically do *you* see in the skin tones that you want to duplicate in Resolve? That's (to me, of course) useful information to build a discussion around.

Sent from my SM-S908U using Tapatalk

It’s not that I’m seeking a particular look. Though when I see Arri and Canon footage right out of camera they look more pleasing and filmic to me. Many others agree - it’s a general reputation.

There are probably many ways to skin that cat. Just like I loved both Kodak and Fuji film even though they were both quite different.

Arri made a look that works right out of camera. RED chose the “you can make it look like anything so we’re not going to bother making it look great out of camera” approach. (IMHO)

Arri’s choice won. When I was doing computer startups this dichotomy often came up. Time and again the “make it great without need for a ton of customization” approach always won. People said nah - just let the user make it how they want. And those startups died.


Sent from my iPhone using Tapatalk

For AGES I heard that Red cameras have a "bad skin tones" while the problem was not the camera per se but a combination of two factors:
1) Red cameras should be rated 800 or 400 ISO. stop using 1600, is too effing noisy and the sensor doesn't get enough light in the first place.
2) the LUTs/color science after you debayer teh data was plain crap. (This is the 800 pound gorilla)

Yeah, that was my experience as well for 6-7 years. Finally with Red Helium around 2016, I think they completely cleaned up their act: pictures got less noisy, color science was cleaner, and in particular, skin tones got a lot more "real" (at least to me). Once CST nodes came out and had Red IPP2 included, it became pretty easy to get perfectly nice pictures.

But only if it were well-exposed. If it was underexposed... not good.

There are probably many ways to skin that cat. Just like I loved both Kodak and Fuji film even though they were both quite different.

If you think back to the film Kodak was accused of making stuff that made caucasian skin look good. To the point of being accused of being racist.

Fuji was better for Oriental skin.

Reality is most of the differences related to exposure and printing being messed up.

If you think back to the film Kodak was accused of making stuff that made caucasian skin look good. To the point of being accused of being racist.

Fuji was better for Oriental skin.

This thesis sounds a bit esoteric to me, as if there was just one specific skin color on each continent which, of course, is not the case. The range goes from pale to dark brown in almost every region of the planet. Even Norwegians can get brown when they stay in direct sunlight for several hours :-)

Also, I would say, the skin tone hue variation is extremely tiny across the world. And when considering even that little hue variation: it varies within one continent just as it varies within any other continent, depending on the indivual's age or blood pressure etc.

I would like to see visual examples of the above thesis. I agree that there are differences when looking at the statistical average between "Caucasian" and "Oriental", but I see an overlap of 90%. They overlap so much that it's impossible to divide them into distinct categories.

WB/BB picker targets singular points whereas skin tone is represented (roughly) as an entire vector. 40% generated grey 102, 102, 102, becomes 253, 253, 253 after WB picker selection. How would this yield good results for skin tone? All skin regardless of tone or exposure would be mapped to a singular rosy pink point? The tool would need to ignore luma and solely focus on chroma. As a discrete function perhaps, but it wouldn't work using the same math as the WB/BB picker.

Good Luck

Skin Tone.png (140.12 KiB) Viewed 1925 times

My suggestion may be unworkable. But it would be nice!


Sent from my iPhone using Tapatalk

If you think back to the film Kodak was accused of making stuff that made caucasian skin look good. To the point of being accused of being racist.

Fuji was better for Oriental skin.

I would like to see visual examples of the above thesis..

Google

Kodak film is racist

You'll have everything from the NYT to the Guardian to even academic papers pop up.

Thanks. I'm seeing many Google hits that are about black versus white; this difference is obvious, of course. My comment, however, was about the generalization of "oriental skin". I think in Asia ("orient") we find skin tones ranging from dark brown to "Caucasian" white. How can such a broad range be generalized into a single color tone? Maybe we have a language barrier here and we should define the word "orient".

Fuji being Japanese I'd think that would be the target for their engineers.

Thanks. I'm seeing many Google hits that are about black versus white; this difference is obvious, of course. My comment, however, was about the generalization of "oriental skin". I think in Asia ("orient") we find skin tones ranging from dark brown to "Caucasian" white. How can such a broad range be generalized into a single color tone? Maybe we have a language barrier here and we should define the word "orient".

It was something about using the word "yellow people" "yellow skin" for Asian and Asian America as denigratori and Kodak stock having a pronounced yellow/ warm cast in the light skintone, therfore when Fuji was manufactured a more red/ magenta cast was introduced for a less warm skin color.

For black/ darker skin, the dark you go along the Kodak stock, the more blue/ desatturated/unnatural the skin it become, and that made Africans and African American look awful on screen

I see. Thank you.

Kodak negative and print combinations whether for stills or movies had a number of characteristics that could make certain skin tones problematic.

For example, many redheads skintones were simply sickly blue/magenta. Other redheads could be very yellow.

My wife is a "standard" brunette, but her skin always looked very yellow greenish. We couldn't use her for our in-house "Shirley's" because her skin recorded so far off from "normal". We used me instead.

African tones were a struggle. Some of the really dark ones would shift very blueish, while another who looked near identical to the eye stayed "warm". You could do a portrait of a family, who looked very very close to the eye, but every face was different in the prints.

Sometimes the lighter tones were warm/yellow, darker tones blueish. On the same face.

Asians could also be an issue. Most Koreans did fine, and most southeast Asians we photographed were decently recorded.

Some of the very pale Japanese were as difficult as the redheads to get "right". And many Chinese had the same issues as my Scottish/English (ethnicity) wife, of showing far more greenish yellow than the eye saw them.

So group photos could be an interesting challenge to put it mildly.

For awhile in the 80s we got Hunt paper, which was overall a bit more neutral for skin than Kodak paper. But then they stopped distribution through the channels we had access to.

Shooting groups with mixed skin tones was a chore. Darker tones needed to be closer to the main light and well-filled to keep "warm". Redheads/light Japanese needed to be in the middle. Not as brightly lit. Yellow/green did better (sort of) with more shadows.

Try and set that up in a 30 person group shot, and you've got six minutes from when they walk in the room to pose them all and get 5-8 shots off. Yowza.

Now ... find a way to make a print so no one looked too bad ...

No ... I don't romanticize shooting with film.

It's easier to shoot and especially fix digital.

Sent from my SM-S908U using Tapatalk

Kodak negative and print combinations whether for stills or movies had a number of characteristics that could make certain skin tones problematic.

I would argue that motion picture prints (which comes closer to the kind of color handled by Resolve) is a much different area that still-photo prints on paper.

I don't think Kodak was racist per se, but they did reflect the world in which movies were made. I can tell you by the 1980s for sure, Kodak had test films that had all kinds of people in them, lit under a variety of circumstances, with different kinds of makeup. Even before the outcry of racism, Kodak actually did start using people of color in a lot of their promotional materials and internal color science testing. And by the way, I can't emphasize enough the need for good makeup and good lighting. Without that, skintone often goes to hell, and it's not so much a color-correction problem as it is a production problem.

Having digitally color-timed a lot of both Kodak and Fuji negative film in the 1980s and 1990s, what I saw was that Fuji film tended to be a little more saturated, and the shadow detail kind of "fell off a cliff" -- that is, was non-existent -- if it was underexposed. Kodak negative film looked a bit more natural to me, and when people moved into shadows, the toe of the exposure was a lot more gradual and gradiated. Some producers tried to save money by shooting on Fuji stock, but inevitably it looked a little weird and "grainy" compared to Kodak, at least through the mid-1990s.

But something changed around the time Kodak created the T-Grain Vision stocks in the mid-1990s. Fuji stepped up, figured out how to match it, and suddenly Kodak and Fuji films started looking very similar. We were pleasantly surprised by how good the Fuji negative looked, particularly in terms of skintone, and Kodak was forced to start giving discounts to TV shows shooting film. But that tapered off and pretty much evaporated by 2010-2011, once production largely shifted to digital.

You can point to a lot of digital cameras as having specific "looks" in terms of skintone, and I know in particular that Arri Alexa cameras had a reputation for looking more natural -- which is to say, more like Kodan film -- than anybody else. I think many would agree that Red cameras suffered by comparison for skintones for the first 6-7 years of the Red One, the MX, and the Dragon, but I think by the Red Helium in 2016, something changed... and I think the pictures from Red started looking fantastic.

Chasing down skintone accuracy is tough because things change so much with exposure, with certain lenses, and with specific hues and saturation/luminance levels. We've all had projects where you have 3 people in a scene, and one looks great, one looks OK, and the other looks awful... and a lot of that is more due to makeup issues than anything else. Trying to solve this in final color is a huge problem. You can do it to a point, but it's annoying when you realize that you have to spend an hour in post trying to fix something that could have been averted in 5 minutes with a good HMU person on set.

Thanks Marc for that illuminating post.

Oh, yes, Kodak had *massively* improved skin and general quality in the 80s. I used both stills and film stocks at times.

And there were further improvements in the 90s. Both in Kodak and Fuji.

But the stills really showed the issues, printed from either stock type. Being as there's no motion, constant changes of scenes, all of that "thing" of motion capture which minimizes many small things.

Not as big as the 70s, but still there.

With decent lighting and capture, skin in post now, is to me SO much easier to deal with than in Days of Yore. Which is why we get tasked with fixing things that were never touched before.

And I agree that if someone pays a *small* bit of attention to skin/hair prior to shooting, that's a huge lift in post.



Sent from my SM-S908U using Tapatalk