skintone

[Blair S. Paulsen]

I agree that RC3 really needs spot on white balance to shine. IIRC when RC3 first landed Jim made it clear that nailing the color temp was the key to getting a purer color palette.

FWIW I will resample white point several times if my first grab doesn't seem right. This also brings us back to some earlier posts about the value of shooting charts/grey cards to improve color alignment in post. I am a big fan of stylized creative looks, but still believe that starting with untwisted color is the preferred way to get there in most cases. YMMV.

Cheers - #19

 

[Matt Ryan]

that's without any additional grading, just the LUT?

yes sir. Didn't even adjust any levels, which I could have. The link I posted is to a couple more samples where I literally just added the output LUT to them.

 

[paulherrin]

yes sir. Didn't even adjust any levels, which I could have.

well, it looks pretty nice!

 

[Matt Ryan]

well, it looks pretty nice!

thanks! I've been really happy with the quick looks you get from SpeedLooks without any tweaking. It gets me to an almost perfect starting point and finish in this case.

 

[David Battistella]

This also brings us back to some earlier posts about the value of shooting charts/grey cards to improve color alignment in post. I am a big fan of stylized creative looks, but still believe that starting with untwisted color is the preferred way to get there in most cases. YMMV.

Cheers - #19

agree. there has to be a reference, or baseline, even if thelook is going to be taken in whatever direction.

David

 

[Bruce Allen]

LogC input, 3500Fuji Output LUT:

http://wp.me/p242Io-j0

I know this is a popular and cool look right now. Unfortunately, that is a good example of what I'm personally not a fan of for every project. I understand that it hits the memory color for flesh tones so it is of course a great idea for lots of things.

My only issue is that those people just look like one orange color. Like one of those old black and white photos that has been hand-colored by just putting orange tint over it.

I just want to see if we can make RED do something with more variation going on in the color that still looks natural. Eg like the top or bottom photos in this (not a huge fan of the middle one):
http://photographylife.com/nikon-vs-canon-vs-fuji-in-a-studio
That's straight out of the camera color rendition for a 5D Mk3 and a D800.

I don't know if it can be solved by a RGB -> RGB LUT. The world is not made up of just red, green and blue, but a spectrum of light. Maybe there are certain wavelengths which are causing crosstalk and if we filter them at capture time, we can sort this out? I know there's like a 1% chance of this panning out but it'd be cool, right? Maybe we can do a series of 3D LUTs that solve the issue for certain lighting conditions?

Promise to do a test soon. Not sure if I can make anything public (I hope to of course!) but if I come up with anything cool for a RED camera, I hope to share it with the community.

If anyone has any links or info on:
- what makeup works and doesn't work well with RED under certain lighting conditions (there are a couple threads but I'd love to actually get hold of the good and bad ones and graph their spectral response)
- where I can rent a high-end spectrophotometer or spectral camera (these things are like >$10,000 so rental is the only option sadly)
- where I can rent either a set of very narrow-band spectral filters or a tunable spectral filter (this will cost hundreds of dollars at the very least)

Or any input on test ideas, please email me at boacinema at gmail or post here if you would rather continue the conversation in public.

Thanks!

Bruce Allen
www.boacinema.com

 

[Matt Ryan]

I know this is a popular and cool look right now. Unfortunately, that is a good example of what I'm personally not a fan of for every project. I understand that it hits the memory color for flesh tones so it is of course a great idea for lots of things.

My only issue is that those people just look like one orange color. Like one of those old black and white photos that has been hand-colored by just putting orange tint over it.

I just want to see if we can make RED do something with more variation going on in the color that still looks natural. Eg like the top or bottom photos in this (not a huge fan of the middle one):
http://photographylife.com/nikon-vs-canon-vs-fuji-in-a-studio
That's straight out of the camera color rendition.

I don't know if it can be solved by a RGB -> RGB LUT. The world is not made up of just red, green and blue, but a spectrum of light. Maybe there are certain wavelengths which are causing crosstalk and if we filter them at capture time, we can sort this out? I know there's like a 1% chance of this panning out but it'd be cool, right? Maybe we can do a series of 3D LUTs that solve the issue for certain lighting conditions?

Promise to do a test soon. Not sure if I can make anything public (I hope to of course!) but if I come up with anything cool for a RED camera, I hope to share it with the community.

If anyone has any links or info on:
- what makeup works and doesn't work well with RED under certain lighting conditions
- where I can rent a high-end spectrophotometer or spectral camera (these things are like >$10,000 so rental is the only option sadly)
- where I can rent either a set of very narrow-band spectral filters or a tunable spectral filter (this will cost hundreds of dollars at the very least)

Or any input on test ideas, please email me at boacinema at gmail or post here if you would rather continue the conversation in public.

Thanks!

Bruce Allen
www.boacinema.com

The lighting levels I lit this for would never give skin tones like the stills examples you just posted. Was never going for the "beauty" photo look. Here's the set:

password:muse

I understand what you mean, but I'm not using the still as an example of beautiful skin tones and rendition. It was just my sample of the SpeedLooks LUT.

On a side note, skin tone is a subjective thing. Unless you see the subject in person, commenting on how their skin tones and coloring should be is subjective. Theres a ton of factors that influence what the camera is capturing. For example, the link you posted with the three images, the middle still skin tones look different from the first and third. First and third have more variation and subtleties between the cheeks and her face while the second is kind of flat and should I say RED looking?

 

[paulherrin]

very nice examples there, bruce... i'm definitely with you on the idolized 'film look'

i still think the best test would/could be a short. either way, you get into a lot of hassle and decision-making about how/who does the shooting/grade for each camera... :/

 

[Blair S. Paulsen]

One aspect of Bruce's skin tone issues that I find valid is the concept of subtle variegations in tonality being important. Due to evolutionary imperatives (friend or foe) humans have a highly developed facility for judging facial characteristics including nostrils flaring, eyes narrowing, flushed cheeks, etc. If the photography misses some of these cues due to "plastic" skin tones then something important has been lost. That said, there are too many RED originated projects with good skin tones for me to conclude the imaging platform is inherently at fault.

IMO the power of RAW color information provides immense control that can be mishandled. I would even go so far as to say that one of the reasons (yes, there are others but that's another thread) the Alexa is so popular is that its harder to screw up skin tones. I also expect that better dyes on the Bayer CFA used for the Dragon will make it easier to get pleasing skin tones "out of the box" than the current MX sensor tech. Graeme has done a terrific job of managing the data set created by the MX sensor to yield good results but with purer spectrometry to begin with I fully anticipate a better starting point.

The really cool thing about RED's approach is that virtually nothing is thrown away when creating the R3D files, so as we have seen with new iterations of the color science, better characterization can produce better color linearity.

Cheers - #19

 

[Steve Sherrick]

Okay, let me see if I can also arrange a test. It may have to be a RED only test for now as getting hold of an Alexa or F65 may be difficult at the moment. Although I am not a fan of comparing digital to film at this point as I feel the capture formats are just inherently too different, it might be interesting to have a professional still photographer on hand to take stills with a film camera (and digital just for the fun of it) just to have as a reference point.

In this test there would be a RED One and an Epic. The goal would be to achieve a natural looking skin tone (one that meets Bruce's tough criteria. Love ya man!). We would be looking to achieve a close match to what we see by eye. Would try to capture subtleties that Bruce mentions. Will shoot under straight tungsten, Daylight, and mixed color temps.

I'll need that following people to pull this off.
- Gaffer
- Makeup artist with a lot of experience with RED
- Professional photographer who still shoots film
- Model. I find RED cameras capture darker skin very nicely, so in this case I think lighter skin with some variations in it to see if we can capture all of it would be best.

If anyone in the northeast fits these roles and wants to do it, get in touch. Makeup is huge part of this, so needs to be someone who knows their stuff.

I will get a hold of a Red Cambook.

In the meantime, I will also try to investigate possibility of getting Alexa and F65. Maybe I'll just make an event out of it and call it Organic Beauty in a Digital Age.

 

[Dave Blackham]

Here are a couple of other observations. With the current cameras capturing so much detail in terms of resolution and also colour information cleanly is a huge benefit but also seems to me that its sometimes capturing some potential problems too. Any lighting instruments that may not output continuous spectrum or have spikes in the spectrum ie LED's, HMI's etc are now much more apparent and skin tones are the first to suffer. Also the tools available in grading now are hugely powerful and the tools are much more widely available in the industry, hence powerful grading tools are in less skilled hands. Trying to correct a spike or hole in spectrum can lead to all manner of strange issues.

The ability to create something not so great looking by the less skilled is equally as possibly as creating something of beauty in skilled hands. It may be we are seeing more imaging issues due to the volume of content thats now available also some flaws we now see may have been masked in some way in the past with cameras that are not so good as the camera we have available now.

 

[Björn Benckert]

Hey Bjorn, what do the different abbreviation things mean? CUSPclip, ConstL, etc., and what effect they have?

Ironically, I actually really love that rich positive/print look, which I usually tried to mimic/fake by adding more saturation to things like skintones and shadow saturation (which didn't really work.) I like where these LUTs start you off. I've tried experimenting with RC3 (instead of RC2) and while the colours may not be right, it still looks pretty good. Really have to watch white balance though... things can react weirdly if you're off.

I read all the documentation of the Cusp clip and all that a while back and if I remember correctly it's different falloffs. (might be wrong) So it's actually mimicking different projectors.


What I find work the best (do not know the equivalent workflow for resolve but it should be possible to do the same) I bring in the r3d straight into the flame batch and open the clip setting so I can do all the white balance, iso etc, and I then usually have the clip in redgamma 3 while I do that, Since I find the redgamma 3 high saturation and contrast/gamma to show the native pituce in a good way. When I think I balanced it well enough. I flip redgamma 3 to red log film. but keep all other clip settings. Then I add something thats called a "lut editor" in flame (it's basically a primary grading tool working in log color space. Then after that node I add one of the print 3dl Luts. Thats a basic setup.

then when having the clip, the lut editor and the 3dl Lut I make sure to have my monitor to display the last node, the 3dl lut and then I apply all my grading and effects under/before that node. So basically all grading and compositing is done in log and the 3dl Lut is the last node that transform the clip to Lin.

I know it's very possible to do quite the same in resolve I just do not use it often enough to be able to describe how.

 

[Scott Crawley]

agree. there has to be a reference, or baseline, even if thelook is going to be taken in whatever direction.

David

FWIW I say the same. I start with a color temp reading and a chart. Some will only bother with the color meter in mixed lighting scenarios but I'm not one of those. If I get the temp reasonably right, and have a reference, correct skin tones are not far away. From there, you can carve out whatever look you like, but if shots are not matched in step #1 the look will not apply uniformly and subtle but critical tones like skin are usually the first to take a hit.

Every skin tone problem I've ever had it seems, dating back to my first Sony still camera was related to a blue deficiency. Apparently silicon transmits 95% of the light spectrum... I have read that Germanium is even more efficient at transmitting light. There are chips being made of Germanium and a synthetic Germanium apparently exists. I would love to see Red experimenting with a new substrate. Germanium wafers and imaging chips seem on the surface to be made for one another.

 

[Timothy Eaton]

Or any input on test ideas, please email me at boacinema at gmail or post here if you would rather continue the conversation in public.

Hi Bruce,

I'm curious to know how much of your own testing you have done for skin tones, direct side by side comparisons of Epic to other cameras and systems. I am not suggesting Epic is perfect by any means, but a number of factors make me skeptical about claims that Red skin tone performance is sub par: 1) The number of variables in many comparisons is very great (lighting, lenses, makeup, post, etc. etc. etc.). 2) My own experience has been good with Epic. 3) There are some terrific examples of great looking movies in which flesh tones are complex, rich, and natural. 4) Last but not least, there is a fair amount of "down with Red" hype floating around out there in zones that cross into both professional and prosumer (not saying you are part of that hype at all).

I have had an occasion or two in which an issue of skin tones came up, but it occurred to me later that I couldn't necessarily assume that it was a problem of sensor or color matrix. I am still open to the possibility that when issues arise, it could be a function of extraordinary new and complex sensitivity to color and light. The camera is seeing things (artifacts of light, spectrometry, or nature) that my eyes either do not see, or if they do, my brain compensates with a more pleasing rendition. Given that we are extraordinarily sensitive to the human face, I think this is not out of the question.

It seems fair to ask that a camera like Epic, if it is extremely sensitive to color variability, render skin tones that are pleasing, provided we have taken all the necessary steps to insure that the variables are right: that the scene is lit right, makeup is proper, color temperature is correct, no spectral problems, etc. But I'm not sure we know what all those variables are yet related to the world of digital sensors. In the world of film, for example, I learned that the color temperature of the early HMIs (LTM) was extremely variable, and started ordering up all the flavors of minus green to live on the truck. And even now, I don't use LEDs, because most of them don't deliver anything close to a balanced spectrum.

Which brings us back to testing. For those who say that Alexa handles skin tones better, have you done side by side comparisons, variables controlled, and color corrected in RAW? Or how much of it is hearsay, or simply forming an impression based on different clips shot under entirely different circumstances?

Tim Eaton

 

[Steve Sherrick]

Hi Bruce,

I'm curious to know how much of your own testing you have done for skin tones, direct side by side comparisons of Epic to other cameras and systems. I am not suggesting Epic is perfect by any means, but a number of factors make me skeptical about claims that Red skin tone performance is sub par: 1) The number of variables in many comparisons is very great (lighting, lenses, makeup, post, etc. etc. etc.). 2) My own experience has been good with Epic. 3) There are some terrific examples of great looking movies in which flesh tones are complex, rich, and natural. 4) Last but not least, there is a fair amount of "down with Red" hype floating around out there in zones that cross into both professional and prosumer (not saying you are part of that hype at all).

I have had an occasion or two in which an issue of skin tones came up, but it occurred to me later that I couldn't necessarily assume that it was a problem of sensor or color matrix. I am still open to the possibility that when issues arise, it could be a function of extraordinary new and complex sensitivity to color and light. The camera is seeing things (artifacts of light, spectrometry, or nature) that my eyes either do not see, or if they do, my brain compensates with a more pleasing rendition. Given that we are extraordinarily sensitive to the human face, I think this is not out of the question.

It seems fair to ask that a camera like Epic, if it is extremely sensitive to color variability, render skin tones that are pleasing, provided we have taken all the necessary steps to insure that the variables are right: that the scene is lit right, makeup is proper, color temperature is correct, no spectral problems, etc. But I'm not sure we know what all those variables are yet related to the world of digital sensors. In the world of film, for example, I learned that the color temperature of the early HMIs (LTM) was extremely variable, and started ordering up all the flavors of minus green to live on the truck. And even now, I don't use LEDs, because most of them don't deliver anything close to a balanced spectrum.

Which brings us back to testing. For those who say that Alexa handles skin tones better, have you done side by side comparisons, variables controlled, and color corrected in RAW? Or how much of it is hearsay, or simply forming an impression based on different clips shot under entirely different circumstances?

Tim Eaton

You're getting warm.
In fact, you just about nailed it. RAW is wonderful, but get the lighting and makeup wrong on set and one can easily end up with a problem that can also be magnified if further mistakes are made down the line in the processing.

But, I feel we could probably all go back and forth about this all day and really the proof is in the pudding. I'm working on putting together an event that will focus on this topic. Stay tuned....

By the way, being there at the event is probably the best way to judge because you'll see it all unfold. But it will probably not be in the best location for a lot of you.

 

[Blair S. Paulsen]

Perhaps lighting matters after all, who knew ;-)

Perhaps lighting matters after all, who knew ;-)

I think David B. makes a very salient point about the influence of the lighting CRI (color rendering index) on skin tones. I work with a producer who was unhappy with skin tones and perceived sharpness on some green screen stand ups so he asked me if I thought the lights we were using might be the culprit. I suggested we rent some Kinos for the next shoot and the improvement was substantial. Perhaps the concept that more precise readings of color made possible by modern imaging platforms makes lighting quality more critical is the key. My analogy would be how much more aware of motion blur I became when first watching "Crossing The Line" in 4K from a digital projector - without the gate weave, flicker, stochastic grain and other attributes of film origination and exhibition to mask it the motion blur was far more noticeable.

Perhaps if Steve manages to produce a test he could include some lighting options as exemplars. Yes, I know its easy to armchair quarterback when I don't have to assemble the resources (sorry Steve) ;-)

Cheers - #19

 

[Steve Sherrick]

I think David B. makes a very salient point about the influence of the lighting CRI (color rendering index) on skin tones. I work with a producer who was unhappy with skin tones and perceived sharpness on some green screen stand ups so he asked me if I thought the lights we were using might be the culprit. I suggested we rent some Kinos for the next shoot and the improvement was substantial. Perhaps the concept that more precise readings of color made possible by modern imaging platforms makes lighting quality more critical is the key. My analogy would be how much more aware of motion blur I became when first watching "Crossing The Line" in 4K from a digital projector - without the gate weave, flicker, stochastic grain and other attributes of film origination and exhibition to mask it the motion blur was far more noticeable.

Perhaps if Steve manages to produce a test he could include some lighting options as exemplars. Yes, I know its easy to armchair quarterback when I don't have to assemble the resources (sorry Steve) ;-)

Cheers - #19

Yes, lighting variations would definitely be part of the test. In fact, we'll probably attempt to do it wrong as an example.

I have had more than one occasion now where I have had to have a bulb swapped out on set due to spikes in the spectrum that might have been missed if I didn't speak up. One in particular where I was told it's a fairly new bulb and should be fine but upon a color temp reading was way out of spec.

 

[Dave Blackham]

Yes, lighting variations would definitely be part of the test. In fact, we'll probably attempt to do it wrong as an example.

I have had more than one occasion now where I have had to have a bulb swapped out on set due to spikes in the spectrum that might have been missed if I didn't speak up. One in particular where I was told it's a fairly new bulb and should be fine but upon a color temp reading was way out of spec.

I guess what Im trying to say is whilst we can capture very accurate and great looking images we are also now seeing the defects in some cases subtle in other imaging tools. ie lighting and also as has been pointed out some other areas, such as make up.

Remember some years back when HD for TV was launched complaints were made that scenery, make up, etc had to be so much better as we could now see all the blemishes, I think that we can now see much more of the detailed defects. My guess is the better the camera the more apparent some of these issues become.

 

[Mike P.]

A little OT, but I've noticed excessive (to my eye, anyway) motion blur as well on RED... I didn't think it would be because of traditional film artifacts (like grain) not being present.

Speaking of lighting, I find that skin tones lack "thickness" when using tungsten, which I presumed was solely because of the lack of blue (you can see it in the skin tones; lacking sharpness and strength compared to the other two channels). When you white balance tungsten the entire blue channel (which is underexposed compared to R and G) gets bumped up on the entire image and not just the skin tones/mids. It's obviously very noticeable when the amount of light is limited, otherwise it goes away when you downscale. I still haven't tried LED lights, but I was fairly optimistic (maybe too much so) that they'd play nice with RED/CMOS sensors (they have a tendency to go green though don't they?)

I've also noticed the "yellow" skin blotches when trying to balance under florescent lighting... And in those cases the subjects were NOT wearing make-up. When something like that happens, you almost have to ride it out and go with stereotypes (like a matrix look for florescent or dragon tattoo look for tungsten) since trying to correct it in post would not be feasible. Lots of needless variables that *seem* to be compensated for with other camera systems.

Bjorn: Where did these LUTs originate from? Did you guys do them in-house (Syndicate) or does Kodak/Fuji provide them? I didn't realize they were mimicking output/projection... very interesting (and somewhat confusing). So you go Balance/Expose tweaking footage as RG3 before switching to RLF >Primary node being a Look LUT > Final Node being one of these Output LUTs? Doesn't that negate your white balance that you did when you first loaded up the footage?

 

[Bob Gundu]

SpeedLooks RedLogFilm to Log C

SpeedLooks Clean Fuji

SpeedLooks Clean Kodak

Kodak 2393 log2HD (lifted mid tone with luminance curve and pumped saturation)

Arri Log C to REC709

RedColor3 RedGamma3