A question about CRI

[Are Luke Muggerud]

After understanding the basics of color rendering I am left wondering about one aspect of this value

When using lights with low CRI to light a person, would the skintone be "unrecoverable" in a RAW based DI? Could you not in theory make the worst lights look as good as an ideal lightsource by applying just the right grading curves and increasing saturation to counteract the shaky color response or are you physically missing out on valuable raw image data?

 

[Blair S. Paulsen]

In the abstract it might be theoretically possible to do enough processing to the RAW data set to get virtually any result, but in reality, good luck with that. If all of your low CRI sources matched each other, you shot charts, the reference color temp was over 4000 Kelvin, you had a kick ass colorist with all the tools, exposure levels on key things like skintones were middle gray or higher - then you'd have a decent chance. It would also be fine if you wanted a surreal look or had other creative motivation for odd color rendition.

When you say "unrecoverable", to me that implies making a finished RGB image that looks like the subject's actual skin tone. If your goal is a pleasing skin tone, accuracy not important, then a proper colorist could simply "engineer" skin tones in DI - if that works for you, then yes, low CRI lighting could work. Be aware that low CRI sources with a higher reference color temperature, nominal daylight for example, will give the color grader a lot more to work in the RAW file than a 3200/tungsten source. You also better have some budget for DI.

From a practical perspective, I find it hard to make a case for deliberately using low CRI light sources for budget reasons considering the level of sophistication (read cost) necessary to fix the colorimetry issues in DI. If you're trying to shoot a passion project on nickels and are considering employing cheap work lights from Home Depot, I suggest you buy one light - shoot a couple of test clips with real people - take that footy into DI/CC and see what you get. If the colorist can build a LUT that does a decent job of compensating for the deficiencies of the lighting, and you use those same lights throughout the show, you might be able to get decent results - especially if some of the inherent character of the lights suits your creative vision.

FWIW, there are some pretty decent low cost lights on the market in terms of output characteristics but build quality usually tracks with price. If you handle them carefully, use them mostly indoors, have a couple of spares on hand, that's an option. If it were me, and for some reason I had to capture nice looking skin tones without proper lighting I would augment available lighting (daylight vastly preferred) with mirrors, reflectors, show card, etc before I would spray the set with an ugly light source.

Cheers - #19

 

[Matt W.]

You're actually missing out on important color information. But with digital you can get away with a decent image using pretty mediocre lighting.

 

[D Fuller]

After understanding the basics of color rendering I am left wondering about one aspect of this value

When using lights with low CRI to light a person, would the skintone be "unrecoverable" in a RAW based DI? Could you not in theory make the worst lights look as good as an ideal lightsource by applying just the right grading curves and increasing saturation to counteract the shaky color response or are you physically missing out on valuable raw image data?

You are physically missing out on valuable raw image data. I am working right now on a project for a major lighting manufacturer that involves color-controllable LED systems. One of the fascinating things I'm seeing is that you can produce the same color temperature with different mixtures of light colors. And while they look the same on a white card, they render the color of things that have color, like wood or fabric or skin, completely differently. It's bizarre, really, but when you produce light by mixing colors, it behaves very differently than light produced by heating a black body (a filament) or glowing gasses (flos and HMIs).

To correct skin in post is possible to some extent, particularly the overall color, but what we call skin tone is a complex patchwork of many different shades of color. Low CRI lights change those color relationships. There aren't enough qualifiers and secondaries in Resolve to get all of those back to what they would be in "natural" light. What's more, it would be impossible to know what the colors were supposed to be without a reference shot in natural light. So as Blair suggests, you probably could make a nice-looking picture, but an accurate picture would be another matter altogether.

 

[Marc Wielage]

When using lights with low CRI to light a person, would the skintone be "unrecoverable" in a RAW based DI? Could you not in theory make the worst lights look as good as an ideal lightsource by applying just the right grading curves and increasing saturation to counteract the shaky color response or are you physically missing out on valuable raw image data?

No. The picture winds up looking "forced" and "unnatural," for lack of a better word. It's better to shoot it right to begin with, rather than artificially skewing the color temperature and other factors to push it in the correct direction.

I've been presented with gross mixed-lighting issues before in certain scenes, and once you can't get clean whites and natural-looking skintones out of it, you're screwed. If they specifically want​ the scene to look bizarre, then there's no problem. Bizarre is easy; natural is hard.

 

[Peter Moretti]

While high CRI is better than low CRI, please don't put a lot of faith in high CRI. CRI measures how certain patches of color look under the light source. But there are large swathes of color between those color patches, and CRI is blind to how those look.

So a 5600 color temp HMI and a compact fluorescent can both have CRI's of 95. But PLEASE don't think they will look anywhere close to the same, because they won't. The CFL will have a NARROW bright green spike at about the 546 nm wavelength caused by mercury. The rest of the spectrum will be filled with alternating bands of no light and comparatively weak light. There may also be a narrow bright red spike as well. Kinoflos don't have the red spike. LED's dark patches aren't nearly as big.

An HMI, on the other hand, will have a pretty, continuous distribution of light. A tungsten incandescent will definitely have a smooth full distribution of light as well. (This is why color temperature is accurate enough to judge the quality of HMI and tungsten sources.) Plasma lights by Hive also emit a VERY nice full spectrum. That said, NONE of them are sunlight.

What you have to realize is that CFL's, sodium vapors and LED's (to a lesser degree) simply will not emit certain wavelengths of visible light. So when you hear that fluorescents have a green spike, that green spike is very bright and VERY NARROW, so the CFL is not emitting even all the wavelengths of *green*. And sodium vapors are yellow, but don't try using a blue filter to eliminate the yellow, because sodiums don't emit much else but yellow. That's why cars can look like they have odd colors or even black when parked under sodium vapors.

Hope that helps a bit.

 

[Jaakko Rinne]

After understanding the basics of color rendering I am left wondering about one aspect of this value

When using lights with low CRI to light a person, would the skintone be "unrecoverable" in a RAW based DI? Could you not in theory make the worst lights look as good as an ideal lightsource by applying just the right grading curves and increasing saturation to counteract the shaky color response or are you physically missing out on valuable raw image data?

The main problem is that if you're completely missing specific wavelengths of light, you can't "recover" them, since they were never recorded. I think Blair put it well in his post: a colorist could theoretically engineer the tones in, but that would most likely end up costing more than just making sure you have proper lights to begin with.

I work on the post on a number of reality shows and struggle with this issue all the time when we get footage shot in people's homes, where for some reason the crew has just used the "available" light from the fluorescents some people have instead of properly lighting the scene.

 

[Jason Comparetto]

I hope what I'm saying isn't redundant based on all the great answers everybody has already suggested, but here's sort of how I think of it:

With a low CRI source, its quite possible for the spectrum curve to be quite erratic. For example, you might have a low value for 467nm, but then at 475nm it will spike, and then be low again at 490nm, and then be normal at 514nm, but then spike again at 560nm and then be *really low* at 600nm and then get normal again around 660nm

So when you go to properly expose, what are you exposing for? Do you choose to expose for skin tones between 600nm-700nm ? but that value could be really low and so essentially you are OVER-exposing the other bands which leads to loss of information and a weird look. What about if you expose for the spectrum peak? Then you've under-exposed all the other parts of the spectrum, which might be ok if those are colors you artistically chose to hide. but in my example they happen to be the skin-tones. so now the skin tones are underexposed, so you end up having to boost them in post, which could look weird but passable, except other colors got rendered just fine so now the contrast is very noticeable in the same frame. even if the viewer doesn't notice it directly, internally they might pick up that something is off.

At the end of the day, ultimately you have to decide if the lighting choices you've made support the artistic quality you are trying to bring about in the film. If its a sickly sci-fi piece, perhaps excessive green tones in the skin are exactly what you want. if you are shooting blue iguanas, perhaps you could find lights with a big blue spike around the same nanometers as the blue reptile skin. But when you are shooting pretty women wearing pretty things in romantic environments, those skin tones need to be dead on, no pun intended. Those skin tones better look very healthy and support the story you are telling.

 

[Are Luke Muggerud]

Thanks everyone for very informative facts and theories which are also backed up in some of my tests with cheap knockoff LED panels. What i see when painstakingly trying to fine tune skin tone and removing spikes is that minor banding starts to appear and the compression starts breaking up before i can nail it. So - would anyone like to recommend a LED panel for someone looking for beautiful smooth 5600k light?

 

[Blair S. Paulsen]

Area 48 Softs and Kinos are my faves.

I also love the old big eye Moles, but they suck power and generate prodigious amounts of heat.

Cheers - #19

 

[Jimmy Gilmore]

A lot of good info here.

One thing to consider is what is what is acceptable to the eye for the context of the piece. For reality or news skin tones don't have to be perfect or necessarily match backgrounds. Good LEDs could be perfect for an action shot of a toddler beauty queen wrestling an alligator.

We shot an A lister last month and he was all into our lighting. Lower cost LEDs would have ruined the day whereas a lot of soft/bounced tungsten fill and hard backlight provided exactly what he wanted and at a much lower cost.

Another thing to consider is that post is used to working with Kinos, Tungsten and HMI. LEDs are trickier and that trickiness is time consuming and costly on the back end. The color suite is the most expensive room in post and screwing around there is way more expensive then bringing a natural looking, high CRI key light to the set.

my two cents,