Why does the color in this A7s video look so fake?

DPStewart said:
Youve just picked the two DSLR type cameras that seem to have the most problems delivering good skin tones. Right out of the camera Canon and Nikon render much better and more pleasing skin tones - hence their dominance in the wedding video market.

All cameras have their own unique look - sometimes called "color science".
I have a Panasonic, but I like its natural colors far less than what comes out of a 5DmkIII or a Nikon D750/810.

Everyone will have their own preferences in camera character.[...]
Click to expand...

Yes, I agree, it's also a matter of personal preferences indeed. I never liked the colors from my 7D: they're too saturated/contrasty, unless you tweak the PP (and still I never been really happy about it).

I'm using a Sony A6000 and for the first time I'm loving the colors even just straight out of the box. Somehow I find Sony and Nikon quite close in terms of color.
 
dsp_418 said:
Yes, I agree, it's also a matter of personal preferences indeed. I never liked the colors from my 7D: they're too saturated/contrasty, unless you tweak the PP (and still I never been really happy about it).

I'm using a Sony A6000 and for the first time I'm loving the colors even just straight out of the box. Somehow I find Sony and Nikon quite close in terms of color.
Click to expand...

Wow, yeah. My sentiments exactly. Our eyeballs must have been manufactured at the same facility.

That is EXACTLY what I instantly noticed about the Sony A6000 - "Damned if this thing don't look EXACTLY like what's coming out of the Nikon D750!"
Of course Nikon is putting a Sony sensor in the D750. Also in Stills Mode, the A6000 does NOT look like the A7s to me, and I prefer the A6000.

I guess cameras are like partners - you gotta pick the one you like the look of.
(snark)
 
I'm going to dip my toe in the water here. I'm not going to address Jason's photography as we all have our own recipes for exposure, and lighting. And that's our secret sauce but I want to point out the sloppy exposure techniques I see, and read about here and elsewhere.

Everyone says the a7s is noiseless then in the same breath state that ETTR yields the cleanest images (which is an admission that correctly exposing Slog is not noiseless). There seems to be the belief that the practice putting the camera's light meter in Matrix metering mode (aka Wide, Zone, or Center) and opening up/gaining up until it reads +1.0 or +2.0 = 1 or 2 stops over proper exposure is false.

SLog 2 correctly exposed according to Sony is 32 IRE as read on a WFM of an 18% grey card, NOT THE OVER ALL IMAGE. Now you can open up from a determined F stop to the next increment on your lens barrel to correctly overexpose by a stop. Using Matrix metering can put you very high in the tonal range and wreaks havoc on flesh tones as correctly pointed out in this thread. Without an 18% grey card you can rough in exposure using a white card exposed at 59%, again, you need a WFM. If you're exposing the sky at 90%, you're way over and shouldn't expect flesh tones to look optimal. That's not how the curve is designed.
 
bill totolo said:
I'm going to dip my toe in the water here. I'm not going to address Jason's photography as we all have our own recipes for exposure, and lighting. And that's our secret sauce but I want to point out the sloppy exposure techniques I see, and read about here and elsewhere.

Everyone says the a7s is noiseless then in the same breath state that ETTR yields the cleanest images (which is an admission that correctly exposing Slog is not noiseless). There seems to be the belief that the practice putting the camera's light meter in Matrix metering mode (aka Wide, Zone, or Center) and opening up/gaining up until it reads +1.0 or +2.0 = 1 or 2 stops over proper exposure is false.

SLog 2 correctly exposed according to Sony is 32 IRE as read on a WFM of an 18% grey card, NOT THE OVER ALL IMAGE. Now you can open up from a determined F stop to the next increment on your lens barrel to correctly overexpose by a stop. Using Matrix metering can put you very high in the tonal range and wreaks havoc on flesh tones as correctly pointed out in this thread. Without an 18% grey card you can rough in exposure using a white card exposed at 59%, again, you need a WFM. If you're exposing the sky at 90%, you're way over and shouldn't expect flesh tones to look optimal. That's not how the curve is designed.
Click to expand...

I've heard this before about overexposing skintones in SLOG2 causing problems, but I've also seen tests and shot lots of material myself that indicates the opposite. Could you site any sources for your claim? Or perhaps offer an explanation? My understanding of SLOG2 is that because it's slope remains relatively constant (without knee, minimal shoulder) it is quite easy to apply a log offset operation to footage to simply increase exposure in a way that does not significantly affect color.
 
nyvz said:
My understanding of SLOG2 is that because it's slope remains relatively constant (without knee, minimal shoulder) it is quite easy to apply a log offset operation to footage to simply increase exposure in a way that does not significantly affect color.
Click to expand...

Yeah, that's what you'd do with log footage to manipulate exposure since cineon times. The issue with a7s apparently is that there is a shift of color throughout the tonal range in S-gamut: tones subtly go from magenta in the darks to greenish in the whites. Shouldn't be a problem in other color modes with more consistent color.
 
cpc said:
Yeah, that's what you'd do with log footage to manipulate exposure since cineon times. The issue with a7s apparently is that there is a shift of color throughout the tonal range in S-gamut: tones subtly go from magenta in the darks to greenish in the whites. Shouldn't be a problem in other color modes with more consistent color.
Click to expand...

http://i1119.photobucket.com/albums/k636/hingsberg/ScreenShot2015-01-20at44128PM_zps1b71ab7e.png

S-gamut is so much bigger than Rec709 that lots of funky things can happen trying to wrangle it back in. On the FS7 Sgamut3.cine has been a god send as I find it so much easier to get better colour in less time because its a smaller colour space but still much bigger than REc709.
 
The best cure for plastic faces is the S-curve.


a7s-face-1.jpg

a7s-face-2.jpg

a7s-face-3.jpg

original

color removed

S-curved

Disclaimers:
  • The "original" I'm working from is an 8-bit, 4:2:0 screenshot of a highly compressed Youtube video of a still that has already been graded once by the saboteur.
  • As has been said, the lighting and make-up are so surreal that I cheated and took away the color. The plasticky look remains, though. The plastic or video look usually is not from the color but from contrast. Take a plasticky, videotic-looking picture, and change it to black and white. Voila! The video look remains.
  • My cure-all S-curve was hastily done, again from a limited original. So go by the gist of what you see, not by the perfection of it.

The S curve is a two-syllable phrase for "raising the contrast of the mids while lowering the contrast of the highs and lows." How you do it in your editor may vary. I did these in The GIMP. Regardless, to my eye at least, the S curve always makes it look more like film. The final image may not be great, but I think it does look a bit more like an old 1950s glamour shot --- which was shot on what? Film!

Here are a few more S-curved still frames. Again my "originals" that I was working from were 8-bit JPEGs scavenged from the Internet, stills from soap operas.

1.jpg2.jpg
3.jpg4.jpg
5.jpg6.jpg

Again, the ones on the right may still be bad, but at least they're better, in my opinion. With the higher-contrast mids, the faces cut through to a greater degree. Meanwhile, with the lower-contrast highs, highlights are softened.
 
Last edited:
nyvz said:
I've heard this before about overexposing skintones in SLOG2 causing problems, but I've also seen tests and shot lots of material myself that indicates the opposite. Could you site any sources for your claim? Or perhaps offer an explanation? My understanding of SLOG2 is that because it's slope remains relatively constant (without knee, minimal shoulder) it is quite easy to apply a log offset operation to footage to simply increase exposure in a way that does not significantly affect color.
Click to expand...

Sony has been pretty clear about how to expose using SLOG in its various forms. There is lots here in this Sony forum on DVXUser on the subject. Use search and poke around. There is also Dennis Hingsberg's excellent distillation of the subject on his blog. Look here...

http://www.hingsberg.com/index.php/2014/04/f3-f5-f55-gamma-curves/

He and others have also written about it here. Again, check the archives.

EDIT:

When I mentioned above "Sony forum" to get information on Slog I should have been more specific. The majority of the commentary on the use of Slog can be found in the Sony forum here that focuses on the F series cameras, the F3, F5, F55 and for now the FS7. You can find it here...

http://www.dvxuser.com/V6/forumdisplay.php?219-PMW-Cameras-(-amp-temp-FS7)
 
Last edited:
I hope that folks won't be angered by my comments - a pseudo-amateur here - but A7s and GH4 clips shot via Atomos Shogun seem not to have the Sony or Panasonic skin tones to me. I realize that it's a clean HDMI signal out of those cameras that is recorded in ProRes but I find the final images far more pleasing than a majority of the FS7 and A7s clips that I have seen in the past even if everything is distorted along the way somehow.
 
Great post combatentropy!

Regarding S-Gamut: while I generally get the best results with the A7S not using S-Gamut with Slog2 (instead Pro, Color, Rec709, etc.), in my tests using various LUTs (Alister's, Sony's, Resolve's), S-Gamut can look decent. Grading S-Gamut without a LUT which first converts to Rec709 (or ACES, P3, etc.) is more time consuming. Clearly if Sony added Slog3/SGamut3.cine to the A7S it would be very useful (looks currently reserved for their pro-level cameras). Perhaps given how aggressive Samsung has been with the NX1 in quickly providing new features based on customer feedback, Sony could be motivated to do the same with the A7S (or successor(s)).
 
DPStewart said:
Wow, yeah. My sentiments exactly. Our eyeballs must have been manufactured at the same facility.

That is EXACTLY what I instantly noticed about the Sony A6000 - "Damned if this thing don't look EXACTLY like what's coming out of the Nikon D750!"
Of course Nikon is putting a Sony sensor in the D750. Also in Stills Mode, the A6000 does NOT look like the A7s to me, and I prefer the A6000.

I guess cameras are like partners - you gotta pick the one you like the look of.
(snark)
Click to expand...

Ha! :)

That's interesting. I bought the A6000 because I wanted to get acquainted with Sony cameras as I'm thinking to buy the A7sii.
You're saying that the A6000 is more pleasing than the A7s for stills which makes me think it could be a market move made on purpose by Sony, that btw would make sense.
A6000 is in fact a tremendous still camera whereas the A7s is some of the best choices for video (in that range and above).
Nonetheless, the idea of a FF camera is still quite appealing for stills as well :)
 
nyvz said:
I've heard this before about overexposing skintones in SLOG2 causing problems, but I've also seen tests and shot lots of material myself that indicates the opposite. Could you site any sources for your claim? Or perhaps offer an explanation? My understanding of SLOG2 is that because it's slope remains relatively constant (without knee, minimal shoulder) it is quite easy to apply a log offset operation to footage to simply increase exposure in a way that does not significantly affect color.
Click to expand...

Yes, just refer to the Sony white papers to see where to map middle grey. Slog, Slog2, and Slog3 are well documented.

There's nice write up on the Abel Cine website if you don't have the Sony papers.
Refer to this diagram of the slope and you'll note it's not constant at all which is why you need to apply a mathematicaaly correct reverse curve:
http://blog.abelcine.com/2013/01/18/sonys-s-log2-and-dynamic-range-percentages/
 
combatentropy said:
The best cure for plastic faces is the S-curve.
...​
Again, the ones on the right may still be bad, but at least they're better, in my opinion. With the higher-contrast mids, the faces cut through to a greater degree. Meanwhile, with the lower-contrast highs, highlights are softened.
Click to expand...


do mine :) (skin exposed around 65 on both, give or take) right out of the shogun (slog2, sgamut):
SHOGUN_S001_S001_T006.00_00_05_21.jpgSHOGUN_S001_S001_T006.00_00_19_14.jpg

tiffs for editing:
https://drive.google.com/file/d/0BwO55ILG1OL-bTZxOVBENDNfTzg/view?usp=sharing
https://drive.google.com/file/d/0BwO55ILG1OL-N2RzV2pGQW5DUDA/view?usp=sharing

slog2 coming out of the shogun is a bit different (crushed) from what I've read compared to the camera, sp that might make it handle differently than it would with in camera slog2.
 
The s-log2 over hdmi is full range where s-log2 from the camera is video range. It looks crushed since it is being treated as video range so the blacks and whites are being pushed out of range. The data is still there it just needs to be brought back in range.
 
Okay, i've skimmed the replies but here's IMHO.

I'm using Lightroom for stills, as default it is awful with A7s files, there are a number of custom created ICC profiles out there that are much much better than Adobes default. I hated skin from the A7s in Lightroom until i realised this. If you have a macbeth and time you can use adobes profile creator to make your own profiles, dual illuminant ones too (i made a load of these when the O7Q first came out) or google A7s profiles for lightroom. Your world will suddenly look right.

So here's the rub, the basic colour science of Sony RAW is good. But not all apps are supporting it correctly.

Secondly the sony red they use is out of gamut in a lot of post software (anything in sRGB), this is true of the raw images and also of slog too. You have to be careful in post to ensure a workflow that uses colour primaries that can hold the sony colour properly. Resolve, for example, can't. It will crush/clip/scale that red in ways that make skin look horrible. I've ended up with a number of methods, some in beta but others through post software like nuke that allows me to manipulate colour spaces properly. I keep meaning to write this up because it caused me months of pain. When you compare a Resolve render with one done right it's like fresh air (i'm mainly referring to FS700 RAW here). Having said that there are also plenty of other issues with the A7s colour and various profiles.

Thirdly looking at that video:

- Is that light any good? Most LEDS are awful for skin. Remote phosphor is awesome and the best little light at the moment is the Aladdin A-Light - very very good colour rendition
- The make up looks, well, a bit over the top - that can't helping?
- It's a false environment anyway, that may well be the objective.

But there are a number of caveats when working with this (and many other) cameras!

cheers
Paul
 
jcs said:
Great post combatentropy!

Regarding S-Gamut: while I generally get the best results with the A7S not using S-Gamut with Slog2 (instead Pro, Color, Rec709, etc.), in my tests using various LUTs (Alister's, Sony's, Resolve's), S-Gamut can look decent. Grading S-Gamut without a LUT which first converts to Rec709 (or ACES, P3, etc.) is more time consuming. Clearly if Sony added Slog3/SGamut3.cine to the A7S it would be very useful (looks currently reserved for their pro-level cameras). Perhaps given how aggressive Samsung has been with the NX1 in quickly providing new features based on customer feedback, Sony could be motivated to do the same with the A7S (or successor(s)).
Click to expand...

A log file needs delogging before grading. You never grade from a log file, it's a way to transport scene linear information. This is true both of the gamma and the colorspace. So as you observe SGamut is fine so long as you LUT it back into a colorspace to grade with. There should be no need to grade *from* a log file, you're just guessing - whereas there are tested and quantified methods to get that into something very gradable for you...

cheers
Paul
 
You must log in or register to reply here.
Back
Top