Why shoot with low saturation for grading in post?

NickHope wrote on 6/16/2014, 5:10 AM
I'm playing around with settings on my GH4.

Can someone explain to me why you would want to shoot with low saturation for grading in post? Most GH4 users seem to be turning it right down if they're going to grade in post, but surely that just means you've got less colour data to work with. Wouldn't it be better to shoot with more saturation and reduce it in post if required? I must be missing something.

I'm talking just about saturation here. Not luminance.


royfphoto wrote on 6/16/2014, 7:28 AM
The answer spans about 2 days worth of internet reading .One school of thought is that by making the image as flat as possible one can increase the dynamic range of capture, the "big boy" cameras shoot in a log (very flat) palate. The other school believes that cameras that shoot in log are 10 bits or more and as such have much more information to throw away when you wrestle the image back to something that looks good, the 8 bit produced by the GH4 does not have enough information to suffer the same transformation without side effects (banding and noise).
Both schools have ardent defenders and mathematical "proof". I take the middle ground: Cine D slightly flattening the image -1sat,-1contrast, -3 NR &Sharpen...not the -5 everything some propose. I'm sure some "expert" will point out what a blundering fool I am.
dlion wrote on 6/16/2014, 7:37 AM
i have a T5i and i shoot with the technicolor cinestyle preset. i recently shot a party at a catering hall - not a lot of light even when they were up full, and i was shooting black folks, and i was wide open.

just working with magic bullet looks, i was shocked at how much i was able to do with the original image. i can't speak to the tech details, but this would have been a nightmare if i shot with, say, the camera's indoor setting instead.
farss wrote on 6/16/2014, 8:05 AM
My best guess is to reduce the chances of clipping a channel. That said I certainly wouldn't advise reducing saturation much from the default as that'd increase the risk of banding.


musicvid10 wrote on 6/16/2014, 8:25 AM
By capturing as many colors as possible, a grader has the most control in post, where inevitably something gets sacrificed for the sake of viewability.

That said, I've found it a time-consuming process, and despite having more control, not substantially better for a non-artist than shooting 2.2 gamma.
royfphoto wrote on 6/16/2014, 8:34 AM
The fellow that did this video is the same person that worked out most of the advanced hacks for the GH2, so he knows his way around the GH series, check out his settings:http://vimeo.com/90794067
wwjd wrote on 6/16/2014, 8:49 AM
I too with gh4 am not reducing saturation in 8-bits. it needs all the help it can get. :)
Rainer wrote on 6/16/2014, 4:56 PM
My thinking is always shooting flat was a fad introduced largely by people who sold "looks", and eagerly adopted by those who think playing around in post equates to professionalism. Yes, shooting flat does gain some range, even with 8 bit. But there's nothing for nothing - it's at the cost of resolution. Shooting flat costs you fine detail - please don't take my word for it, it's easy enough to test for yourself, you don't need a test chart, just shoot some foliage and do some pixel peeping. It might not be a bad thing. The important thing is to be aware of what you want and to know what the tradeoffs are. The dichotomy seems to be between those who say "I always shoot flat, it gives me most flexibility in post" and those who better understand the results they want. Real life situations - you might want more range and less fine detail for weddings, you might want more more saturation for underwater (in your case Nick, just keep doing what you've been doing, we all bow before you).
wwjd wrote on 6/16/2014, 7:46 PM
I can see the point of "flat" when the scene has lots of dark shadows AND lots of high brights.... reeling them both down to a more equal level, they hyping in post, gives you a more evenly "lit" scene while viewing - maybe closer to HDR stuff but more natural looking. Also to condense 24 stops of human viewable range into 16 or 14 that a camera can handle... it makes sense. Yet, with the 8 bit limits of current internal GH4, we don't want to flatten down too far, because expanding back out to look natural will BAND like a mother.

On my older camera, I shot FLATTER than it did normally, but no where near LOG.
Maybe flat does provide more flexibility in post, but 8 bit cameras have noticealbe limits and we must work to avoid it being noticed.
Former user wrote on 6/16/2014, 7:52 PM
the colorist that I worked with always complained about this. He said he spent so much of the range of the DaVinci to get the color back to natural that he had nothing left to give it a "look".
Serena Steuart wrote on 6/16/2014, 8:42 PM
Not having a GH4 I can speak only from my own knowledge of other cameras. I wonder whether there is confusion between "flat" and "desaturated"? Images recorded with a flat gamma curve (especially log) are low contrast and appear desaturated, but this isn't because saturation has been reduced. The aim is to capture a wide luminance range and 8 bit cameras have a very limited dynamic range on "normal" settings. Flattening the curve and rolling off highlights enables capturing detail (e.g. white clouds) that otherwise would be clipped. In post you "reverse" the curve to restore normal contrast while being able to control how highlights are rendered. The nature of the scene will determine which curve is best for recording, the aim being to record all the data needed for the final product. I must add that I've only adjusted saturation in camera to correct for colour bias and I'm trying to think of a situation where reducing saturation would gain data.
musicvid10 wrote on 6/16/2014, 11:56 PM
When the scene lighting is itself a bit flat (diffuse), flat-gamma, high-bit acquisition can bring some amazing things to life for the colorist. Graphic artists and Photoshoppers have known this for a long time.

With our average sunlit fare though, the source highlights are already chopped, giving little advantage to having those added bits.
Laurence wrote on 6/17/2014, 1:14 AM
I've keyed out the blown out sky and replaced it with a blue gradient. Looked great!
Rory Cooper wrote on 6/17/2014, 2:17 AM
How I understand it is that the RGB in REC 709 and LOG is the same but the luminance range in LOG is bigger so since the RGB pixels sits on top of the luminance pixel in the colour space therefore a larger luminance range is also a larger chrominance range.
So luminance and chrominance are connected in the color space = if I have a red square the more black gradient or white gradient i add to that red square the larger range of red I will get. So the red is red you can’t get more red = so when you saturate reds you limit the luminance.
Serena Steuart wrote on 6/17/2014, 8:11 AM
>>>With our average sunlit fare though, the source highlights are already chopped, giving little advantage to having those added bits.<<<

I might not be clear about your meaning, but the point of extending dynamic range is to avoid "chopping" highlights or shadows. Especially in sunlit scenes. You appear to be saying that clipping is unavoidable in outdoor scenes, which is incorrect. In an interior scene you do it with lighting, so then clipping isn't a problem anyway.
Rory Cooper wrote on 6/17/2014, 8:23 AM
Well Serena that is more to the practical application of the discussion, well to me anyway. If I said shooting with Natural Density filters is the same as shooting LOG, = ND is on the lens and LOG is on the sensor/encoding. would that be correct in principle?
wwjd wrote on 6/17/2014, 8:41 AM
and how does the GH4 compression factor in to all this? If one shoots flattest or reduced saturation (understanding those two are different things), does that help or hinder the compression going on?
larry-peter wrote on 6/17/2014, 9:07 AM
Does the GH4 have the same types of scene file options as the larger form factor Panasonic cameras? If you have options for different color matrices for each gamma setting, those are well worth experimenting with to make grading easier.

On the larger cameras, the Cine-D gamma with the Norm2 matrix is a breeze to grade. Norm2 has much less noise than the default Cine-like matrix, looks more balanced in saturation (to my eye) and gives you a lot more latitude in skin exposure. Cine-like really starts to look bad when caucasian skin is exposed over 50IRE, Norm2 can get you 60 before it falls apart.
musicvid10 wrote on 6/17/2014, 9:42 AM
"You appear to be saying that clipping is unavoidable in outdoor scenes, "

Not "unavoidable," but common; and nevertheless the added bits are often not that much more usable under normal 8-bit / 2.2 Gamma viewing conditions without putting in an awful lot of work.

With precision grading, the criteria is not as much a matter of where the clipping occurs, but that it occurs at all. Any source image pixels locked at 0 or 255 are toxic, so acquiring full subject / scene range at [1, 254] is the theoretical ideal for this kind of work. That is simply difficult to contain when shooting outside the studio (my experience is not with a GH4, however).

Many real-world shooting situations simply do not allow us the kind of control that would take full advantage of flat-gamma acquisition. The result is that post-production becomes incredibly time-consuming and possibly fruitless for the average editor, having been undertaken in the presence of other, sometimes more limiting conditions.

Just another point of view, you should understand. I'm not arguing the merits of the technology, just whether it is worth all the time and trouble needed to see limited practical benefits for the consumer in 2014.
farss wrote on 6/17/2014, 2:42 PM
If shooting outdoors or when the outdoors is making an uninvited intrusion indoors using a non linear gamma curve can help retain highlight detail. By default all the Sony Cinealta cameras have "Autoknee" or "Dynamic Contrast Control" On. Using that I've not found any need to do extensive grading in post, the camera does it and it's doing it with 14bpc to play with.
The only serious caveat is one does absolutely need to expose for skin, having dual or adjustable zebras really helps here. If you do push skin tones beyond 70 IRE when the autoknee has kicked in they'll start to look plastic, yuck.

I also find this functionality very useful for stage although again one needs more care with exposure.

On the other hand using the fixed cinegamma curves one might wish to do some grading in post but I find a bit of a play around with a fixed colour curve gets the job done.

Serena Steuart wrote on 6/17/2014, 7:46 PM
A simple ND filter on the lens is not the equivalent of shooting log on the sensor because the ND filter just cuts uniformly. However you can use a graded ND filter for effective control of highlights when using a "standard" gamma curve, such as to reduce the brightness of the sky (especially clouds) while leaving the rest of the scene untouched.
Using a flatter gamma curve (that maintains lowlight contrast while rolling off above around 50%) doesn't involve extensive grading in post. Just as Farss says, you just apply colour curves to generate a curve inverse to that used in recording. Of course not exactly inverse or you blow out the highlights in post. I'd be surprised if anyone is not doing some grading, at least to match clips through a scene, so this isn't exactly technically arduous.
The gamma curves available in 8 bit cameras are not log because that requires at least 10bit recording, but you can do a great deal better than using standard gammas. This can be illustrated by the gamma curves I measured for my Sony XDCAM EX1, which records at 8 bits. https://dl.dropboxusercontent.com/u/60757494/gamma.pdfEX1 gamma curves[/link]

The EX1 has also 4 "standard" curves that vary slightly from "standard 1" and they are not shown on this plot. You can see how the cine gamma curves extend the range of brightness that can be recorded before clipping occurs at 109% (255) and you'll realise that making use of the full sensor range requires grading to at least bring images into the usual 16-235 range for DVD.
As with all techniques nothing is a universal solution to all shooting situations and if you have a scene that requires good highlight contrast while holding shadows then you need a different approach (probably involving lighting control).

Yes, avoiding clipping in sunlit scenes does involve technique but I think it well within the knowledge and capabilities of people on this forum. However maybe not everyone minds patches of pure white in their images. You see it often enough in TV news coverage.
musicvid10 wrote on 6/17/2014, 9:36 PM
"However maybe not everyone minds patches of pure white in their images. You see it often enough in TV news coverage. "

Hehe, that conjures up an image of a mobile editor asking for more time to get news footage on the air so he can "get some more details in the clouds."

Truth is the only way we're going to eliminate bald spots and maintain acceptable daylit subject contrast is by building curves and masks for each scene.
Now that is time-consuming. It's going to be a while yet before mainstream consumers routinely view 10-bit in their homes.
Serena Steuart wrote on 6/18/2014, 1:24 AM
>>> routinely view 10-bit in their homes.<<<

If the display systems were 10 or 12 bit there wouldn't be such a problem. What we're talking about is compressing a real world scene into the range of current display systems. I did raise the TV news but hardly anticipated you to think I was advocating theatrical grading of their output. The original question was why use cine-style gamma curves, but by all means blow out the highlights in your own material.
farss wrote on 6/18/2014, 5:49 AM
In fairness what's going with cinegamma and auto knee / DCC etc. isn't all that easy to grasp, I know it had me confounded for a while

What is going on inside the camera is the same as what we'd do if we fixed the problem with lighting. If we could we'd light to avoid highlights that refuse to fit within a well exposed shot. Because the sensor in the better cameras can capture a larger range of light levels without clipping then inside the camera the highlights can be toned down so they fit inside the more restricted dynamic range of 8bpc and typical displays. The result for me at least isn't an overtly "flat" image at all anymore than a well lit scene is flat.

I should mention that these gamma curves are not the same as SLog or RAW recorded with 10bpc or more. What Slog does is try to capture all the dynamic range intact. When that's done and it's simplistically displayed on a monitor it will look flat and in need of grading. Cinegamma on the other tries to do what the name implies, emulate the knee in film. One can grade it look more video-ish or leave it alone.

musicvid10 wrote on 6/18/2014, 8:18 AM
"What we're talking about is compressing a real world scene into the range of current display systems."
Yup. With the emphasis on the subject. And we both like clouds.

"I did raise the TV news but hardly anticipated you to think I was advocating theatrical grading of their output."
Your example evoked a funny image, that's all.

"The original question was why use cine-style gamma curves, but by all means blow out the highlights in your own material. "
Hi Serena, we're on the same side here, and sharing different perspectives.
Maybe lighten up?