I've read an article by someone with both high-end credentials in video post color management and obviously connections in Cupertino. His comments were carefully worded, I'm sure to avoid "conifidential" or NDA restrictions.
The essence was that "perhaps" the Apple people looked at the original Rec.709 specs from the late 80's or early 90's ... which were back when 'we' were all using CRT screens ... and decided that was really the "correct" Rec.709, not the addendum that came in Bt.1886 a bit more than a decace ago. Let's look at how we got "here".
The "original" Rec.709 did only include the camera transform function, roughly gamma 1.96. This was to match the image coming from digital-linear of the camera sensor, to match the native 'curve' in shadows of the CRT tubes. That way, digital images were able to be treated the same in video post as film created images.
But then, flat-screen digital screens were created, though many still worked with CRTs for some years. Both type screens were in use. And the digital "flat" screen tech didn't have that shadow response/curve.
So the question was ... 1) drop the original camera transform, or 2) add the transform for digital display, or 3) change the camera transform?
The far simpler process was to simply do option 2. Add a transform function for digital displays to the then current system. That way the images worked fine without any other changes in the entire system. Whether on a CRT or a flat-screen.
Apparently, Apple's engineers decided the "scene-referred" gamma of the original Rec.709 was the 'correct' way to be working. On a theoretical basis, also "apparently". And are all heck-bent on sticking with the "correct scene referred" transform, internally insisting everyone else on the planet is doing it Wrong. And they will not join those doing it Wrong.
Whatever. And though this seems to many colorists I know to have some probable basis in reality, all of them Mac geeks, many with some small connections in Cupertino, it's all speculation. As Apple won't say anything publicly past a couple papers they've published with their operating specs. No explanation or justification.
So take the above as speculation. The rest of this is all "known" techy stuff for colorists, but the direct reasons Apple does what they do ... well, Apple will be Apple.
Even if you're trying to claim, however, that by using the original camera transform by itself is the "correct scene referred" image data, that original camera transform was only done to match CRT's native curve. So if you're trying to really be technically scene-referred, you should drop the 1.96 and just go linear to begin with. And stay linear to linear.
Ah well.
Now, I'll throw in another silly kink ... "full" range versus "legal" or limited range encoding. This one is quite possibly affecting your experience. Without being obvious or that easy to check.
Early digital to analog broadcasting systems needed 'headroom' to hold other signals as part of the file, that told the TV/broadcast systems how to behave. So as the files were then only 8-bit, and therefore 0-255, they set 0-15 and 236-254 as "out of bounds" and used only by control signal data. Meaning the image was to be recorded between 16-235.
But the image was the full image from the sensor, just encoded, in a mathematical formula, to 16-235. The display would automatically transfer the image to 0-255, full signal.
And no one needs that headroom anymore, so technically, we could all run Rec.709 as full range and be done with it. However ... "The System" was already in place. As YUV formats (Technically Y-Cb/Cr) were "legal/limited" and RGB formats (the high-bit 12 bit stuff) which weren't ever used in broadcastm, were "Full".
It doesn't change the data encoded to the file at all, just how it's encoded. Doesn't affect the range of values light to dark actually in the file, it's just a math thing. Which a lot of people ... myself included ... are fooled by at first.
So some systems set themselves correctly so that YUV "limited" formats are accepted as "limited" and expanded by the disply to "full". Some are unfortunately set to assume everything is Full, and those handle the few full range files correctly, but improperly display 'standard' Rec.709 YUV limited files only as data between levels 16-235.
You can probably see where that is an issue from computer and system and displays forth and back in and of itself. If someone's system is set so that everything is displayed as if full range ... standard Rec.709 limited files are displayed with very lifted shadows, weak brights, and lower apparent saturation.
Or the other thing, setting the monitor or system, so that all files are displayed as if there are limited ... means that any files that are correctly 'full' are shown with crushed blacks and clipped whites.
And we've had people on here who insisted that they were only going to use full range because they wanted all their data. They were convinced that there was a contrast range difference in 16-235 and 0-255 files for the brightness range the file could handle.
Which ... isn't correct. Same data can be recorded either full or legal, without losing a bit.
Sadly, that decision, in itself, caused them to lose data. That they were furious about.
Now, I'll throw in the next potential problem: viewing environment!
There's a reason colorist suites have very specific and rather draconian specs for room wall brightness and color, plus room lighting brightness and color. All this past the strict specs for the Rec.709 monitor itself. Because the viewing environment completely alters our perceptions for contrast, saturation, shadows, and whites.
So a colorist works in a very controlled darkened room, with a very low lighting level. The walls are specified as an incredibly neutral gray (and yes, there are color coeeficients specified for lighting meters to check), and ... the monitor is required to have a "bias light" mounted to the monitor or between it and the wall, lifting the brightness around that monitor a precise and yes measured amount.
(And yes, my office is fully and correctly setup according to specs. Because, well, it has to be to produce the materials I make for pro colorists.)
The monitor is carefully calibrated, a LUT generated either parked in the monitor itself or in the breakout device feeding that monitor, because they do not trust computer OSs and GPUs.
(This is the one thing I vary from "full" specs ... I still use my GPU to feed the monitor. BUT ... I've carefully calibrated and run profiles to check the calibration! ... and had my files checked by colorists on their systems for accuracy of hues/sat/brightnesses to verify things.)
The monitor is set to 100 nits pure white, D65, sRGB Primaries, gamma 2.4. And even then, they use outboard scopes normally as a further check of brightnesses and saturation.
Look at something in their suite, watch the same file on a phone on a park bench at noon on a sunny day, or on the same phone lying in your bed in a dark room at night. You will see three different images.
Right.
And even between monitors, if one is set to say around 230 nits in a moderately dark room, and another is up over 300 in a much brighter room, that file will have completely different 'feel' for the black level & saturation.
So between the Apple non-standard Rec.709, the potential issues of screens set or not set to full/legal levels correctly, the wide range in screen brightnesses and room/viewing brightness, pro colorists know there is only one guarantee for their carefully crafted work:
No one watching their work will ever, ever! ... see exactly what they saw in their suite. Period.
Not in a movie theater, not via streaming or broadcast, nor even via DVD/BluRay disc. Ever. Period.
That's ... image post processing.
Neil
9
Replies
AdChoices
