RedcineX color managed?

[Stepan Kment]

I am really missing something here.

Normally, the app has its own color working space, usually very wide to ensure no clipping occurs during the processing. It performs on fly conversion to the display color/gamma space, which is represented by the ICC profile assigned to the display by color measurement device. This way, one always sees right colors and luminance levels.

Then export. One has to fit more data from wide internal processing gamut to less bits of ordinary files. That's why there exist curves like panalog, PDlog... and Redlog, I guess. These curves are designed to store information in shadows more than in hili, which is desirable. Simply compress more bits to less so that the damage is minimized for human eye. Additionally, color space, that is primaries, can be set, too, this tells, when say RGB is 1.0, 0, 0, fully red, which xy coordinates in CIE chart resulting color has.

This works in Nuke, After effects and so on. More or less nice, but it is there. Beautiful EXR file even have primaries' xy coords in its header and gamma is always 1, linear. Very easy it is, color manages itself automatically everytime, it almost can't be -not- managed at all.

HOWEVER: in Redcine, I don't get it. I don't know how to calibrate my LCD, what Redcine expects. And why color space and gamma immediately changes the picture - such a curve is a matter of export everywhere, so what happens here, why it changes the preview image immediately?? This works everywhere else, eg. with Genesis it is completely understandable and easy, why not with RED?

Really - I know I am annoying asking again and again, how RED tools and import dialog boxes handle color, but never got answer. Should I color calibrate my LCD to sRGB... adobeRGB or what and gamma to 1.7, 2.2 or 2.4 to ensure whites I see in Redcine's UI are the same I'll see in any other application, which is color managed and adjusts its output according to display's ICC profile? Whole grading is MAINLY about color managed environment and these Baselights, Lustres, Nukes and other tools, which work as expected, simply support it. Even in after effects, which is a low end toy compared to these, one simply assigns every file its color profile and so AE knows, what RGB values in the bitmap actually mean, what primaries and gamma are and how to convert them to a project working space.
--------
And another question:

RED records RAW data as we all know and I believe 12 bits from AD converters are directly stored in RAW data stream (OK, compressed, but this compression doesn't alter bit depth, just packs data with wavelets). Why, however, it stores information about colorspace and gamma set in camera during recording in r3d metadata? Does it mean internal camera setting somehow alters the r3d data being stored? For beginners, this differs from white balance, for instance, completely different thing - we're storing RAW, which simply should be direct CMOS readout, thus color primaries are these of the chip itself. Color cube can be later rotated so that primaries may be changed, of course.
I'd understand, that it actually maybe isn't a colorspace and gamma, rather it may be a sort of LOOK. That is, when DoP likes what he or she sees on the camera output, then it may come in handy to store info about how camera debayered RAW itself during shooting and to restore this in Redcine... so is this actually a visual look rather than real color space or???

is there anyone who -knows-? Thanks in advance!

 

[David Battistella]

Stepan,

You've asked a total aof nine questions here.

What is the primary concern?

David

 

[Stepan Kment]

primary problem

primary problem

Stepan,
What is the primary concern?
David

Hi David,

yes, the problematic is very uneasy and anyone without good knowledge of color management techniques will be completely lost in it.

1/ the problem is how to set the display.

It may be the case Redcine doesn't support any calibration at all, resulting in worst case scenario, when one simply sets colors to their liking, however totally depends on current display settings, which may be absolutely out of reasonable standards.

RedcineX has some very nice first light grading toolset, however imagine grading with some blue cooling filter in front of your display - would be very bad idea, right? But without color management, this is exactly what You may be doing!

Fyi - changes needed to default settings of -most- LCDs to achieve some default target, eg. 6500K, 2.2 gamma, are nothing but drastic - forget about default color presets displays offer, these are totally off. I calibrated many LCDs and sometimes couldn't believe my eyes, how the image changed after calibration.
-----------
2/ Plus some possible inconsistencies?? in terminology RED uses which are hard to explain when competing to Genesis, D21 etc. and which lead directly to many misunderstandings around this camera on the internet and which influence understanding of the color processing again. I'll rather stop typing here, it'd result in another endless post :bored:

 

[M Most]

RedcineX has some very nice first light grading toolset, however imagine grading with some blue cooling filter in front of your display - would be very bad idea, right? But without color management, this is exactly what You may be doing!

That's part of what scopes and histograms are for. Scopes and histograms don't lie, even if your eyes and your picture monitor do. All three should generally agree. If they don't, either your monitor is inaccurate, or you need to let your eyes adjust for a minute or two. But if scopes say the image is blue, it's blue, regardless of what you think you're looking at. Colorists know not to just walk into a room, look at a monitor, and believe what they're seeing to be accurate. They rely on setup procedures, but they also rely on scopes to keep everything honest.

 

[Paul Leeming]

I believe RedCine-X is set to display correctly in Rec.709 but hopefully someone from the Red team can verify this for sure.

HTH

Paul

 

[Jay Gannon]

We calibrate to REC709, seems to work for us.
But as said above real scopes are your friends.

 

[Stepan Kment]

But if scopes say the image is blue, it's blue, regardless of what you think you're looking at. Colorists know not to just walk into a room, look at a monitor, and believe what they're seeing to be accurate. They rely on setup procedures, but they also rely on scopes to keep everything honest.

Correct, however still it means one knows what these scopes show. If primaries are not converted to different colorspace, then they represent in fact what CMOS chip records, that means how color filters of bayer matrix are designed. I don't think this is documented anywhere oficially, never found it.

That means one can depend on scopes only with knowledge in which color space (primaries, white temp) the processing takes place and what gamma is shown.

Only relatively good observation is - if one calibrates their display to some color target and keeps it during whole grading of the project in R-X, then the result will be at least consistent. However it will be most probably necessary to perform a fix applied to whole timeline before output to final medium when it is displayed in color managed environment - most probably, whole movie will be too bright, dark, yellowish... anything may happen. If one is lucky, then this won't be drastic.

Only exception may be external calibrated Rec709 HD display attached to Rrocket, this may work fine, however I don't see a way to calibrate ordinary PC display to such a target or any other, like film stock, where one grades to ~5500K instead of 6500K.
-------------
http://prolost.com/blog/2008/11/7/what-should-red-do.html

Just read post of Stu, I respect him and only have to agree (though a lot of issues were resolved from the date of his posting) - worked with Genesis and it is quite easy to dig the same look of the recorded image as seen on place. With RED, it certainly is different, RED is more flexible, but to achieve the same result as recorded isn't that easy.

Simply said - calling here for how to guide. We know ISO320 is preffered, we know not to overexpose, however it'd be very welcome to know what to set where in postprocessing to get optimal colors consistent with original look of the recorded scene when recording outdoors, 5600K and 3200K indoor. How color space settings in the camera influence postprocessing, if at all. And so on.

 

[jake blackstone]

Why, however, it stores information about colorspace and gamma set in camera during recording in r3d metadata? Does it mean internal camera setting somehow alters the r3d data being stored? For beginners, this differs from white balance, for instance, completely different thing - we're storing RAW, which simply should be direct CMOS readout, thus color primaries are these of the chip itself. Color cube can be later rotated so that primaries may be changed, of course.
I'd understand, that it actually maybe isn't a colorspace and gamma, rather it may be a sort of LOOK. That is, when DoP likes what he or she sees on the camera output, then it may come in handy to store info about how camera debayered RAW itself during shooting and to restore this in Redcine... so is this actually a visual look rather than real color space or???

is there anyone who -knows-? Thanks in advance!

RAW information doesn't get altered in any way during recording. That is the definition of RAW. The info on camera gamma and color space is recorded only as metadata in a separate file. As a result, because it is only metadata, the gamma and color space can be changed during the post production process, This way the choice of gamma and color space can be made later, depending if the material needs to be developed for the video or film finish, ie linear or log color space. Effectively, this approach is a more flexible and filmlike.

 

[David Battistella]

Going back to the original question(s).

I think that Stepan was asking if there is a reccomended GAMMA and COLOR setting for the monitor when using REDCINE-X.

I can understand why people get confused because they are creating motion media for broadcast applications with a hybrid DSLR/BROADCAST viewing system.

It's like REDCINE-X needs a "broadcast mode" with scopes that people are used to rather than just a Histogram as the way to monitor images.

The problem is also exacerbated by the problems with how rendered outputs look.

I can understand why there are questions about what we are "actually" looking at when looking at r3d in the viewer window.

It's really just a color space provided by RED.

David

 

[Peter Dimov]

Going back to the original question(s).

I think that Stepan was asking if there is a reccomended GAMMA and COLOR setting for the monitor when using REDCINE-X.

I believe that what he is really asking is what is the meaning of the various color spaces and gamma curves provided by RED. That is, what are the primaries and the white points of the color spaces, and what are the formulas of the curves. We know (more or less) what the curves mean, but there is no information on the color spaces anywhere (I know of).

There has been a recent thread in which it was said that the different sensor batches give different colors on the same scene, so the sensor primaries may well vary with the batch - in which case it may not be possible to document them.

(Scopes do not help. Scopes tell you that the color is (0.1, 0.1, 0.8), but they can't tell you what specific shade of blue do these values represent; you need to have the CIE primaries and the white point for that.)

 

[David Battistella]

Stepan,

Can you confirm any of this?

Peter,

You are talking about a LUT for each color space combo?

david

 

[Paul Leeming]

...however I don't see a way to calibrate ordinary PC display to such a target or any other, like film stock, where one grades to ~5500K instead of 6500K.

Stepan, the Colorvision Spyder (Pro version) does allow you to create a Rec.709 colour calibrated workspace. Not sure about other brands.

HTH

Paul

 

[Peter Dimov]

Well, you could use a 3D LUT to go from one color space to another, but it's usually done with a matrix. The point is that, if accurate color reproduction is the goal, the RGB triplets in the source image need to have a color profile associated with them. This color profile describes what 100% red, 100% green, 100% blue and 100% white source signals mean as colors.

In principle, exporting an R3D into Rec709 should produce an image that uses the Rec709 primaries and white point, so one possible workflow is to calibrate and export to Rec709. But then there's also REDcolor, which also targets Rec709 monitors.

This is probably a non-issue most of the time, since the footage is graded to conform to a specific look, so it doesn't really matter what the original colors are.

 

[Stepan Kment]

LUTs...

LUTs...

Hi David,

Peter is correct and I agree with him. Still, the more we all are thinking of this, more and more questions pop up.

What I mean above all is:
We can see RAW as "something recorded" and once imported in a grading system, one may try to recall, how the scene originally looked and try to adjust settings so that developed r3d corresponds to the place recorded on the camera. If there is no reference or a grader wasn't present on the place, they would never know, what to try to reach. For genesis, I have to say they have very clear guidelines and if one complies with them, he or she gets really much the same look of the video as the original scene. With RED, I find this more difficult. The camera is very flexible, which is both good and bad.
--------
Is this important for You? How do You solve this when grader wasn't present on a place and has no idea on how colors looked like originally? Everyone on place takes care on lighting etc., I'd expect they first want to see original look of the place being recorded in post, then may ask for some creative tweaking... or they don't mind and if the result is nice, they don't care it looks different from how it was shot?
--------------
1/ primaries of CMOSes differ batch by batch.
2/ to get consistent results from more cameras (eg stereorig), You need at least one calibrated environment somewhere in the pipeline. Note that displays drift over time, too, so HW calibration is essential to get consistent results. I believe more people ran in trouble and discussed this on reduser...
3/ to load images from R-X into any system, one needs some interpretation rules. Primaries, info on gamma. And LUTs for certain systems. There are workarounds, I don't want to make this overly complicated here.
4/ afterFX, Nuke and other systems work or may work in linear light. If there was a consistent way to get linear light data into these systems with minimal losses, it'd be great. To allow for this, we must know what R-X does with colors and gamma, how DPX luma is mapped when exporting from it, and also LUTs (and ICC profile for AFX) are needed to correctly interpret the footage in there. If R3D is linear light, then OK, output to afterFX etc. in linear, this is just fine. But say, how it works, so that one'd know, what to expect

Sure, with some expensive software for building LUTs and after significant time spent recording various test charts, one would probably end up having things resolved. I don't have either of this.
----------
Just to demo the situation in CS4 on Win: one starts color management in afterFX - this immediately alters what one sees, colors get more or less shifted, that is how color management works. (The comp view gets color managed, ensuring consistent results are provided.)

You run r3d dialog to adjust white balance etc, however it is actually different dialog, which is *not* color managed. That means you set white balance etc in that dialog, then put the asset to the color managed comp and see something different(!!) and have to tweak colors again! And You are still just in one program, not switching among many. Really it is that bad. And even worse, import dialog clips overbrights for some reason. So one develops the image, then, without any levels, zebra etc. in the dialog has to lower exposure not to clip anything, import result to a comp and adjust exposure back again. Most of you will agree this is ...weird. I did this on whole feature film shots, really :banghead: ... and whole r3d import dialog can't be skipped as there may be bugs in metadata and once white balance is set incorrectly, it is too bad.

 

[jake blackstone]

Hi David,

Peter is correct and I agree with him. Still, the more we all are thinking of this, more and more questions pop up.

What I mean above all is:
We can see RAW as "something recorded" and once imported in a grading system, one may try to recall, how the scene originally looked and try to adjust settings so that developed r3d corresponds to the place recorded on the camera. If there is no reference or a grader wasn't present on the place, they would never know, what to try to reach. For genesis, I have to say they have very clear guidelines and if one complies with them, he or she gets really much the same look of the video as the original scene. With RED, I find this more difficult. The camera is very flexible, which is both good and bad.

Stepan.
You keep asking the same question over and over. There is nothing inherently wrong with the way Red records information, as opposed to Genesis. How do colorists balance their picture in telecine with film? The answer is simple, just shoot the grey chart and use it in post. Use your primary RGB control to achieve the correct color balance, right after the color temperature adjustment. If you want to use filters for a specific look, then again, shoot the grey scale with no filters and then use them after the grey scale is shot. You just over thinking it. Calibrate your monitor to Rec 709 for video deliverables or P3 for film and grade to taste.

 

[Andrew Ravani]

for consistency, start with the shot

for consistency, start with the shot

Hi - I have had the greatest results when I have had my head screwed on straight and remembered to shoot a color chip chart for each set up as well as the slate. I can take that snapshot tiff image into photoshop/nuke/aftereffects or anything else and grade it as a reference - one from the chip chart and one frame from the scene. Two frames per set up. Done.

That image gets tossed into the mag file and lives with the original R3Ds as a reference. After that, its a matter of suite calibration and the colorist getting it right (which is often better than the photoshop tiff that I give them, thankfully - cause I sux at photoshop).

~Drew