Comparison of CC workflow’s: Film Log vs. Linear REC709

[Frank Cueto]

Dunno, Obviously both approaches have benefits. I've done them both ways and both have turned out great. I dont think either approach is wrong or or worse than the other one. We are all probably arguing over the last 1-2% worth of quality at the most.
Our LAt flick (The Caller) was done in rec709. and all our deliverables (including film ones) passed QC with flying colors, not even one page worth of notes.

After all, didn't LightIron do Social Network on rec709 and it was nominated for a cinematography oscar, it cant be that bad? ;-)

[Luis Otero]

Frank,

On a side note, compartriota, love your reel posted on Vimeo. My beautiful Puerto Rico all over the place. Was the opening footage with the back with "peñones" from la playa de Vega Baja? I grew up there, and it is strikingly similar.

Great job, man.

[Evangelos Achillopoulos]

The LUT used on the film version is a film targeted, cineon to HD LUT that I used on a number of features while in Florida. My point - which you still seem to be completely missing - is that the natural tendency of a film targeted LUT is to push the colors into a film color pallette and a film greyscale, thus presenting the colorist with a film biased image as their starting point. That is the entire point of this conversation, if it can still be called that. A few years ago, when film distribution was still the only significant theatrical distribution, I did exactly what you're advocating simply because our clients wanted everything to look like film regardless of what it was shot on. That was a specific creative intent. But your constant harping about one image that can be used for multiple masters, and that for some reason that has to be film targeted and Cineon based, is something I just don't agree with. Any log original, whether from film or from Red or any other digital camera capable of log imagery, can serve as a multiple deliverable master. It doesn't need a film targeted LUT unless the deliverable happens to be a film print.

You either don't read what I write or you didn't understand what I propose, because you don't do exactly what I do...

I use either real Film Targets OR an artificial Cineon D65 target that DOESN'T introduce the artistic limitation of real film... because is just a mathematical model with no real color space underlying... Cineon D65 or D55 doesn't mute colours as film does because, I repeat, its artificial....

Cineon D65,55 are artificial targets within CineSpace... these color spaces are like IIF/ACES... NOT limited...

So what I propose is 1. Use a film target to have the aesthetics of Film or 2. Use an artificial Cineon D65 target to have all film mechanics (Log behaviour) but no color limits....

Next point

You have repeatedly challenged the ability to retime from a Rec709 master, but there are many ways to solve that problem without film targeted color spaces. One can just as easily grade from the log original using a P3 targeted viewing LUT over RedlogFilm original and accomplish exactly the same thing, allowing retargeting by switching to a Rec709 targeted output LUT and making any color modifications in log space prior to that LUT if one wanted to.

Yes everyone can do what you say.... BUT.... production then needs to keep the original project available along with all source footage (as well as software and hardware needed to operate the system) for future use in case of a new colour space or a new edit or even for a new distribution medium invented...

What I propose is, have a final master sequence (in a container that can store the equivalent of the originals RAW) and get along a set of LUT's and the original target and store them... in this way you will be able in the future to convert the same final to all desired targets regardless of software that will be used...

If the LUTs are preserved in a ASC format and the files in 10Bit Log DPX format, it means that the whole thing is looking like a TODAY available, "IIF/ACES" workflow...

And since everyone understands why an IIF/ACES is crucial for the industry, to supply to people, a similar solution today is as significant as having IIF/ACES available for everyone today...

You said for IIF/ACES

This is not very different in concept than today's DI processes, in which a 10 bit image container is used that is mapped to a log format, and LUTs are used to properly map the output to various deliverable formats.

Additionally I have made the following observation from various posts... (Mark that includes yours)

I have realize that you all do a fundamental mistake.

1. Every delivery format has an expected environment that it supposes to be used with. Film projection expects a dark room and certain brightnes level from screen which is vastly different from REC709 or sRGB, which expect dimmed back light in the room and x2-3 more luminosity from display to 80 - 120cd/m2 against 55cd/m2 of film. Even DCP can have almost 1/3 of a stop difference due to different spec'd tolerances...

2. All have different white points: DCP has a white point x = 0.3140, y = 0.3510, Film has x = 0.3320 and y = 0.3470, REC709 has x = 0.3127 y = 0.3290

Our eyes are dynamic analog "devices" with all settings in auto, like an ordinary cheap photographic machine... When our eyes are exposed to a certain white point they are auto balanced to that... also they auto adjusted to max luminosity levels... all that functionality is part of our perception of vision. Our brain tracks the color range that is exposed to and auto adjust everything limited to daylight range, which is 4900K to 6900K and 13 to 15 stops contrast range that our eyes are able to discriminate.

If you seat for 15 minutes in a Film theater your eyes will adjust and they will perceive as white the Film white point and as max luminosity the current max. In this way the whites will look white and blown whites as clipped...

Now if you do a DCP and a Film projection on the same time, your brain will have to solve a perceptual puzzle... because it has to see two slightly different thinks... here is where perceptual vision "illusions" kicks in. Our eyes will immediately adjust to the stronger elements of the images shown and auto adjust... that's brightness, contrast, saturation, blacks and white point... the resulted perception will be something related to the mix...

If you try to do the same with REC709 the results will be totally devastated... the two are so different that this kind of "butterfly" projection, is totally inappropriate for subjective viewing... since you try to compare apples with oranges...

So this kind of comparison is fundamentally wrong. I'm sorry but our eyes are not linear...

Mike also that is your problem with past conversions from film to REC709... the lack of white point adaptation of the new medium... And I also see it in your examples with Marcie... I will come again on that subject with real movies examples...

Honestly, I don't like having my integrity challenged and my motives questioned. I have no motives in any of this. And if I did, the only motive I would have is to present all sides of a story that has multiple sides. You're not going to convince me that somehow you're right and I'm wrong no matter how many of these posts you continue to write because my experience in my part of the world and the industry tells me otherwise. I've continued this because you're forced me to, not because I either want to or think it's worthwhile any longer.

I'm NOT after your integrity... I'm sorry if that was what you perceived as my motive...

The fact that, we are so far away and had have build our experience without the one getting interference by the other, it might allowed us to see different results from different aspects... These debates in open forums are pushing the limits of our knowledge forward regardless of our professorial interests.

An ancient Greek say "Γηράσκων αεί διδασκόμενος" "I Always Learn as I Age" (Solon)... So there isn't a DEAD HORSE HERE TO BE BITTEN...

[Luis Otero]

Another successful colorist story with the same approach. Instead of following the easiest path, follow the correct approach for the sake of the obvious better final product. Coincidence , or just following logic?

http://www.finalcolor.com/CMS.htm

[Wayne Chiang]

Thank you Evangelos, very clearly show the different color grading workflow, I have a question, if maximum Color range is LOG ，, so how do we see in Grading when check wysiwyg colour, at present the biggest color space display device is DCI-P3, so the most appropriate View see scheme is DCI-P3 display device ?

[Evangelos Achillopoulos]

Always when we see the image data is like we are in a house high in a hill with beautiful very wide view when we are outside in the balcony (that’s all info which is included in the RAW file). When we get inside to colour grade we see through a window the view (the window is the monitor). How much of the view we will see depends on the size of the window (Gamut of the monitor). How accurate will be a representation of the full view is depending on the shape and the filters we have on the window (Custom LUTs).

So the LUTs are forming an image that represents the result. Any lack of performance (monitors performance) will show as a noise or be clipped from viewing. But 80% of the final result on almost all monitors is there… A DCI-P3 device is good, but some times integrated static LUTs are hiding the reality; the performance in the low 10% of blacks is the most crucial part of the image that most of the monitors don’t perform linearly… The new OLEDs seem that are solving that issue… A modern projector with native contrast higher than 3000:1 solves the issue also…

Is common to my workflow to have a partner Grading for Filmout i.e. and while grading to see noise and difficult shadows to grade. In most cases that results to an uncomfortable DoP and Colourist, when they asking me if it will look ok I always tell them to do a free test print. 100% of the cases in the theatre when they see the test, the results are amazing, no noise and accurate colours in the shadows… all very-very smooth… The noise was there because we where pushing monitors gamut envelope beyond limits… so the combination of LUT and low performing monitor resulted to noise… but when the data where transferred to film that can perform, image was unveiled…

Bottom line is, we always need LUTs to accurately represent reality for our specific display device (filters and shape for our window in the balcony view)…

[Steve Shaw]

There's a fair bit of confusion here about range - for colour and brightness, to use two easy to understand words - as well as the real use of LUTs, and when their use is preferable, and when not.

Without going into detail here, I would sugegst having a look at the pages within the Light Illusion website - the LUT pages specifically, and downloading and reading the two main pdfs - the DI Guide and Scene-to-Screen Workflows.

I'm also very happy to answer any questions on the subject of LUTs, calibration, etc.
We have had to do a lot of colour based development work for LightSpace CMS, so do know the subject in some depth ;o)

[Evangelos Achillopoulos]

I understand your positive drive to help, but saying "There's a fair bit of confusion here about..." you just add to the confusion you claim is existing... So to clear it up a bit for those not understanding all, please help me to sort it out...

What exactly is confused or wrong in my explanations?

[Steve Shaw]

http://www.lightillusion.com/usingluts-info.htm

http://www.lightillusion.com/techpapers.htm

But, the biggest single confusion is saying using film LUTs is preferable...

What is actually preferable is to use captured image data that is over-range and wide gamut, and use a non-destructive display LUT that limits the data range for display on a known standard monitor (Rec709 or P3 in all likelihood).

This then leaves the final image data as over-range and wide gamut (no clipping) such that any desired deliverable can then be generated as required.

Choosing to use a LUT with roll-off is a personal choice, but can just as easily be done as a creative process as part of the grading, rather than as being fixed within a LUT.

And this is probably the biggest issue with saying a film LUT is preferable - film profile LUTs have a toe compression region as well as shoulder roll-off, and this is actually not desirable for digital deliverables as it reduces shadow detail drastically.

So, to sum up, you need a workflow that is non-destructive to the captured data - nothing more - and the correct data conversion LUTs to the desired final deliverables.

[David Rasberry]

REC709 is a defined standard log curve for broadcast display. The target values of the color bar and gray scale reference test signals define a reference series of log visual values converted to linear voltage steps in the video signal. Regardless of work flow, if you are grading on a broadcast reference monitor, you are grading to this log curve.
Reference monitors have to match the color space of your delivery format.
The log value range of REC709 is less than 7 stops max.
Theoretical maximum film projection value range in a totally dark theater is about 9 stops. But most theaters don't come close to this.