Arri 3D camera negative LogC2LogC LUT and how I've used it

[Brett Harrison]

MODS please delete this thread. Thanks.

 

[Harry Lipnick]

Brett...sadly I don't have any good info for you, but I'm quite keen to see what other folks chime in with. Just wanted to reply to be subscribed.
Thanks for starting an interesting thread!

-Harry

 

[Harry Lipnick]

New to me, so I'll take it!
Testing at home right now. I don't have a color calibrated monitor, so probably can't make any super informed decisions, but I'll let you know what I find.
Best,

-Harry

 

[Marc Wielage]

This requires you to be able to use multiple LUTs. Full instructions below.

I believe you're overthinking this quite a bit. Realistically, all we ever do is use an input LUT going into the color correction system, then (if necessary) a display LUT on the monitor. Then we apply an actual color-correction based on feedback from the DP and director. It's not that complicated. I would never roll multiple LUTs together, because there's too much danger that one will undo what the other one does, or they'll start to introduce artifacts that can't be reduced or eliminated in the final color correction.

In some cases, custom curves and little tweaks are required, but it's not a question of LUTs per se. I think a lot of people overuse the term "LUT" without understanding that it's not a color-correction per se -- it's a matter of color management and getting the image into the color space you need it to be in, vs. a flat LOG file.

The explanation on Siggraph's "Cinematic Color" website go into LUTs, color space, and color management in great detail:

http://cinematiccolor.com/

Alexis Van Hurkmann's book Color Correction Handbook also draws the distinction between LUTs and actual color correction, which are both very different things (and yet overlap at some point):

Don't ever assume that a LUT will do all the heavy lifting for you. As far as I'm concerned, the LUT only gets it in the ballpark. The real work begins when you or an experienced colorist sit down and do the actual scene-to-scene correction on the project. Everything else before that is just temporary.

And don't forget, unless you have a terrific, calibrated monitor, the whole thing is moot.

 

[Brett Harrison]

I believe you're overthinking this quite a bit. Realistically, all we ever do is use an input LUT going into the color correction system, then (if necessary) a display LUT on the monitor. Then we apply an actual color-correction based on feedback from the DP and director. It's not that complicated. I would never roll multiple LUTs together, because there's too much danger that one will undo what the other one does, or they'll start to introduce artifacts that can't be reduced or eliminated in the final color correction.

In some cases, custom curves and little tweaks are required, but it's not a question of LUTs per se. I think a lot of people overuse the term "LUT" without understanding that it's not a color-correction per se -- it's a matter of color management and getting the image into the color space you need it to be in, vs. a flat LOG file.

The explanation on Siggraph's "Cinematic Color" website go into LUTs, color space, and color management in great detail:

http://cinematiccolor.com/

Alexis Van Hurkmann's book Color Correction Handbook also draws the distinction between LUTs and actual color correction, which are both very different things (and yet overlap at some point):

Don't ever assume that a LUT will do all the heavy lifting for you. As far as I'm concerned, the LUT only gets it in the ballpark. The real work begins when you or an experienced colorist sit down and do the actual scene-to-scene correction on the project. Everything else before that is just temporary.

And don't forget, unless you have a terrific, calibrated monitor, the whole thing is moot.

From the article:

"Note: It is possible to convert the ALEXA-specific wide gamut color space to film color space using a 3D LUT created with the ARRI LUT Generator. The according settings in the LUT generator are shown below."

This LogC2Film LUT doesn't do any of the functions you mentioned. It starts with a log curve as input and ends with a Log Curve as output, therefore it's not a viewing LUT or a gamma conversion LUT - so it's expressly for adding in a film profiling (edit: it says 'film color space' which I assume entails a kind of profiling but I'd appreciate others' input on this).

Other parameters available in the 'conversion parameters' area are REC709 or DCI but they are greyed out when the 'image (destination) gamma' is set to LogC - for clear reasons.

 

[Brett Harrison]

I did a bit more reading on the Arri website and found out more about this film matrix LUT. This page explains with GretagMcBeth chart examples what the intention of the LUT is:


http://archiv.arri.de/AlexaColorProcessing-23Feb2011/html/filmstyle.html


They don't designate exactly which print film emulations (PFE) they're talking about. The ones Bjorn released are of the generic type for "HD Broadcast monitor (ITU
REC-709)"


page 13 http://download.autodesk.com/us/systemdocs/pdf/lustre_color_management_user_guide.pdf


So they could be the print film emulations refered to in the Alexa white paper.


But the basic point is that Alexa LogC doesn't look like ARRILASER LogC files (i.e. film).


Interesting to note that this LogC film matrix is also available in a setting of the Alexa firmware as of software update package 3.0, so your LogC can match up with film out of the camera.


Firmware version (SUP) 2.0 didn't have this feature and relied on the LUT generator.


SUP 2.0 LogC files came in a Camera RGB format:


"The last point needs some additional explanation .In
SUP 2 .x the Log C image was in native camera RGB .
Starting with SUP 3 .0 the Log C image is transformed
in a wide gamut RGB .The gamut is larger than that of
Rec 709 and DCI P3 .
The changes to the Log C in SUP 3 .0 have two visible
results .The images have slightly more saturation and,
for higher EI settings, the black is lower ."


http://www.arri.com/?eID=registration&file_uid=7347 (page 7).


So the SUP 2.0 is the better version of this film matrix for Redusers since r3d can do Camera RGB.


I am aware that the Alexa LogC curve varies according to the EI that camera was shooting in. IOW it's not the standard Cineon/DPX linearisation (perhaps one is?)


http://4.bp.blogspot.com/-p2ZZ2hUg4.../2012-08-02_Alexa_LogC_Curve-Usage_in_VFX.bmp

Therefore it's not clear to me if one or any of these curves are 'valid' for linearising redlogfilm. Of course, the gamma curve (1D transform) applied to redlogfilm is an artistic choice and EI800 seems to be a fairly workable starting point.
(to get the other EI version of this LUT choose 'legal' as the output format in the LUT generator).




So in summary:


this LUT is intended to make Alexa LogC look more like film (scans from ARRILASER),


There's a version of this LUT that expects Camera RGB color space (SUP 2.0)


The curve isn't exactly apropriate but is in the style of the standard cineon curve (the 'C' in LogC stands for cineon)


It's cusom designed for Alexa and ARRI laser.


It's ultimate goal is for use with PFEs like the ones recently out on Reduser and via Juan Melara (http://juanmelara.com.au/print-film-emulation-luts-for-download/).


Caveats:


I don't know if the r3d 'Camera RGB' space is the same as the Arri LogC 'Camera RGB' space. Presumably - but I'd love for someone in the know to shed light on this.


I don't know if the anonymous PFEs referred to in the above Arri whitepaper are comparable to the generic REC709 HD monitor ones from Bjorn.



edit:

Here is (one aspect of) the 'Film Style Matrix' found in the Arri whitepaper 'Alexa LogC Curve usage in VFX'

http://www.arri.com/?eID=registration&file_uid=8026

1.271103 -0.284279 0.013176
-0.127165 1.436429 -0.309264
-0.129927 -0.510286 1.640214

 

[Marc Wielage]

A good colorist can create those precise looks just using the available controls within the color-correction software.

Did you read the links I provided?

 

[Hrvoje Simic]

Brett, this thread has the potential to be very misleading. Can you clarify what are you attempting to accomplish ?

 

[Brett Harrison]

Brett, this thread has the potential to be very misleading. Can you clarify what are you attempting to accomplish ?

To accomplish the usage of the ARRI 'Film (3D)' LUT on RED footage.

It's just an experiment in using another of the ARRI LUTs on RED footage (the ARRI LogC2Video LUTs are popular de-log LUTs with some Redusers). In this case the 3D 'film' LUT, which is intended to make Alexa look more like log film scans before viewing it in a print film emulation LUT (or scanning it out to print stock), like the ones recently released on Reduser.net.

Unfortunately I started the thread before I had a proper grasp on the intention of the LUT so that makes it a bit confusing to read. I thought it might be good to have the thread nuked by a mod.

 

[Marc Wielage]

My expectation is that the Arri LUTs are intended for use with the Alexa. Why not just use Red's own color science built in to Redcine-X Pro? There's also some terrific ones out there on the net from various sources, particularly within Baselight and Resolve. Once you have a regular Log file, there's any number of ways to prep it for final color.

Many, many, many projects use custom LUTs, but it's just a question of turning the knobs until you like what you see. Once that happens, you check various scenes, see if they work, and if so, write down those numbers and use it for the whole project. I'm not the kind of guy who would change LUTs for every scene or every shot, assuming the DP shot the film on one camera using comparable settings.

 

[Brett Harrison]

My expectation is that the Arri LUTs are intended for use with the Alexa. Why not just use Red's own color science built in to Redcine-X Pro? There's also some terrific ones out there on the net from various sources, particularly within Baselight and Resolve. Once you have a regular Log file, there's any number of ways to prep it for final color.

Many, many, many projects use custom LUTs, but it's just a question of turning the knobs until you like what you see. Once that happens, you check various scenes, see if they work, and if so, write down those numbers and use it for the whole project. I'm not the kind of guy who would change LUTs for every scene or every shot, assuming the DP shot the film on one camera using comparable settings.

Not all LUTs are born out of nob turning though. Some of them are based on the recorded measurement of physical effects. This was my hope for this ARRI 3D LUT which takes a log container and makes modifications to it but leaves it in log space. It's meant to add in a certain film effects that are missing (see the GretagMcBeth charts in the above link) from LogC when viewed next to filmed McBeth carts. It's 'arrilogfilm' if you will. Remember that this method involves switching to Camera RGB so it's quite divorced from Redgamma3 which is just a LUT applied to RLF (in Graeme's words) so this method uses the Cineon-compliant curve of RLF but the colorspace of RGB, the same color space that the 2.0 version of this LUT expects. That's as much as I know and there may be one or two assumptions in there that are wrong about the equality of Camera RGB in each platform and the extent to which the log curves match up (but according to a lot of people they're pretty compatible).

I am no expert but I liked the results--I guess you'd have to try it to see what I mean.

 

[Marc Wielage]

Not all LUTs are born out of nob turning though. Some of them are based on the recorded measurement of physical effects.

Naaaaa, we did it all with trial and error at Technicolor and Cinesite (Kodak) over a period of years and years. I've never been convinced that the MacBeth charts are useful except when printing film-to-film; I think there are better charts that are more accurate for digital correction, and I also find greyscale charts to be much more useful overall. The sad reality is that almost nobody uses charts on the set anymore, not even on the first day of production.

Bear in mind that I learned a long time ago that log value images are whatever the person or company wants them to be. I've worked on projects where film from the same shot was scanned at eFilm, ILM, Cinesite, Technicolor, and Imageworks, and all five images were different. There is no single standard for what a log image is. You can't even take a visual effects shot from one facility and have it look exactly the same at another facility. This is why you have such vastly different color science between Red, Sony, Arri, Canon, and so on. In truth, all of them can work in the end. But I wouldn't dream of using a LUT intended for one camera and to use it with another, for the simple reason that I've seen too many trainwrecks in post where people experimented with dodgy LUTs, especially 3D LUTs (all of which are explained in detail in Jeremy Selan's technical paper on color science and color management that I referenced earlier).

I again say that a LUT is not intended to be a look -- it's intended to take a LOG image and get it into a given color space for further correction, or to correct for a specific display device. It's just one link in the chain, just a starting point, not the final destination for what the image is going to be. You can't avoid the use for a color-correction system or the help of a good colorist, because there's always going to be specific shots that look sorta/kinda OK on there own, then look terrible when cut against each other.

I think the Red color science people worked very hard to get this part of the equation right, and I think it's a mistake to try to assume you can figure out another way around it.

 

[Brett Harrison]

Naaaaa, we did it all with trial and error at Technicolor and Cinesite (Kodak) over a period of years and years. I've never been convinced that the MacBeth charts are useful except when printing film-to-film; I think there are better charts that are more accurate for digital correction, and I also find greyscale charts to be much more useful overall. The sad reality is that almost nobody uses charts on the set anymore, not even on the first day of production.

Bear in mind that I learned a long time ago that log value images are whatever the person or company wants them to be. I've worked on projects where film from the same shot was scanned at eFilm, ILM, Cinesite, Technicolor, and Imageworks, and all five images were different. There is no single standard for what a log image is. You can't even take a visual effects shot from one facility and have it look exactly the same at another facility. This is why you have such vastly different color science between Red, Sony, Arri, Canon, and so on. In truth, all of them can work in the end. But I wouldn't dream of using a LUT intended for one camera and to use it with another, for the simple reason that I've seen too many trainwrecks in post where people experimented with dodgy LUTs, especially 3D LUTs (all of which are explained in detail in Jeremy Selan's technical paper on color science and color management that I referenced earlier).

I again say that a LUT is not intended to be a look -- it's intended to take a LOG image and get it into a given color space for further correction, or to correct for a specific display device. It's just one link in the chain, just a starting point, not the final destination for what the image is going to be. You can't avoid the use for a color-correction system or the help of a good colorist, because there's always going to be specific shots that look sorta/kinda OK on there own, then look terrible when cut against each other.

I think the Red color science people worked very hard to get this part of the equation right, and I think it's a mistake to try to assume you can figure out another way around it.

Right, I never meant it to be a dis against RED. It's just an experiment (albeit naive!). Anyway thanks for pointing me in the right direction