Kodak Vision 2383 Greenery Reproduction

[Anton Meleshkevich]

All the 2383 print film LUTs, I worked with, shift all the greenery towards yellow.
But movies, where this print film was used, have green greenery.

I tried to use and compared Davinci Resolve LUTs, Steve Shaw LUTs, ImpulZ LUTs and so on. So it seems like this is typical thing for 2383 colorimetry.
Yes, you can say that LUTs can't be close to film and every time colors are different. But as far as I know, Kodak provided it's colorimetry to Steve Shaw (light illusion). So I think I can trust it's yellowish thing.
I use macbeth colorchecker, set up an exposure, contrast as accurate as possible, adjust WB very accurately. Also I tried LogC gamma, tried cineon gamma, tried not to use wide gamut color spaces and used (as it's recommended) rec709  gamut with cineon gamma and so on. But I still get yellowish greenery.
Also I tried to apply these LUTs to test pictures to see what's exactly going on and did a lot of other things. All these LUTs ACTUALLY shift greenery to yellow.
I can fix the greenery quickly by using hue curve. But I guess this is not the correct way to get green greenery.
I saw a lot of hollywood movies with 2383 print film, but most of them have usual green greenery.
Is it a standard operation with 2383 print film to adjust greens using hue curve? Or may be there is some matrix I should apply before, or anything else?
I'd like to know what is correct way of working with 2383 print film. I can fix it easily without any artifacts using, for example 3d lut creator and it's AB curves in YUV mode, but I want to know the correct way. I guess solution is simple and obvious.
And I know that using LUTs for look purposes isn't what cool colorists do (actually they do), but this answer makes sense only if actual 2383 film print DOESN'T shift green to yellow. Otherwise I'm trying to solve the same thing, what happens with a real 2383 print film.

Edited July 26, 2020 by Anton Meleshkevich

[Bryan Tran]

I've noticed this too. I think what we're seeing is caused by the difference between the color rendition of film negative and digital.  I applied the Light Illusion 2383 LUT to test images from https://cinematography.net/Valvula/valvula-2014.html.  All I did was match contrast and add saturation to the Alexa footage before the LUT.

Kodak Vision3 250D 5207

Arri ALEXA XT

Here's the same two images white balanced to the gray patch on the color chart.

Kodak Vision3 250D 5207 (white balanced)

Arri ALEXA XT (white balanced)

Converting ALEXA Wide Gamut to Rec709 cools the greens a bit, but still doesn't come anywhere close to the film.

If any experts could chime in, please do!

Edited June 17, 2018 by Bryan Tran
Added call for any experts to chime in.

[cameronrad]

I've tried using Adobe's Printing Density ICC profiles to emulate printing density data and got this as the result. https://helpx.adobe.com/after-effects/using/color-management.html I used the Kodak 5218 profile.

______

Adobe Camera RAW Default Color Rendering

Adobe Camera RAW (Custom Cineon Log DCP Camera Profile)

After conversion to Kodak 5218 Printing Density ICC Profile + Kodak 2383 LUT. (Exposure bumped slightly to match ACR Default)

 

Edited June 18, 2018 by cameronrad

[Anton Meleshkevich]

I need some time to dive in and understand what you did 

[cameronrad]

I'm not sure how accurate it is to be honest. I wish I had more first hand experience with this stuff to know for sure. I tried piecing together what I understand from various sites, this one included. Mitch Bogdanowicz's posts help a lot. 

I picked up some stuff from Adobe's old white papers and patents also https://patents.google.com/patent/US9443327

Edited June 18, 2018 by cameronrad

[Marc Wielage]

Well, you could just avoid the LUT entirely and color correct the image by hand. I never trusted the Kodak LUTs, and I worked for Kodak for 2 years. I think they were vague and specious, and there was a lot of guesswork and trial & error involved. (Same thing with Technicolor, and we went through a whole slew of LUTs there.)

If you're going to use a Color Checker chart, I'd look at the Color Checker video (the big one) and place it where the actors are standing. Be warned that the saturation values are not accurate. The DSC charts are fairly accurate, but they cost a lot more money. In truth, some of this stuff is very subjective, and sometimes the filmmakers' artistic intent overrides accuracy.

[cameronrad]

43 minutes ago, Marc Wielage said:

Well, you could just avoid the LUT entirely and color correct the image by hand. I never trusted the Kodak LUTs, and I worked for Kodak for 2 years. I think they were vague and specious, and there was a lot of guesswork and trial & error involved. (Same thing with Technicolor, and we went through a whole slew of LUTs there.)

If you're going to use a Color Checker chart, I'd look at the Color Checker video (the big one) and place it where the actors are standing. Be warned that the saturation values are not accurate. The DSC charts are fairly accurate, but they cost a lot more money. In truth, some of this stuff is very subjective, and sometimes the filmmakers' artistic intent overrides accuracy.

I agree about color correcting by hand, I always tend to get the best results when I just do it manually. I'm a color geek though so I just like to experiment with these types of transforms to try and see what happens. The ICC profile experiment was based off this presentation I saw online: http://www.color.org/ICC_Chiba_07-06-19_PM_DMP_Float.pdf No clue if it's even accurate or this technique is even relevant anymore, I haven't heard any updates and I believe Lars Borg from Adobe is currently working on ACES. Interesting technique to try though.

With the Colorchecker chart, I use the SG but my primary use case is a bit different. I deal with still camera RAWs which benefit from having the extra patches and extended gamut of the SG vs. the standard or the CC Video. The biggest issue I have with the SG is lighting it evenly due to the gloss

[Marc Wielage]

I have no problem with experimentation, and I have to admit, I've occasionally stumbled on some interesting looks just pounding on the knobs. But I'm much more a fan of digging inside the correction to understand why it's happening, and that's not readily done with a LUT.

[bruce alan greene]

These LUTs were designed to be used with scanned film negative, not Arri LogC.  And they are designed not to make your digital file look like film, but to preview what you will see on a film print after printing your digital file to a film negative.  And they are useful for that if one is creating their own LOG master for an Arri laser for example.

These LUTs do have the "benefit "of limiting your colorspace to the intersection of REC709 and a film print. I sometimes use these LUTs to do just this (usually at far from full opacity).  But they do not create a color graded image on their own.  One still has to do that. The color grade that fits into this LUT will not be the same color grade that fits directly into REC709.

-Bruce Alan Greene

DP/Colorist

[Anton Meleshkevich]

@bruce alan greene, you are wrong. @Steve Shaw LUTs, for example, are actually created to use with REC709 GAMUT and cineon or LogC gamma.
The thread isn't about "hey guys I wanna get some film look!". It's about yellowish greenery on a REAL 2383 print film. And if you worked with real 2383 and you can say that there is no any yellowish greenery on that, please tell it me. So I will know that this yellowish greenery is just a part of those particular films in those particular conditions, the LUTs based on.
@Marc Wielage , I have read a lot of your posts about LUTs on other forums, you always say that those LUTs are just print of one day's process. And another day colors will be different. I trust you and I understand these things very clearly. But my question is about yellowish greenery. Is it typical thing for REAL 2383 or this is just this LUTs thing?

Edited June 18, 2018 by Anton Meleshkevich

[bruce alan greene]

Anton, thank you for the blunt reply!   Just kidding...

I have personally gone through the process of creating master files for output to Arri laser film recorder, using these kind of LUT's for previewing what a film print would look like from this negative.  The color grade one does on a computer for output to film recorder is fundamentally different than the grade for video or digital theatrical display.  These LUTs are quite good (within the variances that Marc mentions) at predicting what the film print will look like.  Usually some adjustment of the printer lights at the laboratory is also necessary to make the best print.

When starting such a project, one needs to decide at the beginning what the most important delivery will be.  Film print or DCP or home displays.  If it is film print, than the grading was often done in log space while viewing the preview LUT for the print stock.  But this is kind of a moot point these days as film print delivery is dead.  In our case, way way back in 2012, we graded for REC709 as our master.  For the film print this render was converted to Log C and viewed with the film print preview LUT.  Another pass was made to optimize the film for film out.  Short test negatives and answer prints were made at the lab, and then further adjustment of the Log curve was made to make the best film print.  And the best film print did not look identical to the DCP or home version.  The idea was the best film print, and not the REC709 recreated poorly on the film negative/print.

 

You are most likely correct that a grade for REC709 (from Log) will have yellow greenery when viewed in the film emulation LUT.  Also the shadows will be quite cyan as well.  Remember that the film print stock is designed for optical printing from camera original analog negatives, not digital capture.  So, shooting film in the camera and printing to 2383 will yield pleasing color.  The print emulation LUT is designed to show what happens when a digital file is forced onto the print stock.  It is up to you to fix any color issues before recording the final master negative in the film recorder.

 

It seems to me, you are trying to use these LUTs for something for which they are not intended.  So, I should ask you Anton, what are you trying to achieve here?  Clarify that and I'm sure you will get some good suggestions here on this forum.

[Anton Meleshkevich]

I'm sorry, I didn't want to be rude. I've been google it for months with no success. I'm just trying to understand this thing and everything I get is "LUTs can't make film look" which is a standard answer when anybody ask something about film print LUTs.

Talking about input for Steve Shaw LUTs:
"Note: The Free LUTs provided here expect the underlying image to be 'corrected' for a Rec709 display, with the image either being a LOG gamma image, as defined by the standard Cineon video Log format (which is basically Arri Log C), or have a gamma of 2.4, both with a Rec709 gamut. If you have source images that are of a different gamut or gamma (say, raw camera images that have proprietary gamma and gamut values), they will need to be 'corrected' to match Rec709 gamut with Log C or 2.4 gamma to attain the correct end result."
https://www.lightillusion.com/free_look_luts.html
This is what I meant, talking this: "...are actually created to use with REC709 GAMUT and cineon or LogC gamma..."

Seems like this is what I want to understand:

45 minutes ago, bruce alan greene said:

Remember that the film print stock is designed for optical printing from camera original analog negatives, not digital capture.  So, shooting film in the camera and printing to 2383 will yield pleasing color.  The print emulation LUT is designed to show what happens when a digital file is forced onto the print stock.  It is up to you to fix any color issues before recording the final master negative in the film recorder.

I want to understand workflow for real 2383 print film.
Is it's colorimetry becomes corrected to usual colors by the theater projector? I guess, no. Because preview LUTs still have this yellowish colorimetry.
May be this yellowish thing corrected by input film stock colorimetry wich makes greenery more cold? And it becomes normal after 2383 filmprint? Seems like this isn't true. Because Steve Shaw LUTs are created for rec709 Gamut input.

I'm trying to figure out, why 2383 looks so strange, but hollywood movies still look natural.
May be I missing some simple step that makes 2383 LUTs look normal?
Or the typical way of working is just add 2383 preview LUT and fix all the yellowish greenery using various color correction tools?
I can fix it easliy. But I want to know if it is the correct workflow.

Usually I grade in LogC through 2383-log-to-rec Steve Shaw LUT or D65 Davinci Resolve  built-in preview LUT. I grade music videos for youtube and TV.
And I always fix yellowish greenery using hue curve. Is it correct way of fixing this?
To understand how to fix this I need to know why there is no any yellowish greenery in hollywood movies.

Edited June 18, 2018 by Anton Meleshkevich

[Anton Meleshkevich]

Guys, thank you all for the replies. I'm not a native english speaker so I can sound rude. In my language common way to make speech sound respectful is to use articles, but not words like "please"or similar. So I'm sorry if I sound rude 
I try to speak English correctly.

Edited June 18, 2018 by Anton Meleshkevich

[bruce alan greene]

I don't know Steve's Luts, so let's just stay with the Resolve REC709/2383 preview LUT.

Yes, if you are grading through this LUT you will need to adjust your original Log footage to make a pleasing image through this LUT.  If you have your 2383 preview LUT in the middle of your node tree, than you can make further corrections after the LUT if you wish, using traditional Lift, gamma, gain controls.

 

If you are not pleased with your results using this LUT to de-log your material (assuming it's Log in the first place!), you can use a LUT provided by the camera manufacturer.  This will give a much more neutral representation of your image than the 2383 preview LUT.

 

But you do not need to use any LUT at all, or, as I sometime do when I want a bit of the "film look"  I use the 2383 LUT set at maybe 30% effect.  And even at 30%, I need to grade for this LUT.

 

In your case, you are using a tool that is designed for making a film negative in a workflow that is not making a film negative.  You are not "doing something wrong" that makes the green colors unpleasant. If you see unpleasant colors when you use the LUT then, don't use the LUT, or you will need to grade for this LUT.

 

There are also LUTs that take an existing REC709 gamma 2.2 image and simulate a film print.  But they are derived from the preview LUT and might not make you really happy either.  But, you could apply this LUT at just 25% or something like that and get a filmish "feel" without going too far off base.

 

I really think that if you really want to start with a kind of normal image, from LogC camera original, than use one of the manufacturer's LUTs and start with that.  I know that Marc likes to do the basic color matrix correction himself by eye, but, from my experience, I like to start with the manufacturer's LUT and make it wrong by eye from there  (to my tastes of course!)

 

If you like what you're doing with the 2383 LUT and making your corrections for it, than you can keep doing that.  But I think eventually, all your work will have the limitation of this "look" and you will want to try something different.

[Anton Meleshkevich]

@bruce alan greene Thank you so much! Finally somebody explained me this thing!
I usually use matrix, I've created from colorchecker and macbech official data. We shot it on our camera at daylight at exposure which is perfectly match macbeth gray squares if footage would in rec709. Then I adjust WB in RAW accurately watching at gray squares on colorchecker.
Then I create gamut conversion matrix in tools like 3D LUT Creator. So I get our camera to sRGB  gamut conversion matrix. I remebmer about gamma influence to colors and so on. Finally i get perfect Rec709 image exposed and color-gamut-transformed correctly. And then I get those unpleasant yellowish greenery.
But now I understand that this is normal. I understand that I didn't miss something. And now I can move on and try to deal with it or to create log-to-rec conversion luts by myself using colorchecker and third party apps which is much more natural compared to 2383 LUT.
Thank you, Bruce!

Edited June 18, 2018 by Anton Meleshkevich

[Cary Knoop]

To use the Resolve Rec.709 Kodak  2383  LUTs I would use Cineon Log as a base, but you have to make sure the black values are slightly raised before you apply the LUT and lowered back afterwards because otherwise your blacks might clip.

Soft "Video" look

 

Hard "Film" look

 

[Juan Melara]

Since the norm on forums these days is to judge and then lecture you about what you're doing wrong rather than actually help you out, I'm going to actually help you out.

The yellow/orange greens are normal. There are a few methods around this.

You can implement a 3x3 matrix via the RGB mixer or DCTL to re-align the primaries to compensate for the shifts the LUT introduces. This is tricky if you don't know what you are doing. If you're working with Alexa footage they actually supply a film matrix to get AWG into a colour space closer to a film negative, which conveniently also compensates for the shift. Bottom of page 7 in this document shows the matrix required: http://www.centron.sk/Storage/Document/Product/3453-2012-08-02_Alexa_LogC_Curve-Usage_in_VFX.pdf

Or as a DCTL, last post on this thread: https://forum.blackmagicdesign.com/viewtopic.php?t=61073&p=348495

Or even as a LUT via the Arri LUT generator: http://www.arri.com/camera/alexa/tools/lut_generator/lut_generator/

A simpler method is to just use a hue vs hue curve prior to the LUT to shift the yellow greens back to actual green. If it looks right it is right.

The method I use is to set a node to YUV mode then use the RGB mixer with Preserve Luminance set to off, to interactively place the greens where I want them. First time you do this, use a colour chart and try to match the look of greens under an FPE to the look of greens from the cam manufacturers LUT. Find my Brim LUT rebuild tutorial on youtube to get an idea how thats done.

[Bryan Tran]

I never had much luck with Arri's film style matrix, but I tried using the RGB mixer in YUV mode and this is the best I could get:

Alexa adjusted with RGB mixer in YUV mode, +20 saturation in raw tab, white balanced to gray patch on chart

Kodak Vision3 250D 5207 (white balanced)

Here are my RGB Mixer settings if you would like to try them, but they don't seem to work well on other shots.

Edited June 23, 2018 by Bryan Tran

[cameronrad]

Quick video of Printing Density ICC method I was experimenting with. Using the Resolve Rec.709 Kodak 2383 D65 LUT and Adobe's Kodak 5205 Printing Density ICC. 

http://cloud.cameronrad.com/sVQ6

Edited June 23, 2018 by cameronrad
Had a mismatched system ICC. Rerecorded video

[Anton Meleshkevich]

@Juan Melara Thank you! I used ARRI film matrix numbers on one project. It brings back saturation. But also it makes skin too yellowish.
 

Quote

to set a node to YUV mode then use the RGB mixer with Preserve Luminance set to off

Sounds interesting! I'll try it.

Also I'm going to create custom 3x3 matrix to compensate important colors' shifting.

[Juan Melara]

Alexa balanced for skin tone > AWG to REC709 CST > Resolve 2383 D65 LUT. The yellow/orange greens are quite visible.

YUV RGB mixer node inserted before CST to shift greens.

Comparison Kodak 5207 with Resolve 2383 D65 LUT.

Settings for YUV RGB mixer node.

 

 

 

[Jeremy Dulac]

I tried @cameronrad 's method through the icc profile, using AE color profile converter and it seems to produce a pretty darn nice looking cineon file, fairly close to the look of what I was getting through @Juan Melara 's tutorial on FC hidden film mode (minus the difference in film stocks, Juans tutorial is like the first quality cineon emulation I have come across online, visioncolor is kind of garbage once you really start messing with it). Definitely seeing the extra saturated greens that help counter the yellowish shift from FPE. I in know way claim to be an expert on what film stocks SHOULD look like or how accurate Adobe's Kodak icc profiles are, but I definitely like the results I am achieving.

What I find interesting is the lack of quality film stock emulation on the web, but burried in AE is this profile that seems to work really well and not much info on it. Certainly far better than anything from VisionColor. FilmConvert has been reborn for me after Juan's tutorial, and they told me they should have Print Density profiles implemented before the end of the year. Very cool!

[Anton Meleshkevich]

@Jeremy Dulac I contacted to FilmConvert and and they said me this:
"FilmConvert applies a combined emulation for print, negative and a viewing simulation for projected film.
The result is an sRGB projected film simulation."

[Jeremy Dulac]

@Anton Meleshkevich

Right, but i spoke to them recently after Juan’s tutorial and they said they will be releasing print density versions of the film stocks without the print applied, so essentially a cineon file I believe.

[Travis Ward]

@Anton Meleshkevich did you try out Juan Melara's recent go at this stock? http://juanmelara.com.au/store/kodak-2383-powergrade

He has a video explaining his process, too.