Nicolas Hanson

LUT emulation input data

Recommended Posts

I understand that film emulation LUTs have different tone curves that reflects the tone curves of different film stocks. 

But how precise is this calculation really? As the log input data will change depending on exposure levels, which camera it is shot with, how you modify the contrast before the Film emulation LUT is applied etc. Wouldn't all of this change the Film emulation LUTs tonal range dramatically?

Edited by Nicolas Hanson
Link to comment
Share on other sites

Hi Nicolas,
I like to work in similar way as Juan Melara expains here:
http://juanmelara.com.au/blog/an-easier-way-to-grade-log-footage

For film grade I would use P3-D65 as my working colorspace with an output transform to P3. Film emulsion LUTs for P3-D65 I create them with LightSpace. I prefer to use contrast from the emulation LUTs, sometimes look needs less contrast, than LightSpace LUT manipulations come in handy to lower the contrast in LUT or even only get the gamut emulation from LUT/film profile.
https://www.lightillusion.com/free_look_luts.html

Edited by Teo
  • Like 1
Link to comment
Share on other sites

Hi Teo, thank you for your inputs!  I know how log tools on log data behave and how to transform data from and to different color spaces but I question the accuracy of film emulation LUTs as the input data is different from project to project.

My initial question is, to be more specific - how precise / accurate is this film stock calculation after I have modified the contrast of the clip prior to the LUT. Not much of the film stock preference any longer I guess?

And, how precise / accurate is the film stock calculation as the input data (footage) change based upon the initial exposure levels, different camera sensors etc. How can the film LUT emulation creators claim that the LUT mimic a film stock 100% when the input data varies that much? I guess different exposure levels and camera sensors, modified contrast prior to the LUT etc will make the film stock emulation behave differently from time to time. 

Link to comment
Share on other sites
11 minutes ago, Nicolas Hanson said:

Hi Teo, thank you for your inputs!  I know how log tools on log data behave and how to transform data from and to different color spaces but I question the accuracy of film emulation LUTs as the input data is different from project to project.

My initial question is, to be more specific - how precise / accurate is this film stock calculation after I have modified the contrast of the clip prior to the LUT. Not much of the film stock preference any longer I guess?

And, how precise / accurate is the film stock calculation as the input data (footage) change based upon the initial exposure levels, different camera sensors etc. How can the film LUT emulation creators claim that the LUT mimic a film stock 100% when the input data varies that much? I guess different exposure levels and camera sensors, modified contrast prior to the LUT etc will make the film stock emulation behave differently from time to time. 

They build off standardized code values. Like standardized Cineon values place 18% gray at 10 bit code value 470. 

http://dicomp.arri.de/digital/digital_systems/DIcompanion/ch02.html

http://dicomp.arri.de/digital/digital_systems/DIcompanion/ch03.html

 

Edited by cameronrad
  • Like 1
  • Thanks 1
Link to comment
Share on other sites

Film contrast isn’t a constant value, if you soft light an image and shoot it with film it’ll be flat, if you hard light it it’ll be contrasty, just as with digital.

The output of an image going through a print LUT is entirely dependent of the kind of signal you’re pushing through it, it’s not like every project shot before digital capture had an identical contrast ratio. 

  • Thanks 1
Link to comment
Share on other sites

bwahaaaaaaa - in short... no....

print 200 copies of a reel on friday,  print the next 200 on Monday from a diffrent batch of the "same" stock, and fresh bath, and the resuating prints are from diffrent planets

they will still be within lab tollerance, and that says more about the needed allowable margin of error with chemistry and mfg processes.....

so is your lut made from print stock run on Friday, or Monday or Wednesday, or, or, or....

the only print stock lut that matters is one from a lab that's plotting out your show... the rest are "creative" and there's no need for accuracy there, only doing no harm....

 

 

 

 

Edited by dermot.shane
  • Thanks 1
Link to comment
Share on other sites

All film colourimetry data is measured from the provided film stocks using a highly accurate hybrid spectrophotometer & densitometer combined with an extremely stable and fully automated stepper motor controlled film transport system.

it is very, very accurate+

  • Like 2
Link to comment
Share on other sites

It varies batch to batch, and lab to lab, and day to day...

Welcome to the world of film.

We are accurate to the stock provided to us, and the processing it has undergone.

When building film emulation LUTs for real film productions we generate new emulation LUTs when any one of the above changes...

Steve

Link to comment
Share on other sites

as posted elsewhere... two luts, both claimd to be "accurate" versions of 2383, one an obvious bag of trouble, with banding and suboptimal maths, and costs you $200....Osiris.thumb.jpg.2f2272887e1d5816f53057f99b53a6d9.jpg

the other is free and has no obvoius issues at all;

LightSpace.thumb.jpg.5a98ae0772dc06f387e8fc81e8ce24b0.jpg

the one i question is Osiris, the one i'd use is from Steve Shaw....

the last time i graded a project for a filmout the facility asked for a P3 CTM, they had no more use for accurate emulation luts in 2018

i have worked on projects where the viewer emulation lut changed several time during the grade sched as the lab's chemistry & stocks changed, not recently tho, that was maybe 2008 ish....

Edited by dermot.shane
  • Like 1
Link to comment
Share on other sites

Osiris LUTs are commonly shitty stuff for result. Like maybe 99,9% of luts on web with no difference free or paid. Might be good as monitor creative luts on set I don't know. Never met DPs who used something like that.

More than can't even imagine LUT as emulator. That sounds crazy. It's static file with poor enough bitness inside it for final delivery. That's really wrong way. To Emulate stock I'm sure we need a kinda Stock Emulator Machine with maybe AI core (hey Dado do you plan something like that in your current work) which can analyse digital source and do something extremely smart with that to get organic behave on exit.

  • Like 2
Link to comment
Share on other sites
6 hours ago, Teo said:

Have anyone used FilmConvert? https://www.filmconvert.com Does it supports also P3?

Yap. Works well but sometimes not good I say. Time to time I don't like its color compressor. And there is nothing special what can't be repeated without that plugin. The bottom line is just to use it as grain machine.

Edited by Serge Kosevtsov
  • Like 2
Link to comment
Share on other sites
On 14.2.2018 at 6:54 AM, Serge Kosevtsov said:

Osiris LUTs are commonly shitty stuff for result. Like maybe 99,9% of luts on web with no difference free or paid. Might be good as monitor creative luts on set

How do you determine if the LUT is shitty? By using taste or scientifically testing? 

Edited by Nicolas Hanson
Link to comment
Share on other sites

Pixel sniffing will not really prove anything, and nore will a grey scale image, as a viable film emulation LUT will have cross-colour contamination, varianle grey scale colour temp,, and other artefacts, as that is what film has by default... It is a very, very non-linear colour medium!

See: https://www.lightillusion.com/free_look_luts.html

But, as the page shows, there are a lot of bad LUTs out there.

Steve

  • Like 2
Link to comment
Share on other sites

i agree that in depth anylisis is far better than pixel peeping, but dropping an Osiris 2383 lut over a grey scale takes 2 seconds, and pretty much imedatly shows massive issues with banding on the monitor and easly visable garbage on the scope

but i'm thinking of a double standard.. one for technical luts, there pixel peeping is not enough, either trust your vendor -or- up the game on anlysis

an one standard for "creative" luts like Osiris, there the bar is pretty low, if it can pass "do no harm" it's gonna be good enough for the task at hand

a quick glance at a grey scale is a reasonable first step to eval the "do no harm" thing? At least one can throw out a frightningly large amount of the "creative" luts straight away and focus on the few that make it past the first step? would there be any reason to continue eval'n the Osiris 2383 lut i screencap'd above after looking at it over a grey scale?

end game is the non linearlty of film is not a optimal choice to replaceate if the dammage done is substaintial, but Steve's 2383 has been smoothed out, and is useable without creating banding

i have accurate filmout luts from many vendors, Technicolor, Deluxe etc, but they are only used in the monitoring chain, and removed from the rendering path.. there the non-linearity is a wanted thing, nothing gets rendered with it burnt in, unlike  a 'creative" lut

 

 

 

 

Edited by dermot.shane
adding more thoughts!
  • Like 1
Link to comment
Share on other sites
1 hour ago, Abby Bader said:

Are you still delivering for print?

not any more, last time was about three years ago, for a south Asia based distb, and they asked for DPX@P3

i have lut's going back to the Kodak display manger's LUT's from some time in the 90's, i'm a digital packrat i guess....

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.