Color management setting when working with Cintel scans?

Hi, we just got a Cintel Scanner, there's not much documentation regarding the specifics of the colour space Cintel scans to. The manual only mentions 12bit log, but what's the gamut? No ACES IDT yet? What settings in the Resolve colour management page should we use? Our monitors are all calibrated to BT.709 100nit with 2.2 power law gamma.

[James Little]

Hi Luke,

the negatives are effectively Cineon log, however to make it look better we don't set the white level so low. So it's a density based gamma of 2.046. For PRINT we use 2.2.

We've had a few people get confused when they want to change colour spaces, as this is not trivial on film as it is with video. You can use our 1D LUTs to get access to the full Linear data (i.e. 12 bit stored form sensor), however the exposure and film type used all play into the colourimetry. Unless you have a MacBeth chart in the shot and can use Resolve to colour match against, it's best to stay in the default colour space.

Cheers

Hi Luke,

the negatives are effectively Cineon log, however to make it look better we don't set the white level so low. So it's a density based gamma of 2.046. For PRINT we use 2.2.

We've had a few people get confused when they want to change colour spaces, as this is not trivial on film as it is with video. You can use our 1D LUTs to get access to the full Linear data (i.e. 12 bit stored form sensor), however the exposure and film type used all play into the colourimetry. Unless you have a MacBeth chart in the shot and can use Resolve to colour match against, it's best to stay in the default colour space.

Cheers

Thanks James, so could you please elaborate on the colour management settings in Resolve? Regarding YRGB/ACES, input, timeline, output colour spaces?

Thanks!

I do film restorations all the time, and the trick is that everybody operating scanners (including Arriscanners, Imagica, Spirits, Cintels, you name it) employ different settings. So there is no "one size fits all" setting that will always make the film look good.

Also, different films from different eras look different: 1960s negatives look different that 1970s, 1970s look different from 1980s, and 1980s negatives look different than 1990s. Kodak stock looks different than Fuji, Fuji looks different than Agfa, and on and on and on. (Anything made prior to that is even harder and potentially more unstable.)

My method is generally to ask the scanning operators to balance the RGB at least a little bit so that a white signal is vaguely in the direction of white, nothing is crushed, nothing is clipped, and there's a bit of range between (say) 20ire and 80ire. Given that, I can generally use Offset/Printer Lights in the first node to further balance the signal, plus the RGB Mixer (with lum mix off) to fine tune the adjustments, followed by a custom curve to restore the exposure and gamma characteristics for viewing in Rec709. Everything after that is subjective, related to scene-to-scene adjustment, windows, and keys. Ideally, we have a previous transfer for reference, and/or we get approval from the filmmakers as to direction and intent.

There is no LUT that can do all this, because it's different with every film... even if the scanner settings stay the same. Emulsions change, opticals change, the lab changes, the film ages over time, and sometimes projects change from 35mm camera negative to interpositive to print. There is no way to solve this except by using good judgement and turning the knobs.

Marc and James, thanks for that information.
I never worked with film and only now begin to get a grasp at the complexities of trying to solve the "color puzzle" for grading film. It certainly is harder than I imagined it would be.

In the words of one of the first few pages of the Resolve manual: "you, the colourist; are the colour management system."

[James Little]

Hi Luke,

the negatives are effectively Cineon log, however to make it look better we don't set the white level so low. So it's a density based gamma of 2.046. For PRINT we use 2.2.

We've had a few people get confused when they want to change colour spaces, as this is not trivial on film as it is with video. You can use our 1D LUTs to get access to the full Linear data (i.e. 12 bit stored form sensor), however the exposure and film type used all play into the colourimetry. Unless you have a MacBeth chart in the shot and can use Resolve to colour match against, it's best to stay in the default colour space.

NOTE: There is a 'converting colour spaces' section in the manual to depict what I've discussed above.

Cheers

Thanks James, so could you please elaborate on the colour management settings in Resolve? Regarding YRGB/ACES, input, timeline, output colour spaces?

Thanks!

Basically, we provide the option of you going to a linear colour space (See 1D Luts in Resolve) so that you can 'attempt' to use typical workflows as per other spaces. The way Resolve would normally work, is to assume that it's input is Linear (as all digital cameras are). So normally Resolve is just forward annotating to your selected colour space from the linear domain. From digital cameras etc.. the exposure is linear, so if you don't have a good black, you just shift it down to black and then it mostly comes out ok from there. In Film, if you don't have a good black point, it's resting somewhere on a logarithmic curve due to exposure versus sensitivity of the dyes, as such, unless that's well characterized (what we're doing in the Cintel colour spaces, at least as close an approximation as you could hope for), it's kind of hard to recover from.

If I had any recommendations to give, it would be to use Resolve's colour matching to achieve a near perfect match in the Cintel colour space on a MacBeth chart in the scene, then go to Linear via 1D LUT and then into your desired space via nodes as then you should mostly be ok, always best to keep an eye on whether you are happy with the result or not as others say as ultimately it's whether you're satisfied, but the Cintel Colour Spaces to grade quite well and most people don't bother to leave them. I do know some people have been converting to REC colour spaces as above etc.. and achieving results they are happy with, so don't be too discouraged and give it a go. Without a MacBeth chart in your film scene to match to, I'd recommend you to stay in the Cintel colour spaces provided and grade them as best you can. Film was never really intended for all the tools of modern day, rather simple primary colour grading (i.e. Telecine's) to which the default colour space is more than adequate.

Regards,

James

[DavidS]

Please note using the convert to linear 1D LUT gives you a linear representation of what is on the film not a linear representation of what the film camera was shooting so the output of this cannot be treated the same as if it came from a digital camera. In order to do that you would need to undo the gamma response of the film being scanned.

Negative mage on Negative stock has gamma -0.6. Apply gamma -1.666 (1/-1.66) to linear image after LUT to get back to real world linear.
Positive Image on Intermediate stock has gamma +0.6. Apply gamma +1.666 to correct
Negative Image on Intermediate stock has gamma -0.6. Apply gamma -1.666 to correct
Positive Image on Positive stock (Not Teleprint) has gamma 1.5. Apply gamma +0.666 to correct

I think these values should be true for FUJI stocks as well but we can't find the relevant data for that or teleprint.

As far as colorimetry is concerned we calibrate the machines to match colorimetry to a set of Kodak supplied analysis film.

Film scanning does not easily match an ACES workflow as not only would the IDT have to match the scanner but it would need a different one for every film stock.

(Thanks Stuart Hunt for help with the data)