Part I: The State of HDR Film Emulation LUTs

Part II: Moving Beyond Traditional Film Print Emulation

The following excerpts are taken from Cullen Kelly’s Grade School Episode 5 in response to the question, ‘How do we move digital image mastering beyond the inherent limitations of physical film?’

The heritage of celluloid

[…] I really do feel, as a hardcore devotee of traditional film print emulation and of borrowing from the incredibly detailed and fine work that’s been done over the course of the last century with color science as regards film negative and film prints and that whole system, that’s the best way for mastering aesthetically pleasing images that we’ve ever come up with as a species, by far. So we have a huge debt that we owe to those things and we owe a lot of diligence in terms of understanding how did that work and using that as our baseline instead of saying ‘I’m gonna write my own script and play around with my lift and gamma and gain’ – all the work’s already been done, and until we can do it better, that’s really where we owe our diligence, in my opinion.

The limitations of film

However, film – in addition to being an engineered medium – is an organic medium, with physical limitations to it that we are already bumping up against, especially in the realm of HDR; and as we move forward with more advanced imaging, there’s stuff that a film negative and positive system can’t do that we can do today; it’s just a question of how can we do it in an aesthetically pleasing way that takes what’s good from film and moves the needle forward, doing things that maybe film couldn’t.

Decoupling the film negative and film print

[…] one of the limitations of the traditional negative to print system is, we go from our negative, we time on that negative before we print out – we basically are setting our printer lights – and that was how we mastered images prior to the digital intermediate process. […] but one of the limitations built into that is that… you guys have heard me talk so much about getting our color management and our technical mapping in place and then building our creative contrasts and our creative attributes on top of that. That’s the ideal way to work; it allows things to be modular, allows us to target different displays without having to recook our creative decisions – super cool – that’s actually something we couldn’t do with a traditional negative to print system because it’s going from negative – like a log scene referred color space straight out to a display color space, i.e. the print that’s going to be shown on a projector – so, the creative and the technical are all cooked into that film print and they’re really inseparable. So, part of what I’ve tried to do over the last couple years with all of my fussing over traditional film print emulation is to try to decouple those things and get to where like, all right, I have my color management and then inside of that I want just the creative, subjective characteristics of a film negative to film positive system that I can deploy at will; and even within that, I want to be able to control the individual attributes. Super cool. So, that’s one limitation right off the bat that we’ve already moved past, or at least those of us who get obsessed with this stuff and care to work in this way have moved past.

Stretching a traditional film print an additional 9,900 nits

[…] if we think about the behavior of a traditional film print, it’s taking the massive dynamic range of a film negative and compressing it down to the dynamic range of what a film print can reproduce in a theater, which is 48 nits – half the luminance of what we see on a normal SDR display – so it’s tiny. So, there’s only so far out we can stretch that and we can use color science and good engineering to say like, all right, well, let’s take that target of 48 nits or that target of 100 nits for an emulation that was designed for a Rec.709 monitor and we can say let’s explode that out, let’s hit the ceiling at 300 nits or 500 nits; we hit a point around 700 nits or so where we really can’t stretch it any further, where things really start to break in terms of adapting this legacy data into a high dynamic range container. So, that’s maybe the first and biggest one for me, is that I would like to be able to emulate the characteristics of a film negative to film positive system without having that hard cap of 700 nits when I’m in a container that can go up to 10,000. Now, there’s no display currently that can reproduce anywhere near 10,000 nits, but that doesn’t mean that it’s a bad idea to master in such a container, that’s why this container was devised, namely in the Dolby PQ system which we use for Dolby Vision. So, there’s advantage there to be had in terms of being able to design what I was just telling Gedaly I’m calling a synthetic curve that allows me to mimic the characteristics that I care about from a traditional film print but to climb all the way up to 10,000 or 9,000 or 3,000 or 800 – wherever I want to park that peak luma and not be limited by how far I can stretch out the fabric of that initial measured data from that film projected onto a screen before it starts to break apart. So that’s kind of one advantage there.

Reimagining an ideal film stock

[…] in addition to the contrast that we get out of a film print emulation, you guys have heard me talk lots and lots about all of the preferential color mapping that comes along with a film print that takes memory colors and slots them into a sweet spot and takes more saturated stuff and generally kind of drives it down to the floor, makes it denser; so there’s all these kinds of subjective, preferential color re-mappings that go on inside of a film print or a film print emulation, but one of the things that is kind of a fixed ceiling for the most part is, I want you guys to look, not only at the overall kind of shape that I’m seeing move around here – let me get this scope out – there we go – not at just the overall shape that’s me that’s being moved around here but it’s actually getting smaller and getting choked in when I compare it to the full color gamut that I have available to me inside of say, my Arri wide color gamut color space that I’m working in right here, a film print, and even a good film print emulation, even one that has been adapted for Arri wide color gamut, simply can’t reach the full extent of that color gamut. So, we’re limited in terms of the most vivid, most saturated colors that we can express ourselves with when we’re working underneath a traditional film print emulation – again, even when it’s been adapted for those things, there’s just a point at which you’re stretching it out so far and reimagining it so extensively that you’re really no longer meaningfully using the source data that was used to derive that film print emulation. So, this is another advantage that we can have when we’re making, again, what I’m calling a synthetic print is, I can set my color gamut wherever I want. If I want to limit myself to the traditional gamut of a 2383, or, for that matter, to be more scientific, the color gamut of a 5219, for example, plus a 2383 – that being a negative and a positive print stock, or a negative and positive film stock – I can do that; or, if I want to use a color gamut that we simply were never able to encompass with that negative to positive system, I can do that as well while still maintaining the characteristics that I find pleasing about a traditional negative to positive film system. […] our ultimate goal in terms creating looks should not be to get a one-to-one mimicry of traditional negative to positive print that – we should be able to do that, if that’s what we want – but we ultimately need to be moving beyond what that medium will allow for so that we can take best advantage of the DI system as it exists today which allows for much greater dynamic range, much larger gamuts and the displays that are catching up to that, that allow us to be more expressive, to paint with a bigger canvas, if you will.

The main difficulty in data interpretation is that whatever target we use for film sampling, we always deal with a limited set of colors. It can be 24, 116 or even (say) 4000 patches, but this set is always limited and discrete. Real life gives us millions of color tones (for example 16,772,216 in case of digital RGB model in 8-bit color representation).

That means, we always know what exactly 24, 116 or even 4000 specific colors should become. But we don’t have information about what should happen to the other 16,773,216 colors in between (in case of RGB 8 bit, for example). Whether we like it or not, we will have to ‘invent’ new values for them. Of course, based on the measured data we already have. – How we build film profiles, Dehancer blog

Note: 12-bit is able to reproduce 68 billion colors!

Part I: The State of HDR Film Emulation LUTs

Part II: Moving Beyond Traditional Film Print Emulation

One thought on “Moving Beyond Traditional Film Print Emulation

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.