During a live stream, Cullen Kelly said that 1,000-nit specular highlights are unpleasant to look at and that windows should be closer to middle gray. Do you agree? Can intense specular highlights serve an expressive purpose?
“HDR is fairly new, and as typically happens with new technologies, we have impartial parties really, really pushing for a particular aesthetic. So what do I mean when I say that? Well, I mean as an industry we’ve come up with like isn’t this great we have these displays that can get as bright as, you know, 700 or more nits on a consumer display and that’s led us to come up with standards that say well, the standard for an HDR deliverable is a 1000-nit luminance level which by the way, if you don’t know what I’m talking about, luminance is just the peak brightness of your display, roughly speaking. So where normal SDR standard dynamic range is going to have a peak luminance of around 100 nits, the standard for HDR when I go to do an HDR master, is a 1000-nit ceiling. So it’s a lot brighter, there’s a lot more we can do up there, but I think one of the challenges that has come about is because we have all these technology and display and HDR solutions manufacturers in the space talking about HDR as loudly and as quote unquote ‘expertly’ as anyone who’s shaping HDR images, the message has become ‘oh, HDR is a 1000-nit image’. That’s the conventional wisdom that I hear a lot, like, the conventional wisdom would be like, ‘oh, if you’re doing a 1000-nit master, your brightest stuff in your image should be a thousand nits, you should have plenty of like speculars and highlights and things in your grade that are regularly hitting at a thousand nits’. That’s the conventional wisdom. It’s not even conventional wisdom, it’s more like an untested assumption, I would say. ‘Oh here’s HDR, it goes up to a thousand nits’ and we as colorists sort of assume ‘okay cool, I guess that means that my grades are now going to have peaks that hit around a thousand’, at least in some cases, right? I think that’s a problematic assumption. Personally, I wish we could look at it here together. We can’t really look at an HDR image over YouTube, at least over this live stream, but what you’re going to find and what I’d be curious to hear anyone’s thought out there, when you look at an HDR image and you look at like a true thousand-nit image that has like some speculars and highlights that are hitting that area I don’t find that pleasant to look at. I really find it unpleasant to look at. So the assumption, or the convention for HDR, is well, if you’re doing an HDR master, maybe the simplest way to put it is your HDR master should look quite different from your SDR master. That’s actually not the way that I grade anymore after having done lots and lots of HDR work. If you come to me tomorrow and you say ‘Hey, I’ve got an HDR project I’d love to grade with you,’ I’ll say ‘Okay cool’, and very early on in the process I’m going to say ‘Let’s set our speed limit in terms of luminance.’ And if you ask me for my take and you want to go with my suggestion, I’m going to suggest that we set a speed limit of around 250 nits because I don’t want to see an image brighter than that. […] You need to remember, HDR is not about a brighter image per se. It’s not like the body of the image is being exposed up. HDR is for sparkly, twinkly stuff in the frame. […] So once you start to connect that up and say like, ‘Okay, HDR is not about how much brighter you want the image to be; HDR is about do you think there is creative advantage in letting bright things – speculars, highlights – be reproduced more linearly, closer to their ratio out in the real world when they’re photographed? Do you find that in some way creatively advantageous to you? That to me is the fundamental question of HDR. Do you find that advantageous and to what extent do you want to exert your ability to let those things be brighter? And the counterpoint to this that I would offer, not just for my POV but other filmmakers who I really, really admire and respect, who will say ‘I don’t need it. I don’t want it. 100 nits is plenty for me. I don’t want highlights that have the same ratio to the subject as they did in the real world.’ I made this example in a conversation with Steve Yedlin, the ASC cinematographer, who I have had the chance to teach a couple times with and chat with every now and then and we had a conversation where like we we’re talking about HDR and we kind of pointed out like okay, if I’m standing in front of a window right now and you and I are having a conversation and you’re sort of facing that window and talking to me as you do without even thinking about it you’re going to walk over to the side because you don’t want to look at me backlit by that window because that contrast ratio is not that pleasant. So, in a sense that’s part of what we’re doing when we master images: making that frame which is something you see in films, shows [and] commercials all the time – a person in front of a window – my take would be that the ideal reproduction of that person in front of that window is not a one-to-one recreation of the contrast ratio that was there on the day because that’s not a very pleasant visual experience. You actually want the highlight to be closer to the middle gray or to the subject exposure because it makes it easier on the eyes…”
Perhaps most disconcerting, for someone who considers themself a color scientist, is that Kelly provides no justification, either psychophysiological or aesthetic, for his decision to arbitrarily limit peak luminance to 250 nits.
Attached, find a downloadable file with 1,000-nit specular highlights. Do they hurt your eyes?
I was the one who asked the question to Cullen that prompted this reply.
I think it’s clear he’s not a huge fan of HDR and it doesn’t really fit with his teachings (ie set a base exposure and then everything you do after should compress/bring down your levels. / Burn not dodge in photography terms).
I also think HDR peaks can be distracting because the gap between them and where the rest of the image sits is so obviously vast that it draws your gaze. If this is intentional then fine, but if the colourist is making this shot decision rather than the Director – you probably do want to establish that speed limit Cullen is pointing out.
There is absolutely no industry consensus that HDR peak highlights must reach 1,000 nits as Cullen claims there is, nor is there any such requirement by any producer that I’m aware of.
I’ll add that Cullen is a big proponent of exploiting the camera’s dynamic range in post and it follows that filmmakers like Yedlin should be doing the same when shooting. He’s also written that those who don’t learn to make compelling images in the new medium will be replaced by those who do, so very disappointed to learn that HDR-loathing Yedlin has convinced him that HDR is an abomination.
I have discussions like this almost every week with DP’s, directors, tv channels and the industry. Finally I can understand why many colorists are not big fans of HDR, but I can’t understand why they oppose it or badmouth it. It is the future and it will permeate everything, it has been happening for about two years. Displays are getting cheaper and cheaper and in a few years it will be the new standard. I understand both sides, being a big proponent of HDR myself.
In principle it is right to develop the highlights with diffuse white not necessarily much higher than 250 nits, in this respect I understand Cullen very well and I agree. What is much more interesting is how to deal with specular highlights, as opinions differ greatly on this. Finally, in HDR 18% grey is key for healthy images and to know where it must be set in the exposure is the most important thing. I’m watching your videos dear Jon and I recognized that you mentioned me in a article some years ago about HDR dispalys. I really like your videos on YT!
Thanks for the feedback, Tim.