Steve Yedlin on HDR Cinematography

Looks like Blackmagic's Hook and Brawley were on hand for this "Debunking HDR" presentation from Steve Yedlin. It's an interesting watch.

https://www.yedlin.net/DebunkingHDR/?fb ... eIBPCqhJtQ

Look, this is going to sound pedantic but he's not really discussing HDR cinematography; he's discussing HDR as a finishing format and an post-pipeline. They're entangled but very different things.

Look, this is going to sound pedantic but he's not really discussing HDR cinematography; he's discussing HDR as a finishing format and an post-pipeline. They're entangled but very different things.

Out of interest could you give some more information on HDR cinematography or where to look for it on te net? Thanks.

A great piece by Yedlin and highly recommended for anyone working with imagry. Yedlin clears up misconceptions about HDR, just as he has done with resolution. Yedlin explains some important concerns for skin tones and typical surfaces. Where I would differ with him is with subjects of water-like, specular reflections and emissive light sources. He does talk about the "punching through the ceiling" which does give some response to these. However, overall I think he may overlook some of the features of HDR not attainable with SDR. Although he makes excellent points about the majority of images looking identical as either HDR or SDR which I also find critically important.

Yedlin uses the term "garish" for much of the HDR look. I agree, but this is only when HDR is not utilized in a properly controlled manner. Too much by many done in HDR has this garish look. Now that said, there is likely a place for such look, but not as most are currently using it. Further, things are complicated by creators of DRTs "baking in" such a garish look. I agree with Yedlin as to his comments on DolbyVision. With ACES2 some individual effort has to be added to maintain consistency between various HDR and SDR "flavors", but it is doable and ACES2 controls many of the issues of previous versions.

Yedlin states several times that he likes his tones and values with a certain creamy look and typically not "punch through the top" or even too close to the "top." And this is where I diverge and incorporate into my look some of the "punching through" values. But be warned, just letting HDR values fall where they may is no excuse for good art. Values need to be placed where they work for the image and story. Many of my own imagery will be close to what Yedlin describes as his preference, but much will differ when I need to add that certain look that I want in my imagery.

Critically important to HDR is having at least 10-bit, although I strongly desire 12-bit. When will displays get there and affordably? And as Yedlin indicates, HDR is a real challenge to distribute mainly because of necessary compression. He makes a good point that SDR can present a same or better look when confined by codec and compression. But when bit-rate presents no issue HDR can and will rule.

Yedlin makes an interesting issue of SDR as relative and HDR as absolute. However I feel it is a bit more complicated in that lighting ratios are the important consideration (as he points out) and will cause the HDR results to be relative and not just an absolute. I think of HDR as a larger box fully containing SDR.

In pointing out from the outset of the program that human perception is relative not absolute, the subjective claims for the superiority of HDR pretty much fall apart, because the same relative measures can be created in SDR within a smaller range (no need to "punch through"). For the viewer, it's all the same, because the work itself sets the baselines and the reference points.

And there's the matter of aesthetics.... Yedlin's preferences aren't idiosyncratic. Everybody in that room internalized the same standards from both earlier cinema, and art history. The folks who are mad about HDR -- brighter! more saturation! pops! -- tend not to share those standards. Maybe they'll end up making something new. Or just produce a lot of vulgar, garish videos....

This is my own opinion, my own interpretation of HDR and SDR as how I like to see them for.
SDR = Theatrical screen, Projector, Organic, and for relative perceptions. Excellent for narrative films.
HDR = Digital panel monitors, large panel screens, high saturated look, vivid for sensory capture. Excellent for commercial content (I.e. ads, events).

I have not had anyone asked for HDR grading and delivery for cinema delivery (even streaming), even on digital projectors. So far, 100% of DCP delivered came from my SDR workflow.

What’s your opinion and experience and what kind of projects do you do (narrative films, documentaries, or commercials/events)?

Look, this is going to sound pedantic but he's not really discussing HDR cinematography; he's discussing HDR as a finishing format and an post-pipeline. They're entangled but very different things.

Oh, I missed this. Totally agree with Alex. See my previous post. Nowhere did I relate HDR or SDR with Cinematography. Just a post-pipeline. I’m not expert but that’s what I know and have experienced with.

For me the value of HDR lies in that ST2084 curve, which Yedlin correctly identifies as absolute, mapping digital numbers to real-world brightness values instead of a relative curve as found in SDR (Rec709, say).

The relativity of the simple gamma curve means that an image graded for SDR can look radically different on two displays, even when both are correctly calibrated for it. Nowhere is this more drastic between an OLED display and an LCD one, as only the former can produce true blacks.

For me the value of HDR lies in that ST2084 curve, which Yedlin correctly identifies as absolute, mapping digital numbers to real-world brightness values...

Yes. I haven’t watched the long presentation yet, but he may touch on the variability in real world brightness.

When OLED was new several years ago, the diffuse white point was at 100 nits, and that’s how I graded my ST2084 deliverables. In the last couple of years I believe the 4KTVs are brightening that diffuse white point to 200 or 300 nits and so lately I graded in that range. The reference is still absolute, but the diffuse white point nit value can exceeds that original target.

When I grade SDR , again I reflect changing nit values. sRGB can now go up to 500 nits on a consumer monitor and Rec.709 for playback on HDTV is typically well below that.

I do three grades for each potential monitor or television I expect a viewer will use, so unlike everything else you see, I’m not limited to Rec.709. No one has complained about my deliverables in terms of colour or brightness as I stay within the recommended limits while grading. Not sure how Vimeo massages those videos I upload to their platform for viewing.

Look, this is going to sound pedantic but he's not really discussing HDR cinematography; he's discussing HDR as a finishing format and an post-pipeline. They're entangled but very different things.

Yes, but this is still important for us as cinematographers. Especially since recently I was having the conversation about getting those new Flanders Monitors for HDR monitoring on set. And, this really enlightened me as to the misconceptions I had regarding HDR.

There’s so much information contained within this presentation. It’s valuable for anyone working in the industry to watch this and get the understanding of this subject matter. Especially when it comes to working on crafting the image we want to master for audiences.


Sent from my iPhone using Tapatalk

One doesn’t need that $30+K top of the line 4K large Flanders monitor to grade in HDR. Although that would be an ideal, having a recent higher-end OLED 4KTV as a viewing monitor and DaVinci Resolve with HDR waveforms and grading tools would likely suffice for non-theatrical releases. If your work is destined for theatrical release you might want one of the industry stalwarts to provide finishing of your grade.

I use the less-than-ideal Apple 6K Pro Display XDR released in 2019 with the Mac Pro 2019. This monitor doesn’t do a perfect job considering it only has an LCD screen and only about 500 local dimming points so bright lights against black backgrounds don’t display perfectly (later screens have thousands of local dimming points so can look much better). What it does support is ST2084 1000nits everywhere and 1600nits on 10% of the screen area. I grade to a maximum of 1000nits for any very bright bits which seems to be the maximum on 4KTVs although most of the screen is about 200nits or less.

Edit
In Yedlin’s terms my screen limit is 1000 nits for HDR when I’m grading which is pretty much never quite reached and my scene white is between 100-200 nits (the diffuse white point). So i’ve been doing things in general conformance to Yedlin all along.

And why I like HDR grading so much is the ability “to punch through that scene white” as my creative choice. It does enhance the beauty of the image having that higher highlight when it’s important as if you reserved that punch for a Star Wars light saber; I would not use it for any of his other example scenes such as an even brighter exterior while filming an interior.

Look, this is going to sound pedantic but he's not really discussing HDR cinematography

But we've all been shooting HDR for years now, and there's not a whole lot to discuss. Even the original BMPCC is/was an HDR camera. So what's really changed for the shooter, other than more consumer-level displays able to display that full range of data in an HDR color space?

And does it really matter, when you can get an equivalent representation of same on an SDR display?

The challenge with calling anything “HDR cinematography” is that it doesn’t have a separate coherent meaning from simply “cinematography”.

As Yedlin carefully lays out in his presentation, “punching through scene white” is essentially a grading look, not some completely separate process.

And yes, if a production plans to deliver that look, it’s useful to ensure that it is what is displayed on the set monitors. But ensuring that the set monitors are displaying the show look is no different from what should be done for any creative look, whether it punches through scene white or not.

But we've all been shooting HDR for years now, and there's not a whole lot to discuss. Even the original BMPCC is/was an HDR camera. So what's really changed for the shooter, other than more consumer-level displays able to display that full range of data in an HDR color space?

That's true, anything with over 12 stops of dynamic range falls into the HDR category, at least as far as potential goes. What's changed is that display technology has improved drastically over the past few years also.

And does it really matter, when you can get an equivalent representation of same on an SDR display?

It does matter, because on an HDR display you can show pure blacks and superwhites with color and detail that far exceed the clipping range of SDR. Trying to make HDR footage look the same as SDR footage is a wasted opportunity.

Truthfully this changes my perspective of the need for on set monitoring in HDR. Now, in the Filmmaker's Academy video live stream on the URSA Cine 12K Tor told Shane that the camera can monitor with LUTs for HDR. However, Steve brought up some incredible points within this presentation that convinced me that on set monitoring in SDR is absolutely fine.

In large part there was some great breakdown on the streaming compression numbers, and on set UHD SDR Monitoring seems a lot more reasonable over a Wireless Video Transmission. His argument truly won me over to the added benefits of relative values.

We're at an interesting time for on set monitoring. And, I do love my Flanders Scientific DM241, which I have used on set along with the predecessor 240 model. Simply put the monitors have always been great even in HD.

The greatest benefit of the new XMP Monitors is more than the HDR, but the UHD 4K level of SDR in the HDR Monitor. Steve stated as much using the 310s with both SDR and HDR on each monitor. Could you monitor with such monitors on set in HDR? Yes. However, SDR allows an approach more akin to how many of the cinematographers have been lighting for a century.

You want something interesting: Shane Hurlburt actually commented on Instagram on the release of the URSA Cine 12K Comparison to the Alexa 35 stating that the Alexa 35 is for cinematographers who don't know what they're doing. Insinuating that cinematographers don't need so much dynamic range. There seems to be a rebuke for such high dynamic range. https://www.instagram.com/p/DJRzFSjp9TZ ... BiNWFlZA==

Screenshot 2025-05-05 at 7.23.39 PM.png (61.47 KiB) Viewed 2093 times

And, this seems to be the unifying cry against HDR Cinematography that is happening across multiple cinematographers. It's an interesting turn.

Sorry, I have to run as I need to generate better proxies and upload today's footage for a client. I shot 880 GB worth of footage on the URSA Cine 12K alone. Felt good handheld. But the shoulder does hurt a little. UMP Shoulder Mount is certainly not properly designed for this camera. Works in a pinch, but I hope that the true URSA Cine 15mm Baseplate Shoulder Pad is more conformable.

By the way, the camera does feel very hot next to your cheek. I feel in many ways the smaller URSA Mini design would best serve for handheld at lower power draw. Instead of BPU I really want to stay in the B Mount family as I love the battery mount. 150Wh batteries would be perfectly fit on an URSA Mini 12K.

[quote="Rakesh Malik]…Trying to make HDR footage look the same as SDR footage is a wasted opportunity.[/quote]

I was surprised as Yedlin seemed to be drinking too much of his own Kool-Aid saying he could have applied HDR punches, but he chose not to do that which demonstrated his tenet that a HDR 4K grade and a 4K SDR Rec709 grade (which I should call BT.1886) could look identical on the monitors his audience viewed. I watched his video on my PQ ST2084 reference (which I should be calling Rec.2100).

I disagreed with his insistence on this approach as there are times when that extra punch enhances the viewers’ experience. I’m not creating ridiculously bright colour on my HDR grades that detracts from the central character as he stated having brighter exteriors in an interior scene would detract from the actor. I use the punch like you might season a steak, to enhance the scene. And I wouldn’t brighten the exterior either.

Yedlin is very smart and very conscious of what he’s doing. Nobody is going to listen to me on this little tweak that HDR affords the colourist. I grade slightly differently with different scopes for SDR and HDR. I see no problem with forcing a video village to use HDR monitors as SDR is very well understood and ubiquitous. HDR grading can be performed in post as your starting point and then slightly tweaked for SDR. Scene white does not have to be an equal value. You’re welcome to follow Yedlin’s approach, but I suspect colourist’s approach will take advantage of each monitor’s dynamic range.

Again I’m not grading for theatrical release which would benefit from grading with a different scene white and a different edit studio.

It does matter, because on an HDR display you can show pure blacks and superwhites with color and detail that far exceed the clipping range of SDR. Trying to make HDR footage look the same as SDR footage is a wasted opportunity.

Pure blacks are not strictly the realm of HDR, and as Yedlin insists, human perception is relative to the other values in the field of view. It's not quantitative, with bigger numbers always being better. As with resolution, more isn't better in itself. It's just more.

And it's not a matter of trying to make HDR look the same as SDR. It just that, for the viewer, it will be next to impossible to tell the difference, with appropriate adjustment of the footage in an SDR color space.

This is not a prescription against working in HDR. He's addressing what he considers misconceptions. Roger Deakins did not sit there patiently because Yedlin doesn't know what he's talking about.

And it's not a matter of trying to make HDR look the same as SDR. It just that, for the viewer, it will be next to impossible to tell the difference, with appropriate adjustment of the footage in an SDR color space.

This is not a prescription against working in HDR.

Monitoring in HDR offers greater detail in both highlights and shadows, but for the average viewer, the difference is often negligible—it’s largely a matter of perception.

Consider the analogy of a 10-megapixel vs. a 100-megapixel camera. A 40x30 glossy wall print from the 10MP image may appear pixelated and unpleasant when viewed up close (around 2 feet), even at 300 DPI. However, from 6 feet away, that same print at 150 DPI becomes much harder to distinguish from the higher-resolution version. Switch to a canvas print and view it from 8 feet, and both images appear virtually identical—even at just 75 DPI.

The same principle applies to HDR vs. SDR in theatrical settings, where audiences are typically seated 10 to 20 feet away. At that distance, the perceived difference between HDR and SDR diminishes significantly. HDR does offer a visual advantage for viewers watching on 4K OLED displays from about 6 feet away, where enhanced contrast and dynamic range are more noticeable.

In my view, viewing distance plays a more significant role than the format itself, and it’s not always worth stressing over the technical differences. Of course, if you’re passionate about the finer points of image fidelity, that’s absolutely valid. But when I make films, I prioritize the experience of the typical viewer—and so far, SDR continues to be a perfectly acceptable and effective choice.

One thing missed in the comments here is his noting that there's absolutely nothing in the Bt.1886 specs limiting image brightness. The 100 nits thing is only mentioned in specs in the accompanying documentation, in the section about suggested grading environment and monitors.

I personally hadn't thought about that. Essentially "SDR" is designed to adapt to whatever the monitor can do, as far as contrast range.

As far as color space and volume, I think he's correct that 2020 gets near ridiculous. P3 is an entirely adequate practical space.

I found his argument that using a gamma curve relative brightness format both sets the image to whatever any particular screen can provide, and at tge sane time drastically cuts required image data, therefore bandwidth required per image quality. Meaning you actually get more usable image data via streaming with that formatting compared to the far greater data requirements for the absolute specs used in current HDR formats.

In all an interesting watch.

Sent from my SM-S908U using Tapatalk

So at 42:06 when he says that 18% gray is usually mapped to 10% in SDR, that's not something I should be seing in the Resolve scopes because he's talking about the actual light output on the display? That's 10 nits on a display where you've set white to 100 nits, correct?

To see 10% (or 10 nits) on the internal scopes I would have to apply the appropriate EOTF and switch the scopes to HDR mode? Then it sort of hits 10 but not exactly.

Pure blacks are not strictly the realm of HDR, and as Yedlin insists, human perception is relative to the other values in the field of view. It's not quantitative, with bigger numbers always being better. As with resolution, more isn't better in itself. It's just more.

The industry standard color spaces for SDR can't actually show black. They can't get darker than dark grey. That enables colors to look a lot more vibrant, and even without a dedicated HDR grade, using an HDR display system can enhance vibrance simply as a result of being able to display true blacks that aren't grey.

And it's not a matter of trying to make HDR look the same as SDR. It just that, for the viewer, it will be next to impossible to tell the difference, with appropriate adjustment of the footage in an SDR color space.

I have met colorists who teach HDR grading techniques that can be summed up as: grade for SDR, then apply Dolby trims. Literally -- that's not HDR grading, it's SDR grading with extra metadata to make it work in HDR.

This is not a prescription against working in HDR. He's addressing what he considers misconceptions. Roger Deakins did not sit there patiently because Yedlin doesn't know what he's talking about.

I know that Yedlin knows what he's talking about, but I have met quite a few people who don't understand what HDR really is, or that it's possible to expand the creative options like Rick described. I also haven't had a chance to watch the whole video yet, so I'm not intentionally commenting on anything he said yet; if it sounds that way it's just coincidence right now

The industry standard color spaces for SDR can't actually show black. They can't get darker than dark grey

News to me. Whether SDR or HDR, zero is zero -- if the display is capable of it. What's your source for this?

I am more than willing to help Steve Yedlin appreciate what HDR brings to the palette if he wants to reach out to us. I’ve been thinking of an example I could share on Vimeo. One of my recent public music video promotional shorts might suffice.
Seriously.

Here’s a 2023 60 second promo graded for internet viewing SDR sRGB:



And here is the version I did for 4KTV with HDR PQ Rec.2100 so you can view this on a HDR monitor or any recent model iPhone supporting P3!:



Just a little difference with the grades almost identical, steak with some sizzle.

At my studio I have an LCD monitor and an OLED TV that I use for grading, both color-accurate to Rec709 Gamma 2.2. Even though they are both "correct" in the eyes of the Rec709 colorspace, the blacks on the LCD monitor can never match the true blacks of an OLED, and the result is that even with the same image, the perceived contrast is much lower on the LCD than on the OLED. Thus, I must create two separate grades for each display technology, even though it is the same colorspace.

So in effect, the Rec709 (SDR) colorspace fails to standardize the look of an image, because of its lack of mapping to absolute brightness values. You could have an LCD with a backlight blasting 5000 nits at the top, 1000 nits at 0% blacks, and still be considered correctly calibrated for SDR, because a gamma curve defines only the relative shape of the curve, not its position. And it'll look very bad.

The HDR spec attempts to fix that, or at least the ST2084 curve, with its absolute mapping to nits. In my opinion that is really the only boon, in practice, for HDR; the wider gamut of Rec2020, for example, will never really be exploited much, as it is more a theoretical standard. Find me one real-world display that covers 100% of Rec2020.

And yes, Yedlin's suggestion that SDR and HDR don't look differently holds true in practice if not theory. How many HDR grades of the same film look radically different from its SDR grade? Turns out that what we traditionally consider to be good lighting and cinematography doesn't really require much dynamic range to begin with, so in some ways I understand that snarky ig reply. For HDR, maybe blow out small things, like a sun or a lightsaber, but not much more.

It is so painfully obvious who in this thread has actually taken the time to listen to the presentation and who hasn't.

The industry standard color spaces for SDR can't actually show black. They can't get darker than dark grey

News to me. Whether SDR or HDR, zero is zero -- if the display is capable of it. What's your source for this?

Zero on a Rec709 display maps to a dark grey. On an HDR display zero IRE displays as zero.

That's one of the big reasons that OLED displays look so gorgeous, and why the Crystal LED makes every other video wall look flat, drab and boring: black is actually black.

Christie had a demo theater set up at NAB once that was light sealed, so when the projector was turned off it was as dark as a cave -- you couldn't see your own hand. They turned the laser project to SDR (Rec709, IIRC) and the screen went grey and lit the place up. They switched it to HDR and the theater went black.

Zero on a Rec709 display maps to a dark grey.

No, it doesn't. The other posters are correct; black is black in Rec709. The only reason you'd ever see the dark grey you're talking about is if the technology is incapable of emitting zero light—e.g., edge lit LCDs—or if your display and image pipeline are improperly configured/calibrated—like if you feed a legal range signal in to a display configured for full range.

Maybe someone is flashing back to Pedestal?

Just a little difference with the grades almost identical, steak with some sizzle.

Rick judging the grades on my I-pad capable of displaying P3 @ 500 nits (that’s within your hdr target, right?) you made the choice to do 2 different grades. In doing so you broke the red shirt In the HDR version of the left guitar player when he is standing up during the first song.
I think the shirt was near clipping when you captured the concert (is that proof there is something like HDR cinematography, or just bad luck as its out of the range of the camera and you do not have control over the stage lights) Stretching the reds into hdr territory didn’t help, although you can break them within sdr just as well (which judging from the sdr clip, you didn’t).

It is so painfully obvious who in this thread has actually taken the time to listen to the presentation and who hasn't.

You got that right.


A few comment on other things noted:
1) One can do a custom grade for each particular particular display, but it is more common to grade for a calibrated norm which is shown on many unknown displays. Just as a painting may be viewed in many scenarios, so too a film/video. Even if one sells the display along with their film, viewing conditions might vary. Same for music.
2) There are available now some laser projectors which display the entire 2020 color space.
3) As with printing photographic images (chemical or ink) maximum black is not the same as the black maximum. Establishing the black in an image is part of the creative process as much as lighting and placing any other values.

Rick judging the grades on my I-pad capable of displaying P3 @ 500 nits (that’s within your hdr target, right?)...

Thanks for your salient comments, Bunk.

I’m grading HDR for 1000 nit displays so there may be problems when viewed at 500 nits but I usually don’t let the levels get to 1000 nits; I liked the brighter emphasis in this grade though as it seemed suitable to the emotional way the guitar player was playing.

I think the shirt was near clipping when you captured the concert (is that proof there is something like HDR cinematography, or just bad luck as its you do not have control over the stage lights...

Shooting in 12bit log raw as I do is HDR cinematography whether or not the captured light levels actually exceed the intended limits. I try to avoid clipping of course in my exposure and frequently monitor for that purpose, but it can happen in some areas of the frame. Near clipping is good as I can manage that.

The grading is different for HDR because I have a greater possible range of values. For the SDR grade, I suppress those levels to try to stay within SDR display ranges. Grading them the same as Yedlin appears to be recommending is not my artistic choice and Yedlin recognizes that HDR display grades can be different, they don’t visually have to match SDR display, the way he appears to do with his deliverables. I posted the two grades to illustrate the possible advantages of the visual differences.

I don’t have any control over the intensity of the lights in this performance which the lighting director often changes from rehearsal to performance or varies within a performance or sometimes during a song. These are live events as you appreciate with no retakes. So I need to compensate in post, but I don’t recall the red being clipped when grading with the HDR waveform (that goes to 10,000 nits). Clipping might occur but I pull the HDR levels down in post so they do not exceed 1000 nits for the HDR deliverables.

Yedlin is right about everything. The problem is that it's now 10+ years on with HDR. It's picked up more inertia than can be undone. With it, directors have lost control of creative intent, likely forever.

He does recognize that heading for HDR that punches through scene white but stays below display peak white is permissible when used artfully by the director. That’s what I’m trying to do in the example I posted above. I’m not doing this with skin, just tiny portions of the frame when I judge it’s appropriate.

I’m not sure I’d find acceptable that second to last shot in his video featuring the highlight side of the actor’s face looking blown out (although the lighting is motivated by the flare from a bright light out of frame). He’s so careful to keep everything in his grade at 200 nits or below (as I try to follow for anything with important detail).

I agree that he has a good approach for suggesting a “scorch” feature be added in the colour space of BT.1886 and drop HDR. But he seems to think that would only be used for a “caustic” grade.

That’s an awfully big chip on his shoulder to insist on using words that have very negative connotations. Nobody wants to be scorched or drowned alive in caustic soda (sodium hydroxide).

Why not call it something like sizzle or spice that might have a positive connotation? Sizzle or spice are understood to be used sparingly to enhance your meal and wake up those taste buds. Like I tried to do sparingly in my example, heightening some visual impact, intending to be artful.

Maybe someone is flashing back to Pedestal?

Analog NTSC was 7.5 IRE but still mapped to zero nits.

So at 42:06 when he says that 18% gray is usually mapped to 10% in SDR…

SDR mid grey sits between 41 and 50% depending on the gamma, 10%???

Good Luck

It's all relative. They had superwhites 400 years ago, no need for HDR:

superwhites.jpg (129.79 KiB) Viewed 1081 times

flame2.jpg (105.9 KiB) Viewed 1051 times

It's all … no need for HDR

+1. I only think of HDR when it’s for vivid looking commercials on a TV/Monitor. More so often they looked plasticky.