Best way to preserve SDR footage in an HDR timeline

I frequently mix SDR clips into projects with a rec2020 ST2084 1000 nit timeline intended for HDR PQ uploads to YouTube. The problem is that the rec 709 clips seem to adapt to HDR levels by default with peaks hitting 1000:

https://ibb.co/bM9fNZfR

I want the rec 709 footage and srgb images to remain accurate to 100-nit SDR levels- one example where this would be critical is comparing the HD Blu-ray release of a film to the HDR 4k release, where both clips would need to accurately reflect the source media as it would appear on the viewer's screen when played from the disc.

The best method I've found so far is to use a color space transform with only the input gamma changed to rec 709 (also changing the input color space results in major undersaturation). This is with the input color space of the clip itself being unaltered from the default rec 709 Resolve sets upon import:

https://ibb.co/vx8c2yR8

I've also tried setting the input color space of the clip in the media pool to match the timeline space, then doing a full color space transform with all conversion options set. The results seem similar:

https://ibb.co/Dg9B0Ssj

I've also experimented with no color space transformations at all and only the "use custom max. output" setting at 100 nits, keeping the input space at rec 709, though this seems to lead to clipping on highlights.

The ACES transform option produces duller looking results that fall below 100 nits:

https://ibb.co/rGkcrwB7

And finally I've tried using the LUT option for ST 2084 "gamma 2.4 to HDR 1000 nits" but it tends to produce even dimmer results that peak around 60 nits.

I've also experimented with different project settings, using automatic color management or fully custom settings:
https://ibb.co/GNnwcrJ

Is there a simple and reliable way to accurately preserve the SDR appearance and levels of these clips in an HDR timeline that I'm missing? Thanks in advance to anyone who can clarify things.


For more information, I'm monitoring in HDR and using clips from many disparate sources- some footage in my next project is Panasonic V-log from a Lumix S5II, some clips are direct mkv rips from a blu ray or 4k disc, some game footage was captured in HDR using Nvidia Shadowplay (which mediainfo lists as rec2020 PQ), and some gameplay footage is being captured from an SDR console using a capture card set to rec 709, and there are a few standard srgb jpg images from the web.

I frequently mix SDR clips into projects with a rec2020 ST2084 1000 nit timeline intended for HDR PQ uploads to YouTube. The problem is that the rec 709 clips seem to adapt to HDR levels by default with peaks hitting 1000...

I think if you did the whole projects with CST nodes, you wouldn't run into this problem. You can always bypass color management very easily this way.

I think it is dicey to mix HDR and SDR clips in the same timeline, but it would only take a second to make sure the color management was ignoring the nit level and just leaving it alone, if that's what you wanted. Usually though, in (say) an HDR documentary, we do routinely bump the SDR material up "a little" so it doesn't stick out too much. It's purely a creative decision.

I can't give you a simple answer, but here's two.
scenario 1: color grading/feature/documentary/fictional/whatever --> creative component
If you want to use the project color management I'd select the HDR DaVinci preset (only one that makes sense tbh), then open the list again and select Custom. This way it will select all the settings from that preset, but let's you change them.
I would now disable the input tone mapping (None instead of DaVinci) and leave everything else as is
1. benefit, you don't get any stupid non-linearities by input tone mapping into the DaVinci timeline color space
2. benefit, you can add input tonemapping for selected clips as a first node in the tree with a CST or Gamut Mapping OFX (doesn't really matter, except the OOTF knobs in the Gamut Mapping OFX don't work for whatever reason, if that's important). You can now map them exactly where you want. I'd recommend setting the input nit value to 100 and playing with the output nit value.

scenario 2: comparing the same film but different deliveries
Just go node based. You have way better control - or let's say, it's way easier to understand what is going on.
Actually if you would just want to see a Rec709 grade in a PQ 1000nits container, you wouldn't need any tonemapping at all. the tonemapping is intended to distort the image so it can work in the desired color space, but that will of course not help you with the analysis you're trying to achieve. If I understand correctly you want to see what the differences are between SDR grading and HDR grading, right?

A note on youtube HDR:
The way youtube implements HDR is a bit screwed. It works ok-ish, if the display is HDR capable. But if it can only show SDR, youtube will apply its own method of tonemapping the HDR file to SDR again. This method is just ... ugly ... it introduces wild oversaturation, unpleasing contrast tones and overall just does not look nice.
Does this benefit your intent of comparing SDR and HDR grades? For anyone with an SDR display .. I'll let you answer that.

Thanks for the advice, setting the color management mode to custom and then input DRT to "none" stopped the HDR stretching on SDR clips.

I also tried getting captures from an SDR source with my capture card set to rec2020 HDR, which allowed me to import rec709 clips in an HDR container that would not be tonemapped. I used those as a reference to compare against the clips that had been captured in rec709 and I was able to confirm the color space transforms I had been using seemed accurate, but I trust this new method of just disabling the input DRT more.

A note on youtube HDR:
The way youtube implements HDR is a bit screwed. It works ok-ish, if the display is HDR capable. But if it can only show SDR, youtube will apply its own method of tonemapping the HDR file to SDR again. This method is just ... ugly ... it introduces wild oversaturation, unpleasing contrast tones and overall just does not look nice.
Does this benefit your intent of comparing SDR and HDR grades? For anyone with an SDR display .. I'll let you answer that.

I've been using the HDR metadata tool that YT recommends for this. I export a simple rec2020 ST 2084 1000 to rec709 color transform as a .cube file that I can embed in the file before uploading and supposedly YT will rely on that rather than its own tonemapping.
HDR on YT is kind of a mess, especially with browser support still being patchy, so that's probably the best I can do apart from offering the actual exported file itself directly for downloads.

…A note on youtube HDR:
The way youtube implements HDR is a bit screwed. It works ok-ish, if the display is HDR capable. But if it can only show SDR, youtube will apply its own method of tonemapping the HDR file to SDR again. This method is just ... ugly ... it introduces wild oversaturation, unpleasing contrast tones and overall just does not look nice...

Up front, I’m not grading for broadcast so I don’t need to conform to traditional broadcast standards.

For my clients, I generate separate deliverables for HDR PQ Rec.2100 P3 ST2084 for 4KTV and recent laptops, iPads and iPhones since iPhone-3 that support P3 and ST2084 ‘reference’ mode, SDR Rec.1886 for traditional HDTV, and SDR sRGB for web viewing on older monitors. The user can select the version suitable to the screens they have. For several years, my mostly private videos are only on Vimeo, not YouTube. Recently one client has wanted their videos on YouTube and open to the public which I’m in the process of doing.

Having three versions of deliverables from three Resolve timelines for each project is my simple approach. I grade the first timeline for 4KTV PQ with diffuse white point generally 200 nits but it can go up to 300 as the latest OLED 4K TVs support that with peak levels up to 1000 nits; no longer religiously need to set diffuse white for OLED to 100 nits. For sRGB monitors diffuse white can go well beyond 100 nits with it becoming common for recent gear to support 500 nits maximum brightness.

My clients are used to this approach, but it remains to be seen if this will succeed on YouTube. Your feedback is appreciated. I gave up on expecting the streaming services be able to display a single grade well on every device. It can be done but it’s not going take advantages of the unique capabilities of 4KTV and HDR P3 devices, HDTV, sRGB.