True Lies HDTV (Film4 HD) Preservation

[deleted user]

Ah that's good to hear

But you still need an HDR monitor to do that grading and to watch the result - I personally don't have one

And I'm pretty sure the scans are not done in AVC - that's just how they deliver 'em - though even a ProRes 10 bit file in some log-colorpsace should do fine I think

[Stamper]

Actually no, some monitors with DVI accept HDR input. It doesn't show the whole dynamic range, but once you export to Rec709 the colors are accurate to what you saw on the monitor. The only thing you have to do once it's in Rec space is adjust the contrast down a bit like you would do in any 2020 to 709 conversion.

If you do your telecine with dng, then the possibilities of regrading are nearly endless, they are multiplied by 12 in comparison to prores or compressed codecs.

[deleted user]

I don't understand what's the point of editing it in HDR colorspace without an HDR monitor, much less when using Rec709 as output. You need to at least be able to see how it actually looks in HDR to properly adjust basic values like contrast/exposure etc. Yes, you can convert to Rec709 with various methods and it will look O.K., but it will not be an accurate representation of what it would have looked like in HDR, so you may as well do the color correction directly in Rec709, to not create some kind of Frankenstein 2020 that may be unusuable in the end anyway. It would be like mixing a CD on loudspeakers that only go up to 2 kHz.

[Valeyard]

I love the 35mm but clearly the colors aren't 100% right due to it being a scan and not shown through a projector. It has that grey look all 35mm releases have, like watching the film projected from behind on a grey screen.

What do you mean by "grey look"? Generally speaking 35mm prints don't have rich blacks like digital film does and in a cinema you'll notice the blacks are grey not a deep black.

[zoidberg]

Usually the scanner will project the scanned film as a reference to adjusting the scanned footage.
Don't forget 35mm film is timed for projection onto a highly retroreflective screen material with a target brightness off the screen of 14fL (16fL open gate). Most theatres never even got close to that brightness!

[Beber]

I'm not fully convinced by the brightness of the 35mm scan, especially in the beginning. I've not watched the whole thing yet.
I cannot believe that the 1st revealing shot of Arnie, the movie star, being so dark that you cannot even tell it's him.





Same thing inside, in the ballroom. How don't people bump into each other in there? Can't rich people in Zwitzerland afford extra lightbulbs, especially when hosting a party?

[Stamper]

About HDR, thought it may seem color timing for HDR is not very good idea since you can't replay it unless you burn it to 4K HDR. However, the difference between first grading in HDR, then converting in Rec709 is huge. All you clipped whites in direct 709 will show detail if you original adjusted them in 2020 then converted back. Just try it but remember you need to use at least 12bits footage. Also your monitor should have DVI, if not, resolve will not show "HDR" near your nodes when you switch to HDR.
Beber you're right, Here yes, right. The print is a bit dark. I think it should be color matched to the DVHS or your NRJ12 recording. The right colors might be not far off maybe a hybrid of both.

[deleted user]

Ah I see what you might be getting at. There are different ways of converting HDR to SDR. I don't know that much about it, but I know Photoshop has these too, one of them being Tone Mapping. With tone mapping you don't get overexposed stuff, that's true, but you also lose faithfulness, as different areas in the image do not have the correct brightness in relation to other areas. You're basically leaving the area of a faithful preservation and moving towards a creative reimagining, because getting the wider dynamic range back from the tone mapped version is probably close to impossible and you will have a forever changed image in style. Either way, you could achieve the same effect with any plugin that is coded with tone mapping in mind. Dunno if it exists though, so maybe the HDR workaround is not a terrible solution for that kind of approach. But even if this allows for a nice SDR grading, it won't give you an accurate HDR video in any real sense, because all you see is the HDR->SDR converted image (unless you get lucky and it happens to look alright). I think that as long as HDR is not a common format in most households, the current method of compressing the dynamic range (creating a low saturation image as a result) is the most future-proof way.

Plus, I am pretty sure the scans they did were all 16 bit TIFFs, and once HDR is here for good, those can be used to create a good HDR version, as long as they don't already have clipping/crushed blacks. 16 bit, after all, theoretically has 16 times the fidelity of 12 bit (e.g. Dolby Vision) and 64 times the fidelity of normal HDR10. You lose some of that during grading, of course, but it should still be more than enough to create solid HDR versions without oversampling the source.

[Stamper]

Yes it's not about creating a HDR footage, it's about using the extended range of HDR before switching back to regular rec709. It's a bit if you like, when you scan a photo from a magazine at regular size vs 12 times the size then reducing it back to actual size. The scan with more depth will look better than the 1/1. Or when they used to do HD scan in order to release DVD masters. They all looked great compared to SD only masters.

[deleted user]

Hmm, I don't see how that would give a similar advantage. They are already doing something like that by using 16 bit TIFFs as source. There is nothing in the conversion to HDR and a conversion back down to SD that would yield any benefit I can think of. If there indeed is a better image as a result, it's likely through some inaccurate or at least doubtworthy downconversion method that creates some kind of good looking falsification meant to create a viewable SDR version of a video intended for HDR displays, possibly with lots of psychovisual modelling approaches and whatnot.

Not sure which software they use, but After Effects for instance fully supports working even in 32 bit (!!) floating point colorspace. In fact, it has supported this for years before HDR was a thing (I have used the software for around 10 years). This also has to be converted down for an 8-bit output, and you can probably choose between rounding and dithering methods, but the difference would be very subtle, as in visible in very fine gradients.

Check here: https://wolfcrow.com/blog/how-to-handle-...peedgrade/

In other words, what you're suggesting (using a high resolution source and then subsampling that), it's already being done if the source is 16 bit. Using a HDR colorspace for it and then converting to SDR can do absolutely nothing to increase the fidelity of the result, as in being true to the source. An HDR colorspace will not improve the source in any way, it will just do some matrix transforms or whatever.