AVS Forum banner

9581 - 9600 of 10194 Posts

·
Registered
Joined
·
560 Posts
hi, can you please post this in the MadVR support thread, this thread is dedicated specifically to a discussion on improvements with MadVR, thanks!
OK. Thought this thread would be the right one since its specific for HDR to SDR conversion and this error is specifically for HDR to SDR conversion.
 

·
Registered
Joined
·
139 Posts
Thought I would share that I DID in fact get my GTX 1650 Super card to work with Tone Mapping, with Compromise OFF as well - so fully working! Using beta 113.

The limiting factor was originally the Video Card RAM. 4GB gets filled very quickly - so I had to figure out a way to reduce VRAM overall (shout out to the doom9 forums guys for help on this!).

The answer was to reduce the ques and read ahead. I set CPU que to 8, GPU que to 4, and frames to present ahead to 3. This was CRITICAL in order to save a ton a VRAM. And these settings (for me) work smoothly!

This allowed me to uncheck compromise for tone/gamut, and uncheck compromise for luminescence. The only checks I have in the trade performance section are 'don't re-render during fade' and 'scale chroma separately if it saves performance'.

Other settings: Chroma = NGU AA Low, Luma = NGU Sharp High

For HDR to SDR I was able to copy NeoXP's settings (including highlight recovery and shadow recovery settings) from a couple of pages ago - Curve = NeoXP 1.0 High.

I am getting 'comfortable' rendering times of 30-34, maxing at 36 to 38. Just within the parameters.

Was able to watch a full 2 hour 4k remux movie, no glitches. Was also able to watch a 25FPS (planet earth II) without going too high on the rendering!

My other specs: Core i7-3770 (OLD!), 16GB of DDR3 ram, and 1TB Samsung SSD.

So Javs was right: You DON'T need a super computer to use MadVR for HDR to SDR.

PS - If you LIKE to use compromise ON for tone and gamut, this saves even more resources, and would allow me to go up to NGU HIGH for Chroma. Although I could never tell the difference between Low to High TBH.

Thank you everyone for the help and tips!
 

·
Registered
Joined
·
8,940 Posts
I’m also puzzled by your decision to not use some other options and to prefer the performance option that disables a lot of good stuff. We usually have fairly similar taste re picture accuracy, so I have no doubt that what you see and leads you to make these corrections is different from what I see here.

I have some issues with some of the contrast curves, including mine, so I use a very conservative one, otherwise it leads to brightness jumps etc that I find a lot more jarring than any minor oversaturation. I disable the shadow restore option because I never liked what it can do to contrast in order to restore shadow detail that I never felt was meant to be seen. I use a fairly moderate setting for restore highlights too. But I don’t experience the saturation issues you report.
I guess I will follow up on this. Please be aware, this is my preference and what guides my eyes and opinion on most things, like you, film also forms the basis of my day job so there are also things and looks I have been somewhat steered towards which stick with me through all this stuff. People such as yourself are absolutely free to disagree with me, and I am in no way trying to say its the 'only way'.

So, to me, apart from the hue correction which works perfectly (to me) back in beta 10, I dont see any really good stuff on my end (for me). Not to say it has no place, just dont really care for it nor feel like I need it.

I should firmly say that I have never been of the opinion that contrast should be boosted or manipulated very much past the way its presented in the frame, my arve curves ended up trying to retain that with quite shallow roll off also in the end (V3), so, I dont see anything wrong with the contrast in UHD at this time when using good target settings. I dont grade film or photos that way, so I dont expect HDR to fill the scopes either just because MadVR detects that there is still room, no just that, it appears to do its own curve in addition to tone mapping.

With contrast recovery we are potentially modifying intended headroom and balance, I try not to mess with that kind of thing, we can argue we are never hitting the intended brightness already, but it just never looked quite right to me, since at least visually, HSTM appears to be adjusting the shots histogram dynamically and non-linearly. I certainly do not put that in the basket of accuracy either. I actually prefer to run very mild dynamic clipping if anything, because I think that more effectively gets rid of some truly trash rogue pixels and seems to be better suited along with a good dynamic target setting and keeps to one proper tone mapping curve.

Some testing on my end is also showing that straight DTM with no other features ticked is already adding some contrast and saturation compared to the SDR counterpart already (at least on my test file - I will make more), and I will show this properly later, but just testing today with two photos tests I made is showing more contrast and a little more saturation in the HDR tone mapped version vs SDR both same exact colour gamuts encoded incl high target nits - even disable DTN and manually enter 2500 peak, this is a very bright test. Adding Contrast Recovery to the mix just adds even more contrast.

Shadow detail, we both agree just doesnt really seem to yield a good look (not arguing it has no place for some who need it), I never sit there feeling like I need it, I see the details I feel I need to, not really missing anything here. Interestingly though in my test, low/medium gets 'rid' of some of the inherent contrast boosting I am seeing in straight DTM, though I know from experience this option greatly affects certain shots vs others, so remembering that I still dont wish to use it.

Highlight recovery, I never really liked this option, it adds sharpening to specific areas as a by product and makes things a little out of balance with the rest of the image, in my opinion this removes 'depth' in the picture where its used, this has never looked quite right to me.

We just cannot eat the whole buffet with HDR when we dont have the brightness to hit the intended peaks, its good to have these options, but they really should be options, I put saturation recovery in the same basket, off will likely be my preference even if its proven that whats its doing now is correct, because I just dislike it and can see saturation being added vs my tests even high nit targets. I have a feeling though it will meet somewhere in the middle through testing, it still does not change the fact that people notice it and dont seem to like it just as me, so it really should be an option just as all the other things are.

Also, it will not mean things are suddenly not WCG if there is a bit less (not added) saturation, I am not sure where that train of thought comes from. We never thought the image lacked ultimate colour with the arve curves, this is no different, I am not asking for less saturation, just not more of it.

The Greatest Showman seems to be really excellent example of this colour boost, you should be able to clear as day see that, the thing thats even more interesting is the agency I used to work for just signed the director, so I am literally trying to organise to send him an email with a few images and ask him which one is right to him, he will certainly know, because in that particular shot, the differences are far from subtle. In my opinion the version from Beta 10 is by far the best and also closely matches the saturation intensity of the bluray (I dont believe that is an accident).
 

·
Premium Member
Joined
·
9,944 Posts
Thanks for taking the time to type this. I don't disagree with any of it, except the bluray part at the end. I never understood the obsession with matching bluray when looking at 4K UHD HDR content.

One of the great advantages of HDR is to provide more saturation not only because of the increased color gamut, but also because of the increased color volume, which is far more important. This is true even for colors inside rec-709, simply because on most displays the gamut shrinks as you go up in luminance. So a color that would look near-white on bluray (say a sky color that sits within rec-709) because you don't have the color volume would be more saturated on UHD bluray not because the color sits outside of rec-709, but because the luminance is so high that it would look desaturated on SDR and far more saturated on HDR. In that case, the higher saturation in HDR is perfectly accurate. Why would you want to tune it down to match bluray, where it's only the SDR limitation that desaturates it, and not the mastering intent?

So when comparing to bluray, how do you take the color volume into account? Unless you know that it's a color that can be fully saturated in SDR at that luminance level on that display, you are simply undersaturating HDR to match the understaturated SDR. That's 100% wrong in my view because the undersaturation is not by intent, it's due to the limitation of SDR. I completely agree taking a 4,000nits or even 1,000nits pro HDR monitor as a reference (as long as the content isn't masteredt above 4,000 or 1,000nits respectively). I 100% disagree on using bluray as a reference, unless you know for sure that the bluray saturation you are comparing the HDR content to isn't the result of either the gamut cover or the color volume limitations of SDR (sometimes both), which is tricky to assess on content that you haven't graded yourself.

The only colors that can be compared between bluray and uhd bluray are black and white. :)

Any other color could be undersaturated on bluray due to a color volume limitation and not necessarily grading intent, so isn't in my opinion a valid point of reference.

This is why I was relieved to hear that you were hiring pro high-nits grading HDR monitors for your tests, as this is the only way to have a reference HDR picture (assuming they are calibrated properly, which I know you'll make sure of).

In any case, please could you share your settings and the Greatest Showman test timecode? I'll give them a try when I get a chance, I remember doing some tests when you first came up with this issue (The Dark Night, possibly Interstellar) but I wasn't convinced as the issue here looked more in the realm of subjective than something obviously wrong, and the differences here (with 120nits and full P3 cover, using a 21^3 equivalent P3 3D LUT with profile off / filter on so no CMS active on the JVC) were really minute. Having the director's opinion on an obvious test should provide a great indication (as long as you're sending HDR screenshots and he's watching them on a calibrated HDR monitor), so looking forward to that data point.

By the way, none of the above means that saturation in madVR doesn't need a pass, it does and we were patiently waiting for that round. Only that I still have to find a scene where it's really offensive and takes me out of the film. To me, this happens all the time if the contrast enhancement curve is too strong (for example mine, which I stopped using a long time ago!), not only because of the too strong "enhancement" in some scenes but also because of the brightness variations, which I find a lot more disturbing. I suspect that most people complaining of brightness variations in madVR/Envy are using a too aggressive contrast curve in madVR (or the high setting in Envy, that I personally would disable as it causes these brightness variations), especially with low peak nits. These artifacts are super annoying, because they are objectively wrong. Color saturation might be a bit off occasionally, but again nothing here that would cause me to notice when watching a film. On the other hand, although we both work in film, grading is your job, so I have no difficulty accepting that you are 1) more sensitive to this issue and 2) more competent than I am in this area.

So I'm really looking forward to your testing with reference HDR monitors, because even if I'm not annoyed by the minor saturation issues, it doesn't mean that what I'm watching is as accurate as it could be, which is really what I want. Not too much saturation, but no undersaturation either. I'm sure we have the same goal (picture accuracy to the original intent, within the limitations of our brightness starved projectors), so as long as the methodology is correct, I'm sure we'll both be happy with the results. :)
 

·
Registered
Joined
·
8,940 Posts
Thanks for taking the time to type this. I don't disagree with any of it, except the bluray part at the end. I never understood the obsession with matching bluray when looking at 4K UHD HDR content.
I would generally agree with your whole post, but in this case, there is a common thread here, you can definitely check the bluray to see if you are in the right 'ball park'. In this case its significantly different.

Just have a look at the images, one (two) of them doesnt belong as they used to say in Seasame Street :)

This is why I wanted to reach out to the director, no I dont have to send him HDR images, I can send these, its very simple, this colour grade is SO far from Comp On and Beta 10 Hue only correction that I think he would instantly know its either what he wanted or not. Desat 1+2 makes it better, but not enough, and yeah I will be the first to admit I am really sensitive to this stuff, thats why I happened to watch two films in one night back to back which are both very red so both my Wife and I noticed something was up.

This is 120 nits peak, 50 nits dtn, these images are rec709 conversion to share here, but you can check any shot in this whole sequence, its there in P3 / LUT etc. One of the things I also want to deep dive is if there is a problem with Rec709 conversion, before you jump on me for that :) but also know that with comp on and beta 10, there is no issue with the saturation in the conversion so it leads me to believe its just global in the Comp Off DTM.

Timecode, 50:09.

Bluray



Hue Perfect Beta 10. - I dont think its a mistake that this is nearly identical to the overall look and feel of the colour palette of the bluray. The colour will still be slightly more 'rich' in WCG mode no doubt, but not a smack in the face difference.



Compromise On



Compromise Off Dont Desat



Compromise Off Desat 1+2



Compromise Off in both cases, are both very different palettes, this is to me such a difference its like a whole different brief for the colorist. I just dont buy that its meant to look anything like this. there should be a general consistency across the colour in all mediums this is usually very important to the film makers..

On my Panasonic FZ950 this is more more like Beta 10, almost exactly identical actually... That OLED can do over 700 nits, this is an 800 nits shot. I dont think I am far off the mark with that display as it is. I can take a photo of it if yhou like. With the Compromise Off option, I would have to use Desat 1+2 and also use -15 on saturation in the color & gamma setting tab.

I dont have it on hand sorry, but that Panasonic was so good out of the box I literally didn't even bother calibrate it, here is what is does In Rec709, P3 was about the same accuracy, it was stunning. Vincent Teoh gave this OLED a really good review, said the same thing. Of course an HDR Ref monitor is the ultimate, but right now, my FZ950 is telling me this shot is supposed to look like Beta 10. If anything the Panasonic is the slightest bit more on the pale skin side of things.

3050916


Bad phone photo of my Panasonic FZ950 in native HDR mode, I can assure you my phone is actually adding even more saturation here (Samsung). Its 90% the same as Beta 10. Pale skin all day long.

 

·
Premium Member
Joined
·
9,944 Posts
Thanks a lot for the above.

And I agree with you, in that example it's unlikely to be a color volume difference.

I also agree 100% based on these SDR rec-709 screenshots that there is an issue with compromise off.

Now what I'm interested in is what I see on my projection screen when displaying HDR content tonemapped to my P3 calibrated PJ, so as soon as I get a chance I'll take another look and see how close or how far it is from your screenshots.

Please could you post all the relevant madVR settings so that I can get as close to what you see in case what I see here isn't as bad as what you see?

Chroma upscaling, dithering, any processing, DTM settings, etc.

I assume Windows 10 / GTX / RGB 8 bits 4:4:4 and 2.2 gamma for baseline and LUT target.

Please don't spend any time on convincing me, we're really on the same page and after the same thing. I just don't want all the work you're going to do (which is precious and for which we are all grateful) to be done on a basis that leads to wrong results because of low nits, reduced gamut and so on.

I trust your eyes, so if you base your observations on reference HDR monitors with enough brightness to match the mastering brightness and full P3 cover, and if the changes you recommend lead to a picture that is closer to that reference with our brightness starved projectors with at least 100nits peak and full P3 cover, I will 100% trust your recommendations as I can't reproduce that reference here :)

Anything that needs to be done because the brightness is even lower than 100nits or the gamut doesn't cover 100% of P3 should be a different setting.

We want to optimize HDR tonemapping first for HDR able projectors (full P3 cover, around 106nits, which is Dolby Cinema peak), then for further hobbled projectors :)
 

·
Registered
Joined
·
8,940 Posts
Thanks a lot for the above.

And I agree with you, in that example it's unlikely to be a color volume difference.

I also agree 100% based on these SDR rec-709 screenshots that there is an issue with compromise off.

Now what I'm interested in is what I see on my projection screen when displaying HDR content tonemapped to my P3 calibrated PJ, so as soon as I get a chance I'll take another look and see how close or how far it is from your screenshots.

Please could you post all the relevant madVR settings so that I can get as close to what you see in case what I see here isn't as bad as what you see?

Chroma upscaling, dithering, any processing, DTM settings, etc.

I assume Windows 10 / GTX / RGB 8 bits 4:4:4 and 2.2 gamma for baseline and LUT target.

Please don't spend any time on convincing me, we're really on the same page and after the same thing. I just don't want all the work you're going to do (which is precious and for which we are all grateful) to be done on a basis that leads to wrong results because of low nits, reduced gamut and so on.

I trust your eyes, so if you base your observations on reference HDR monitors with enough brightness to match the mastering brightness and full P3 cover, and if the changes you recommend lead to a picture that is closer to that reference with our brightness starved projectors with at least 100nits peak and full P3 cover, I will 100% trust your recommendations as I can't reproduce that reference here :)

Anything that needs to be done because the brightness is even lower than 100nits or the gamut doesn't cover 100% of P3 should be a different setting.

We want to optimize HDR tonemapping first for HDR able projectors (full P3 cover, around 106nits, which is Dolby Cinema peak), then for hobbled projectors :)
Sure, will get to it tomorrow, late here now.
 

·
Registered
JVC RS4500 | ST130 G4 135" | MRX 720 | MC303 MC152 | 6.1.4: B&W 802D3, 805D3, 702S2 | 4x15 IB Subs
Joined
·
9,274 Posts
Man none of that over saturation looks too bad to me until you bring skin tones in and then it just looks really bad :(
 

·
Registered
Joined
·
1,241 Posts
Hue Perfect Beta 10. - I dont think its a mistake that this is nearly identical to the overall look and feel of the colour palette of the bluray. The colour will still be slightly more 'rich' in WCG mode no doubt, but not a smack in the face difference.

This one looks the best IMO, by far.

@madshi any good news to resume madVR's testing?
 

·
Registered
JVC NX5 at 140", Denon X4200W (5.1.2) with Axiom Audio speakers + Bass Shakers
Joined
·
2,183 Posts
This one looks the best IMO, by far.

@madshi any good news to resume madVR's testing?
Yeah, the hue shift from compromise off is what keeps me from using compromise off. The skin just looks more unnatural and sort of more towards green rather than pink.

But with it on and newer builds, it is soo saturated and like sunburn level, thought it's not as bad with desaturate modes which is what I use.
 

·
Registered
Joined
·
139 Posts
If I set Contrast Recovery to Disabled - does that mean I am using the standard BT.2390 curve? Or is there no way to set that curve within the beta?

Just trying to use BT.2390 as a 'baseline' to test the other recovery options against.
 

·
Registered
Joined
·
380 Posts
I know that somewhere I could find the settings, but does anybody has a recommended settings for a LG OLED65CX9LA, which has basically 700nits??
 

·
Registered
Joined
·
1 Posts
I recently got a new laptop with windows 10 and a gtx 1650ti mobile.
Which nvidia driver version is recommended to use (or which ones better not to use) with madvr?

And which madvr beta us the better one to use currently? 112b or 113? I guess for non-HDR stuff it doesnt matter anyway?
 

·
Registered
Joined
·
139 Posts
Finally sat down and just enjoyed watching some movies.

I feel like I have my PEAK nits set right - bright scenes are BRIGHT - and look amazing!

DARK scenes also look good, blacks are not washed out etc.

It's the 'middle' scenes that seem..... too dark to me... almost like my eyes feel 'strained' while watching. Is this something that the dynamic nits setting could help? Right now I am running 100 peak, 50 dynamic. If I upped it to 60, would that increase or decrease average brightness?
 

·
Registered
Joined
·
8,940 Posts
Finally sat down and just enjoyed watching some movies.

I feel like I have my PEAK nits set right - bright scenes are BRIGHT - and look amazing!

DARK scenes also look good, blacks are not washed out etc.

It's the 'middle' scenes that seem..... too dark to me... almost like my eyes feel 'strained' while watching. Is this something that the dynamic nits setting could help? Right now I am running 100 peak, 50 dynamic. If I upped it to 60, would that increase or decrease average brightness?
try 45 dynamic target nits.

higher you go the darker.
 

·
Registered
Joined
·
559 Posts
Some testing on my end is also showing that straight DTM with no other features ticked is already adding some contrast and saturation compared to the SDR counterpart already (at least on my test file - I will make more), and I will show this properly later, but just testing today with two photos tests I made is showing more contrast and a little more saturation in the HDR tone mapped version vs SDR both same exact colour gamuts encoded incl high target nits - even disable DTN and manually enter 2500 peak, this is a very bright test. Adding Contrast Recovery to the mix just adds even more contrast.
How did you create you test patterns? Can you tell us more about them?

I compared many Blu-rays with their UHD counterparts. Some had very similar color grading and steepness and some looked completely different. It is nearly impossible to compare if the grading intent is unknown. Well-defined test pattern may be more reliable to test the tone mapping. I'm interested what you did here.
 
9581 - 9600 of 10194 Posts
Top