madVR - high quality video renderer (GPU assisted)

[huhn]

without measuring you will always be guess.

the idea is pretty simple.
HDR wants more max nits and the higher the max nits are the more useless tone mapping gets(simplified deal with it).
madVR tone mapping should in theory always when when you have the same SDR nits as HDR nits but you can't figure this out without measuring tools and you usually really really don't want to run your screen at such high nits.
if you add output in HDR you are now adding two tone mapping step with can ruin everything.

TL;DR do it by eye if you can't calibrate the screen.

[madjock]

Thanks.

Yes I guess thats true ref calibration, I think in my mind I thought tonemapping is better because at least with SDR you can calibrate it as such, where as HDR is another animal.

So if my max nits is 350, is that what I have to set it at or lower ?

https://www.displayspecifications.com/en/model-display/e0a010c1

[mclingo]

https://imgur.com/a/wQjETu3[/url]

lol. i sent an old screen shot thats all :)

I think if people have a high nits OLED then maybe the TV's own tone mapping might be better, my 400 nit OLED definately benefits from madvr's tone mapping, certainly for highlight recovery.

The other issue is that even though there is a lot of HDR stuff out there now, very little has any real pop to take advantage of high nit units i've found. My dad has a q85 and the few things I watched didnt look that much different to me in a dark room.

One thing that did really pop I watched recently was the latest season of the expanse but I saw that on amazon prime so not through MADVR, i've never seen HDR look so good, the pop from the lights and colours were amazing even on my TV, nothing else looks this good on my TV but I think this is more to do with the fact that a lot of HDR movies out there are fake or poorly mastered HDR.

It would be interesting to see how the expanse compares on hdr bluray via MADVR to amazon prime, this is something i'm eager to test.

[huhn]

Thanks.

Yes I guess thats true ref calibration, I think in my mind I thought tonemapping is better because at least with SDR you can calibrate it as such, where as HDR is another animal.

So if my max nits is 350, is that what I have to set it at or lower ?

https://www.displayspecifications.com/en/model-display/e0a010c1

if you pump the backlight and the brightness to max maybe.

[el Filou]

@madjock your supposedly v113 screenshot isn't. mclingo's is indeed 113. (shame madshi's not including the test build number in the GUI)if you add output in HDR you are now adding two tone mapping step with can ruin everything.Depends on the display really. If you use a low enough real peak nits value you can possibly completely bypass the display's tonemapping and not only avoid ruining anything but get a better result.So if my max nits is 350, is that what I have to set it at or lower ?You need to find out what your actual contrast ratio is at different white nits values. If you push the display to 350 but black level rises in the same proportions then it's much less interesting because you don't gain any real dynamic range.

[huhn]

you get more >dynamic< range that's the point of the active backlight.

the CR stays the "same" if you run your new HDR TV at 1000 nits or at 100 nits.

Depends on the display really. If you use a low enough real peak nits value you can possibly completely bypass the display's tonemapping and not only avoid ruining anything but get a better result.

there is a reason i said it can ruin everything not it will. with the aggressive back light and ABL of modern TVs is more then unlikely that you will ever find a way to disable the tone mapping completely while keeping the same peak nits.

[madjock]

How can his be 113, it has no mention of HSTM and 3 rows of numbers ?

Thats what Neo-XPs recent numbers are for ?

https://www.avsforum.com/forum/24-digital-hi-end-projectors-3-000-usd-msrp/2954506-improving-madvr-hdr-sdr-mapping-projector-293.html#post59248256

He has quoted my one bizarrely.

[el Filou]

you get more >dynamic< range that's the point of the active backlight. the CR stays the "same" if you run your new HDR TV at 1000 nits or at 100 nits.Of course, but this assumes the display has efficient local dimming, which sadly isn't always the case when it comes to LCDs, especially monitors and low- to mid-end TVs.How can his be 113, it has no mention of HSTM and 3 rows of numbers ?I just installed v113 so I'm sure of that. There are no rows of numbers for curve values anymore and no mention of 'HSTM' either, it's been renamed to 'contrast recovery' with the different (non-tunable) curves available via a drop-down list (set to Hill Strong in mclingo's screenshot).
(Edit: check your madvr.ax date, v113 is 2020-01-20)

[huhn]

Of course, but this assumes the display has efficient local dimming, which sadly isn't always the case when it comes to LCDs, especially monitors and low- to mid-end TVs.

never seen an TV in the past 10 years that can't do that. the cheap backlights even have an edge they will not create holos black will simply not look black but it's an LCD.

what so ever his linked TV should be able to do that.

[madjock]

Ah ok well I guess Neo-XP says he has not used that version due to smooth motion causing issues, will try it now. Thanks.

[tramin]

Hi, I need some advice.I have a PC with Radeon 570 and Philips OLED 804.I only watch UHD movies and at a faster scenes bothers me like the blurred outlines of figures.As if they were wrapped in bubbles.I use LAV and madVR(chroma upscaling Anti-Alias medium quality).It can be modified somehow with madVR?I don't know if I wrote it clearly,thanks.

[mclingo]

OLEDs can look bad in faster scenes due to their very fast response time, there is no frame blurring. You can try smooth motion in MADVR, that does help a bit but it doesn't work for some who are using the latest 113beta. You could also try turning on a little motion interpolation, I use a very small amount on my LG OLED and it make a huge difference, although I dont really recognize your description of the issue "bubbles", for me there are jagged edges to lighter areas on fast pans., this as very much reduced with a little bit of LG's tru motion turned on, if you use too much it gets all soap opera, I use 2/10, it looks unwatchable awful anything over 4/10

[Octo-puss]

Is madVR dead? Not that it doesn't work, but with no updates in what, two years?... it seems quite abandoned.

[SamuriHL]

Only if you don't read...

[ryrynz]

Yeah you can't expect to disappear yourself for years (last post 2013) ignoring all conversation and ask a question like that.
Educate yourself. (https://www.avsforum.com/forum/24-digital-hi-end-projectors-3-000-usd-msrp/2954506-improving-madvr-hdr-sdr-mapping-projector.html)

[tp4tissue]

Thank you.

I still do not understand as such and for a layman like myself with no calibration equipment and the same in depth knowledge as yourselves I find it all another language.

A lot seems to be based on OLEDs, which is also something I do not have, not 100% what my real NITs is also which does not help.

edit : so do I need to output in HDR format ?


I highly recommend you buy a colorimeter, either the i1 display pro, or colormunki display, both will work identically for madvr 3dlut purposes, the i1dpro will do it a bit faster, therefore having a slight edge in accuracy, because the longer you take measurements, the less accurate it is.

It doesn't matter how good/bad the display is, without a colorimeter, the workflow is BLIND.

The computer has no idea what it's outputing, madvr has no idea, the TV has no idea, You're getting the end result of several different machines throwing darts blindly at the target, COMPOUNDING their errors one after the other.


Movies without 3Dlut correction arn't worth watching.


The problem with OLED, is that because of the burnin problems, they have to implement a bunch of Tricks to compensate, such as ABL ontop of Secret-Sauce image tuning.

This makes the display very non-linear. Even after you measure the patches, the behavior of those patches do not fully correspond to real image output.

The advantage of oled having infinity contrast, is that it always manage to deliver a picture Pop.

LCDs are much more predictabably behaved, there's less distortion to worry about.

The downside is, LCDs (VA) beyond 300nits, really start to eat into the near black details. Most of the time this isn't an issue, but when you get to the 500nit-700nit panels, those dimming zones do terrible things to the image, imo this is worse than what ABL on OLEDs do.

For me personally, locking a VA panel down to ~350-400 nits, is a good balance for accuracy, just enough black detail, and solid post calibration response.

I really like Oleds, but I had burnins and bad retention after only a month of ownership, I decided it's just not for me. I'm not walking on glass for my TV. Tvs are meant to be watched, not chain you to a bed.

LMCL LCD will rectify most of these issues, and blow away everything on the market including oleds, quarter 3 of this year.
No burnin , ~100,000:1 contrast ratio.


Specific to should you output in HDR, if the TV is fairly linear, and doesn't do weird stuff, and the calibration comes out good without anomalies, then yea.

That's the problem though, because MOST tvs do weird stuff in HDR mode. The calibration and profiling becomes problematic. This stuff messes with Madvr's assumptions about the panel.

HDR at the moment is Cart before the Horse, Neither the software nor hardware are ready. They're even implementing dynamic tonemapping relative to Room lighting. That's kind of like taking HDR and going back to SDR. Because now they're walking away from the absolute tone curve.

BT1886 looks like shit, and that's kind of what they're bringing back to HDR..

[tramin]

@mclingo : Maybe I was wrong. I am concerned with blurred outlines such as the head and face when moving more sharply.

[madjock]

@tp4tissue

Cheers. Yes I appreciate all that and it all makes sense.

The reality is when your boiler has just broken and you have warning signals appear on your car dash it drops down the list a bit.

To be honest I find it hard to justify an expense like that that will only be used very rarely as well. ignorance is bliss in some ways, also TVs in general there are so many settings that you could spend days with that/madVR settings etc etc, and you end up watching nothing. I just need to suck it up and enjoy as is.

[tp4tissue]

@tp4tissue

Cheers. Yes I appreciate all that and it all makes sense.

The reality is when your boiler has just broken and you have warning signals appear on your car dash it drops down the list a bit.

To be honest I find it hard to justify an expense like that that will only be used very rarely as well. ignorance is bliss in some ways, also TVs in general there are so many settings that you could spend days with that/madVR settings etc etc, and you end up watching nothing. I just need to suck it up and enjoy as is.

Between all us guyzus on the doom9 forum, we know almost everything there is to know about calibrating tvs. Yea, you can spend maybe 2-3 weeks, but I'd like to think this is a fun experience, and brings you closer to the whole point of seeing the creator's intent.

The entire movie/tv industry runs calibrated, EXCEPT when it comes home, to the customer, who pay for them to do all that work.

The probe is not 1 time use, you have to retune every 300-500 hours of usage for accuracy.

You can also use the probe on your other monitors and laptops.

I want to say, you can also impress your girlfriend by calibrating her TV/laptop, buhhhh this is a very small subset of girlfriends who would be impressed by that. :D, nonetheless, 3DLut is madvr's greatest most powerful feature, night and day difference, once you see it, you'll never go back..

Of course then there's the conspiracy section, they don't promote calibration so people would buy new tvs every 2-3 years. :angry:

If you divide the cost of the colorimeter by the number of display calibrations you'll do on multiple panels, multiple times, it's chump change vs what you've already spent on video gear.

[tp4tissue]

@mclingo : Maybe I was wrong. I am concerned with blurred outlines such as the head and face when moving more sharply.

The only display that doesn't have this problem is CRT.

Even the fastest lightboosted IPS panels still have some blurring.

We're getting there though with the latest 240hz strobed IPS iterations, but IPS image quality is disgusting and unusable for movies/tvs.

OLED's sample and hold will always produce blur. In the very distant future, theoretically, OLED can be strobed to produce a CRT like image. However it will almost certainly never be able to do this in HDR brightness range, only SDR, and we're talking very distant future on a completely different chemistry, built from completely different factories.

[mclingo]

@mclingo : Maybe I was wrong. I am concerned with blurred outlines such as the head and face when moving more sharply.

Hi, we are probably talking about the same thing, try what I suggested, dont worry about what other people say about these two options, its about what works for you, some people would rather sell their kids then use any frame interpolation or smoothing :)

[huhn]

The only display that doesn't have this problem is CRT.

Even the fastest lightboosted IPS panels still have some blurring.

We're getting there though with the latest 240hz strobed IPS iterations, but IPS image quality is disgusting and unusable for movies/tvs.

OLED's sample and hold will always produce blur. In the very distant future, theoretically, OLED can be strobed to produce a CRT like image. However it will almost certainly never be able to do this in HDR brightness range, only SDR, and we're talking very distant future on a completely different chemistry, built from completely different factories.

can you please stop this nonsense.
it's not only your under standing of eye tracking motion blur and random IPS...
or your mis judgement on patch sizes it is still recommended to sue the biggest number of patches you can handle...

[tp4tissue]

can you please stop this nonsense.
it's not only your under standing of eye tracking motion blur and random IPS...
or your mis judgement on patch sizes it is still recommended to sue the biggest number of patches you can handle...

huhn, I understand you've been on this forum a while. And you've answered questions for me, and I am grateful.

I've avoided calling you out on your BS. I say this respectfully, You don't know everything.

It is NEVER recommended to exceed 1 hour time on consumer calibration equipment, unless the 100% saturation target is anomalous, where extra patches may be necessary.

The i1dp can only do ~1500-2000 patches in that span. The longer you let it run, the less coherence the measurements have. The meter's sensor surface may also heat up too much in warmer seasons/ buildings, causing alot of measurement drift. You've personally touted and bragged about using 10,000 patches, this is mistake, you should not be doing that. Making a number go higher isn't instantly better. 10,000 patches take ~4-6 hours to complete, no consumer display or consumer meter can remain stable in that time span.

Notice pro meters are very long in shape, that's to prevent overheating, because measurements are temperature sensitive, the DISPLAYS themselves are also temperature sensitive. Consumer $200 meters do not have the repeatability of the profession $10,000 meters, they are not designed to be able to work with long measurement times.

And you've also recommended people create p3 luts, this is hugely problematic, because currently the p3 lut causes red-clipping issues on certain chemistries of Wide gamut panels, such as PFS panels. In all of these cases p3 lut should be avoided. instead create a rec2020 lut which will avoid the clipping.

[huhn]

go rewrite dispalycal is open source do now.
your p3 stuff is just hilarious the display never know about this so why would it do anything different compared to bt 709 or bt 2020 which would be created from the same readings...

there is a good reason the display and the meter in contact mode should be heated up before calibration.

i have run 10 hour calibration runs just for the fun of it and guess what the verification was stop on how could that be?

[el Filou]

Movies without 3Dlut correction arn't worth watching.IMHO this kind of completely over the top remark doesn't help at all getting people interested in specialty software, picture processing, calibrating etc. This is akin to saying music isn't worth listening to unless you're using perfectly linear studio monitors and sitting in a fully treated studio-like room, which is a ridiculous idea.They're even implementing dynamic tonemapping relative to Room lighting. That's kind of like taking HDR and going back to SDR. Because now they're walking away from the absolute tone curve.
BT1886 looks like shit, and that's kind of what they're bringing back to HDR..Breaking news, most people don't watch movies in a controlled environment like that of a mastering studio or on a perfect zero-luminance-black display, and manufacturers will want to provide a better experience for those people, whether purists like it or not. Dynamic tuning of the curve depending on the type of scene and the viewing environment is much better than the static curve alterations that Samsung had been doing which were bad practice and weakened the accuracy of HDR and made other - accurate - sets look worse in shops.
This is not at all akin to going back to SDR, and the only case where Bt.1886 doesn't look better is if you're using a CRT or an OLED in a pitch black room, i.e. not at all the common situation people view movies in.Of course then there's the conspiracy section, they don't promote calibration so people would buy new tvs every 2-3 years.There are many reasons why some people frequently buy a new TV, but owners being too lazy to do calibration isn't one of them.

[tp4tissue]

go rewrite dispalycal is open source do now.
your p3 stuff is just hilarious the display never know about this so why would it do anything different compared to bt 709 or bt 2020 which would be created from the same readings...

there is a good reason the display and the meter in contact mode should be heated up before calibration.

i have run 10 hour calibration runs just for the fun of it and guess what the verification was stop on how could that be?

Because you fundamentally misunderstand how the measurement and corrections work.

These are NOT precision instruments. the i1dp, as good as it is, it's the Harbor Freight caliper of the colorimeter world.

The clipping issue isn't a matter of gamut difference/coverage, it's a matter of software. Depending on the angle of the primaries, the software algorithm in this case argyll, or madvr, in conjunction is causing the red clipping problem. This problem varies depending on the chemistry of the panel.

The readouts from the meter are also Relative metrics, they're not absolutes.

[huhn]

ok i'm misunderstanding verification.

and it's not common practise to use an expensive spectrometer to calibrate a cheap colorimeter to do the job with it?

and just because a i1d3 has deltas in the 2 and variance between models that doesn't mean they read the same color differently because they are very stable that's why a correction is so tiny 4 patches because they for our use they are precision instruments.

[mclingo]

Hiya, anyone know if there is a workaround for smooth motion crash with 113 beta, sure this was discussed at some point point searching doesnt seem to reveal anything.

[tp4tissue]

music isn't worth listening to unless you're using perfectly linear studio monitors and sitting in a fully treated studio-like room

I agree with this assertion.

most people don't watch movies in a controlled environment like that of a mastering studio or on a perfect zero-luminance-black display

You misunderstand what I'm talking about, I'm talking about the workflow being BLIND. It's impossible to produce a perfect display for everyone, but at the moment, it's a Blind seeing eye dog, leading the blind guy, and he gets hit by a bus. That's what happens when you don't have calibration in the workflow. All i'm advocating for is a non-blind seeing eye dog, he can have 3 legs for all I care, but you've got to have this dog @ at least this level of NON-blindness. hahahaha

This is not at all akin to going back to SDR, and the only case where Bt.1886 doesn't look better is if you're using a CRT or an OLED in a pitch black room,



BT1886 being shit is the assertion of color professionals. Even pro color house light illusion recommends against it.

Movies SHOULD be watched in a black room. They were created in a black room, they should be watched in one. Let the darkness consume you.

[huhn]

if bt 1886 is "whatever" then gamma 2.4 is the same?

[tp4tissue]

if bt 1886 is "whatever" then gamma 2.4 is the same?

bt 1886 = gamma 2.4 IFF the display black level is 0, like on oled.

The point of BT 1886 is to pick up the tone curve on non 0-black displays, such as LCDs.

The problem is, it looks bad, and almost always heavily distorts the creator's intent.


Messing with the image using a Day/ Light sensitive process is AGAIN trying to do what bt 1886 does.

Maybe it could be done better than bt1886, but I doubt it. 10 years from now, they'll be saying, yea, this was bogus image tuning to upsell our rebadged TVs every year. Again, Ambi-sensors are not new, it's been tried, and they never work right.

[chros]

The point of BT 1886 is to pick up the tone curve on non 0-black displays, such as LCDs.

The problem is, it looks bad, and almost always heavily distorts the creator's intent.
I agree with this, I never liked BT.1886.

[mclingo]

I can never get my picture looking right using 2.4 or BT1886, both seem to accentuate macro blocking and macro grain on poorly masted material, 2.2 has always looked better on my OLED but its not calibrated so this could mean nothing at all :)

[huhn]

bt 1886 on your oled should be the same as 2.4 absolute so what are you doing LG?

[tp4tissue]

I can never get my picture looking right using 2.4 or BT1886, both seem to accentuate macro blocking and macro grain on poorly masted material, 2.2 has always looked better on my OLED but its not calibrated so this could mean nothing at all :)

macroblocking near black, or all over ?

Near black issue is probably due to black roll-off problem on the oled, calibration might not help because the grey balance adjustment doesn't go down that low, but 3dlut might help. Either case you should do both.

if your display is not calibrated, how are you setting bt 1886 or 2.4

[mclingo]

its very similar but slightly worse looking that 2.4, however my TV being gen 1.0 didnt have great near black handeling, this issue was likely sorted in later models, compressed stuff can look horrific on my TV sometimes.

[mclingo]

macroblocking near black, or all over ?

Near black issue is probably due to black roll-off problem on the oled, calibration might not help because the grey balance adjustment doesn't go down that low, but 3dlut might help. Either case you should do both.

just near black but see above, its mainly down to my particular model.

[tp4tissue]

just near black but see above, its mainly down to my particular model.

You might be able to tug on the near black handling by pulling the lowest grey balance value really high or really low,

Then use a 3Dlut (on madvr's side) to correct the difference. You don't care about accuracy here, because you're just trying to either crush the macroblocks out, OR lift the entire black level just enough that it looks even.

[tp4tissue]

and it's not common practise to use an expensive spectrometer to calibrate a cheap colorimeter to do the job with it?

and just because a i1d3 has deltas in the 2 and variance between models that doesn't mean they read the same color differently because they are very stable that's why a correction is so tiny 4 patches because they for our use they are precision instruments.

Spectros are simply not affordable, a good one cost several times more than an Oled TV.

for calibration, we can only rely on generic (but professional) spectro corrections to use with our < affordable > colorimeters.

Those corrections are definitely good-enough for home use and they're done w/ pro equipment and lab settings..

The spectral for early quantum models are some what iffy, because those coatings are less stable over time than the older wide gamut phospors.

The newer quantum corrections should be fine and stable.

i1d3 is not a precision instrument, it will never be a precision instrument. It's good enough for for movie watching and photography work, it ends there. Just like how harbor freight caliper is good enough for wood working, but you need to go mitutoyo or starrett for metal.

[mclingo]

yeah, i have a profile specifically for this material, this sounds like it will look horrible but it works really well. I push gamma to 1.9, this lifts the whole picture, I then lift 5 ire by about 15 points to even out the macro blocks, I then drop the overall brighness down using contrast, OLED brightness on this TV isnt very granular, this was also fixed with later models.

This is far from ideal and will likely horrify everyone who's reading this but I only use this for really really bad compressed streamed stuff, 95% of what I watch is bluray which I have zero issues with generally.

I have tried lots of other ways to do this using 2.4 gamma and lifting 20 point low etc but I have the best results with his hands down, the only issue I do have with it is that it makes the black bars a bit grey but as a lot of streamed content is 16:9 this is also rarely an issue as the bars or very small or none existent.

[tp4tissue]

Have you tried adjusting in HCFR, in the dark room, as soon as you find the value range where it starts to flicker the black into grey, that's your target, either crush of lift beyond that point.

You're saying you need to push the gamma ONTOP of pulling the 5% ? You're using i1dp ?

[mclingo]

I'm doing this by eye mate which I know is ridiculous, I never got around to buying one, i had other financial priorities, i had to buy a new PC. I'm hoping to pick one up next year.

[VBB]

Hiya, anyone know if there is a workaround for smooth motion crash with 113 beta, sure this was discussed at some point point searching doesnt seem to reveal anything.

There is not. We'll have to wait for 114.

[SamuriHL]

Of course there's a work around. Turn it off. HAHAHAHA I'm so funny. No, most people that use it went back to 112.

[mclingo]

is there much to gain PQ wise in 113?

[SamuriHL]

I think this is the build where he added the curve selection dropdown so while the PQ hasn't changed much the build is easier to maintain. I don't remember what other changes he made. I don't use smooth motion so I'm still on 113.

[mclingo]

i'm trying smooth motion over the next few days for reasons I wont bore you with, so far its looking pretty good, no frame drops or glitches, what are the disadvantages using this, is there any loss in PQ at all?

[huhn]

after some time you will start seeing the ghosting it creates.

[VBB]

On an OLED, the little ghosting SM causes is actually welcome. I don't consider it a loss in picture quality, but rather a gain overall.

[mclingo]

ok cheers.