madVR - high quality video renderer (GPU assisted)

[lazostat]

I have a question about Madvr. I have seen some guides that are many pages and i need many hours to read them and make the settings. But i don't know if it's worth the time.

I see movies on my 1080p TV (46") connected to PC, also i have a 2k pc monitor (27") but it's not as good and i rarely watch movies on this.

Can Madvr improve Pc monitor image quality? Also if i watch the same source as my TV, or Pc monitor, and i have a very good mkv quality, will i see any difference?

I have read that Madvr is good for projectors or very large TVs or for scaling methods. So, should i bother with all those advanced guides or just stick to the basics?

[SamuriHL]

It's not a question anyone can answer for you. There's WAY too many variables. It's not just your equipment that influences the answer, it's the quality of the content, your viewing area, if your equipment is calibrated, etc, etc. The likelihood is that madvr will improve image quality in most cases. How much and whether it's noticeable by you is not something that can be answered.

[.m4jX]

You don't have to spend much time if you don't want to have the best quality for various scenarios. Setting a couple of things will already get you to a better place than most regular players.
e.g. "chroma upscaling" -> "NGU" "Anti-Alias, medium"
image upscaling" -> "doubling" to "NGU Standard" or "NGU Sharp", "luma doubling" to "medium", "luma quadrupling" to "medium", "chroma" to "normal"

Play a SD file, a 720p file and a 1080p file in fullscreen, hit Ctrl+J and see if "rendering" stays below 40ms. If it does, you're done. If not, set something to "low" instead of "medium". If you have a good GPU and are far below 40ms you can try "high" instead of medium, if you don't mind using more power. Can be done in a few minutes. Even just setting everything to "NGU" and the rest to "let madVR decide" will get you good results.

[flossy_cake]

Regarding the upscaling refinement setting, can anyone advise what exactly the condition MadVR is testing for when it decides whether to enable its upscaling refinement settings?

I've made the following observations :


Video source res | Fullscreen display res | Is MadVR upscaling? | Did MadVR activate upscaling refinement?
720x480 | 1366x768 | yes | definitely yes
1280x720 | 1366x768 | yes | seemingly not
1280x720 | 1920x1080 | yes | seemingly not


I say "seemingly not" because I don't know if it's active but being suppressed by that pre/post resize behaviour mentioned on the previous page.

I'm guessing there is some threshold like eg. video must be upscaled in both horizontal and vertical by a certain percentage before it qualifies for upscaling refinement -- is this correct?

I need to know because otherwise I could end up with some videos being sharpened when I don't want them to. If that is the case then I need to use tags like [profile=x] to control it all manually instead of trusting MadVR to detect them as I'd like.

[aufkrawall]

I have read that Madvr is good for projectors or very large TVs or for scaling methods. So, should i bother with all those advanced guides or just stick to the basics?
Get a colorimeter to create a proper ICC profile with it. Then you can generate 3D LUTs with DisplayCal for madVR. I think most will agree that after having done this once, nobody wants back to hand-tunes catastrophes.

It can even create a 3D LUT for HDR -> SDR conversion which I find to look even better than madVR's internal BT.2390 conversion. It really looks phenomenal.

[Siso]

It can even create a 3D LUT for HDR -> SDR conversion which I find to look even better than madVR's internal BT.2390 conversion. It really looks phenomenal.

How can it be done?

[huhn]

displaycal can create these 3DLUT they are static tonemapping LUTs so don't except anything like dynamic tone mapping from them.

you take bt 2020 PQ are source apply a tone curve of smpte 2084 to it and use the trade of setting and type in the true peak nits just like in madVR. define a target either measured or something like DCI P3 D65 bt709
after that you can load it in madVR under HDR.

[Siso]

displaycal can create these 3DLUT they are static tonemapping LUTs so don't except anything like dynamic tone mapping from them.

you take bt 2020 PQ are source apply a tone curve of smpte 2086 to it and use the trade of setting and type in the true peak nits just like in madVR. define a target either measured or something like DCI P3 D65 bt709
after that you can load it in madVr under HDR.

Can I create one from my sdr 3dlut settings? Or I have to run the whole calibration again?

[huhn]

yes you could do that.

[aufkrawall]

displaycal can create these 3DLUT they are static tonemapping LUTs so don't except anything like dynamic tone mapping from them.

I find the result very sane, madVR's internal BT.2390 conversion looks overexposed to me. Less than mpv's, but still (though I haven't tried any of the test builds after the last stable madVR release).

[Filmgeek47]

Thinking of upgrading my card to the 3060ti. Is there a stable driver release that will work on that card without stutter/issues for MadVR?

[clsid]

The stutter issue can be fixed by adjusting settings. They have only been mentioned 318 times in this topic so far so you might have missed it...

[Filmgeek47]

The stutter issue can be fixed by adjusting settings. They have only been mentioned 318 times in this topic so far so you might have missed it...

Thanks, that’s all I needed to know. Didn’t have the patience to comb through 3k pages, lol.

[OrangeFlavored]

I've been having issues ever since switching to W11 some months ago that when playing a video with madVR on my main monitor (native resolution 2560x1440) it will always change my screen resolution to 1080p while the file is open, even when "switch to matching display mode" is disabled and nothing is entered in the box of supported display modes. Is this a known issue with some kind of workaround available or is there some kind of quirk on my setup specifically causing this? Tried clean reinstalling a few times over the months but never had any luck fixing it besides simply not using madVR if I want to watch something on that monitor.

edit: Also I guess might as well mention that the "Custom Modes" tab has also been completely gone for me ever since upgrading to W11. This isn't really a major problem for me and I'm not that desperate to find a fix, but maybe it could help someone diagnose the issue, I dunno. (Googling this a while back indicated it might be caused by madVR failing to identify my display properly or something, which could also be why it creates a ton of extra "identification #" entries above but I'm not sure what I can do to resolve that.)

[ashlar42]

Obviously everybody has different priorities but my advice to anyone wanting to keep using madVR is not to upgrade to Windows 11 for the time being. Not that there's anything wrong with the new OS. But we know development of madVR PC is on an indefinite hiatus, if things need to be fixed, madshi won't be fixing them in a hurry. Perfect recipe for disaster, in my book.

Then everything can go well, I don't know. But I will be holding on to W10, for HTPC use, for as long as security updates will be available.

[KrazyLurt]

Regarding Windows 11, as we know, using alt+tab or swich to virtual desktops, triggers MadVR OSD to report only 8-bit color.

Have anyone done any testing so that is actulally happening and not a false positive?

[huhn]

that should be accurate because madVR decides what it sends.

[tp4tissue]

Can I create one from my sdr 3dlut settings? Or I have to run the whole calibration again?


Depending on how well your tv's primary rgb targets track, you MAY want to use the Computer's Gamma-Ramp (1DLut) compensation instead.

The reason is, madvr's HDR tonemapping can cause saturation anomalies when used with 3DLut.

So here's the example workflow for HDR -> SDR tonemapping involving color.

Set TV to max brightness, run rgb flashing colors for 1.5 hours. <LCD>, Oled, ~45min.

Make sure you're in the right mode, 10bit, 23hz, etc. they are unique and have different tracking behavior.

Then go into HCFR, set your RGB balance curve to extract maximum contrast, smoothest gamma tracking.

Then, go into Dispcal, select the SDR setting. Change Tone curve under the calibration tab to the gamma 2.2 default, It MUST be this setting, because madvr's tonemapper assumes purepower 2.2.

Go to profiling tab, gear icon next to profile type, uncheck low quality pcs, click enhanced, 33x33x33 smoothing. change to 596 patches,

Go to 3dlut tab, uncheck create 3dlut.

Click calibrate and profile on the bottom.

After it finishes, manually go into the folder and install the the MAIN, generated .icm, if you created a 3dlut, there will be a 3dlut icm, DO NOT install that one.

The Main .icm, is the one you want. This is the 1D compensation, and it works through the GPU.

Now in madvr, just set the target brightness for HDR>SDR to the peak brightness from that calibration file. Under calibration select this display is already calibrated to either 709 or p3, purepower 2.2. Make sure disable gamma ramp is UNCHECKED. you WANT the gamma ramp.

You will now have 2 profile associations with that single monitor, and they show up in the dispcal loader, so you can switch between them from your daytoday lower brightness mode and the HDR mode at peak brightness.

OR if you always use peak brightness, you could just have the 1, but this is not ideal as it cuts you off from better SDR performance when you're using regular desktop.

[tp4tissue]

It's not a question anyone can answer for you. There's WAY too many variables. It's not just your equipment that influences the answer, it's the quality of the content, your viewing area, if your equipment is calibrated, etc, etc. The likelihood is that madvr will improve image quality in most cases. How much and whether it's noticeable by you is not something that can be answered.

@lazostat
To add to SamHL's answer, Image quality on the Majority of monitors is Unwatchable due to improper/impure gamma tracking.

RGB Gamma tracking/Whitebalance are the most important base for image quality. You can have bonkers saturation, as long as gamma tracking is good, it'll look passable.

If you're after image quality, I recommend you purchase an i1display colorimeter or i1 studio, the cheaper one. Worth every penny.

If you're talking about Upscaling settings, etc, they do make a difference, but inreality, Not much.

COLOR ACCURACY makes a much bigger difference.

[aufkrawall]

You can have bonkers saturation, as long as gamma tracking is good, it'll look passable.

I totally agree that gamma (curve) should not be underestimated (too bad not everything is produced for 2.4, e.g. Harry Potter movies seem to be intended for 2.2...). However, I think saturation is similarly important. Infamous case of red faces with rec709 content on wide gamut displays... ;)

[ashlar42]

too bad not everything is produced for 2.4, e.g. Harry Potter movies seem to be intended for 2.2...

I am surprised that there is no database online for this. We spend hours, days, trying to extract every ounce of performance from our screens. And then we watch stuff with the wrong gamma settings :-(

[Siso]

I am surprised that there is no database online for this. We spend hours, days, trying to extract every ounce of performance from our screens. And then we watch stuff with the wrong gamma settings :-(

Unfortunally this is how it goes with the studios mastering...

[SamuriHL]

You think SDR is bad.... LOL HDR is an absolute mess.

[tp4tissue]

I am surprised that there is no database online for this. We spend hours, days, trying to extract every ounce of performance from our screens. And then we watch stuff with the wrong gamma settings :-(

Unfortunally this is how it goes with the studios mastering...

I'm reasonably sure all modern SDR mastering is done with 2.4 pure power standard. Usually a mastering oled, maybe plasma/crt if it's very early bluray.

HDR is 1 of 4-5 options, 1000nit/2000nit/4000nit but it's not like anything we buy approaches those mastering monitors so.... kind of moot.

[Manni]

I'm reasonably sure all modern SDR mastering is done with 2.4 pure power standard. Usually a mastering oled, maybe plasma/crt if it's very early bluray.

HDR is 1 of 4-5 options, 1000nit/2000nit/4000nit but it's not like anything we buy approaches those mastering monitors so.... kind of moot.

No, most modern SDR mastering is done using BT1886, not power 2.4.

With HDR, it’s worse than that as it’s not only the mastering display luminance that changes, the black floor is often raised as well.

[huhn]

bt 1886 is 2.4.
bt 1886 only changes to a lower gamma on displays that aren't OLED so you can not really change the source depending on that the file it self is not bt 1886 because it can't be. calibrating to bt 1886 is a different story and currently seen as the spec.

and the talk about 2.2 yes it used all the time (or was) to master a movie under studio lights and if that has been done you are "supposed" to use 2.4 in a dark room that's where 2.4 even comes from bt709 is 2.22.

[mclingo]

i've always found that BT1886 is closer to 2.2 on my TV but that's probably just my TV.

[Manni]

bt 1886 is 2.4.
bt 1886 only changes to a lower gamma on displays that aren't OLED so you can not really change the source depending on that the file it self is not bt 1886 because it can't be. calibrating to bt 1886 is a different story and currently seen as the spec.

and the talk about 2.2 yes it used all the time (or was) to master a movie under studio lights and if that has been done you are "supposed" to use 2.4 in a dark room that's where 2.4 even comes from bt709 is 2.22.

No, BT1886 isn't 2.4 power gamma.

BT1886 takes into account the black floor to compensate. Power 2.4 is flat, BT1886 is only equal to power gamma 2.4 if black is black (0nits). Otherwise is gradually goes from 2.2 up to 2.4, and the targets are calculated taking the black floor into account.

[aufkrawall]

Does anyone actually use BT.1886 gamma with an LCD device without local dimming for mere media consumer use case? Sure it prevents black crush ~entirely, but apparently at the cost of distorting of dark shades and killing average scene contrast. I can't imagine that any film maker would describe this as intended viewing conditions.
Also the help text below the 3D LUT section in DisplayCal describes both rel. gamma 2.2 - 2.4 and BT.1886 as correct for rec709 content. So I suspect BT.1886 is correct rather from a theoretical point of view, but in practice not really when it comes to consumer use cases (apart from OLED).

[SamuriHL]

Personally I calibrate two SDR modes on my OLED. Working on it right now in fact. Expert bright for regular TV watching at gamma 2.2 100 nits. (Yes, I know, that's not really "bright"....I've been informed already but I like it dark in here relatively speaking even during the day). And Expert dark for movie viewing at gamma 2.4 100 nits. On OLED bt.1886 and gamma 2.4 SHOULD, theoretically be identical more or less. So why not just calibrate for 2.4 and call it a day in that case? I'm still not clear on how a studio could master to bt.1886 anyway given that it's theoretically "dynamic" on where the black floor is. How does that work for content mastering where you should want it to be quite clear what it's mastered to??

[tp4tissue]

bt1886 is a compromise made for LCDs. 2.4pure power is the studio standard.

Mastering sdr at 2.2 depends on the delivery target, if it's pc games internet, usually 2.2

Major studio movies were usually by the book 2.4, but because different targets increased over time, they've put out special grades for each.

[aufkrawall]

My impression is PC games are mostly 2.4, though of course there are examples pointing the other way or having terrible clipping also with 2.2 due to bad tone mapping/lack of indirect light etc.
Also seems distinction between relative and absolute gamma once again is rather theoretical? At least I don't see any difference with 3D LUTs created accordingly.

[SamuriHL]

There really isn't SUPPOSED to be differences. LOL :)

[tp4tissue]

PC games are typically NOT graded to ANY-Standard.

If it looks good to you, you're done.


Movies on the other hand are stricter.

[tp4tissue]

There really isn't SUPPOSED to be differences. LOL :)

They wouldn't need colorimetry if there was no difference.

It's ok to say there's no "Correct" Gamma. So, 2.2/2.4 are equally correct.

However, there is a "preferred" pipeline. 2.4pure

Bt1886 looks like trash in the majority of cases, so just ignore it.

[huhn]

No, BT1886 isn't 2.4 power gamma.

BT1886 takes into account the black floor to compensate. Power 2.4 is flat, BT1886 is only equal to power gamma 2.4 if black is black (0nits). Otherwise is gradually goes from 2.2 up to 2.4, and the targets are calculated taking the black floor into account.

and how do you save that on a file?

if you master for bt 1886 you master for 2.4 because you have a 0 black on a high end mastering display. they may aim for bt 1886 but they don't master at it well they do by using 2.4 kind of.

it can be lower than 2.2 not that it matters here. easily looking really really bad.

what so ever the new gamma spec is bt 1886 and you are supposed to use it with new masters.

[djsolidsnake86]

there will be no more update of madvr? there is a better render actually?

[SamuriHL]

When you say no more update, it depends on what you're looking for. Right now madvr is being updated periodically with improvements to HDR tone mapping. Other than that, no, there's no updates.

As for "better" renderer, that's highly subjective.

[djsolidsnake86]

When you say no more update, it depends on what you're looking for. Right now madvr is being updated periodically with improvements to HDR tone mapping. Other than that, no, there's no updates.

As for "better" renderer, that's highly subjective.

last update is from 2019

[Asmodian]

last update is from 2019

https://www.avsforum.com/threads/the-future-of-madvr-htpc.3190488/page-6#post-60645444

[chros]

There really isn't SUPPOSED to be differences. LOL :)
Naaah.. . :)
Also seems distinction between relative and absolute gamma once again is rather theoretical? At least I don't see any difference with 3D LUTs created accordingly.
What's your black level? See point B.5.a. here (https://www.avsforum.com/threads/madvr-argyllcms.1471169/page-208#post-40108362).
You can also fake black level in DisplayCal on calibration tab to see this in a greater effect.

[SirMaster]

last update is from 2019

Last update is from Oct 1, 2021

https://www.avsforum.com/threads/improving-madvr-hdr-to-sdr-mapping-for-projector.2954506/page-681#post-61067550

[Octo-puss]

Why is he not posting updates here then?
I thought madVR was long ago abandoned.

[SamuriHL]

He decided to work on HDR tone mapping quite a few years ago (at this point) and was involved in the AVS forum thread. Then the Envy happened which has taken the majority of his time and focus. He's continuing to update the AVS forum with new HDR tone mapping beta builds in an effort to improve quality for the Envy and to give the madvr community access to newer builds. But for the most part, they are ONLY focused on HDR tone mapping and do not include anything else. Whether or not that changes after HDR tone mapping is "complete" is anyone's guess.

[flossy_cake]

Regarding the 2.4 gamma debate, the reality is that gamma was poorly standardised as the Bt.1886 standard didn't even exist until 2011 and lots of content was made before that.

Before 2011 the convention seemed to be just a 2.4 power law -- although it was never officially standardised -- but different room lighting conditions can call for a different gamma to get the same looking image (eg. 2.2 for daylit room) due to a visual effect called simultaneous contrast.

Sources:
https://www.itu.int/dms_pubrec/itu-r/rec/bt/R-REC-BT.1886-0-201103-I!!PDF-E.pdf
https://drive.google.com/file/d/1Ot_wBmMGGh-s6vJUk7H2aXfY_uflNQQf/view
https://app.spectracal.com/Documents/3_Gamma_2.0_2.2or2.4.pdf
http://poynton.ca/PDFs/GammaFAQ.pdf

Here I've plotted Bt.1886 with 120nits white and 0.05nits black which produces a curve similar to sRGB. https://drive.google.com/open?id=1MIwUSfc8HxmWx9pBqidaY7THfhv5rrs4S89Vgy1t2Y4

https://i.lensdump.com/i/g5ADyr.png

So if a movie was mastered on the above bt.1886 curve then it will look wrong on a 2.4 power law monitor, and vice versa.

For a good balance I target 2.4 power law but with black raised to make 1% or 2% grey visible. Then for daytime viewing I'll shift the whole curve down to 2.2. I wrote an app called Calibration Tools (https://bitbucket.org/CalibrationTools/calibration-tools/src/master/) which can do this automatically and gradually during the twilight period according to your GPS coordinates. But it's still not a great solution as ambient lighting isn't consistent due to unpredictable cloud cover causing different room lighting on different days. A proper solution needs an ambient light sensor, however updating the video card gamma table regularly with new values can cause dropped frames which would happen every time ambient light changed.

[huhn]

bt 1886 does never look the same if the black is not the same that's by design.
luckily it is 2.4 with 0 and there are mastering devices and home devices with 0.

[flossy_cake]

bt 1886 does never look the same if the black is not the same that's by design.
luckily it is 2.4 with 0 and there are mastering devices and home devices with 0.

In the google sheet1 cell M3 and N3 you can put in custom values for white and black and see the resulting curve.

[aufkrawall]

Thanks for Calibration Tools, flossy_cake. This tool should really be more widely known, the dither control with Nvidia or the override setting (helpful with some D3D9 games) add up nicely! Too bad Windows is so stupid and still (more or less) randomly breaks color precision for GPU gamma ramps. Microsoft's incompetence is just breathtaking...

I think your recommendation makes sense, 2.4 absolute gamma with 99% black output offset seems to be a good compromise. With 100%, gamma test picture in some games (Doom Eternal or Witcher 3) can be a bit hard to identify.

chros, thanks for linking your post. I'll definitely give it a read.

[Siso]

I think your recommendation makes sense, 2.4 absolute gamma with 99% black output offset seems to be a good compromise.


This is interesting, has anyone tested it for movies?

[flossy_cake]

2.4 absolute gamma with 99% black output offset seems to be a good compromise.
This is interesting, has anyone tested it for movies?

Just to clarify what I meant: I set my display's black level control until 1% or 2% grey becomes barely visible. 0% will still be darker than that.

I'm not sure if this is what aufkrawall meant by "99% black output offset".