madVR - high quality video renderer (GPU assisted)

[6233638]

http://i886.photobucket.com/albums/ac65/XJ6_avs/Video%20Cal%20%20Tieng%20Viet/1-BlackClipping.jpg

So I noticed that at the 0-255 setting in madVR, viewing this black clipping pattern I could not with any change of my TV settings get the 0-16 lines to be visible. When i change to limited range 16-235 though all the lines appear, like in the image above. Is that normal? I'd have thought it should be the other way round.Levels below 16 should not be visible, this pattern exists purely to set the brightness control on your display. If you are using anything other than a CRT, there should be no need for it whatsoever. Brightness jumps in very obvious steps on modern displays, and it should be set at the point where a black field is black, and one notch higher raises the black level. There's no need for specialised patterns.

MadVR uses the correct designation for the 0–255 and 16–235 settings, most other applications have it backwards.

On top all this I get quite significant aliasing with madvr on the reds even if I set everything to soft cubic 100, compared to EVR Custom with Potplayer's YV12 upsampling shader.Please post a sample.

I have a couple of questions about madVR.


Is there any benefit to run D3D11 mode?
Is it okay to run in windowed mode instead of exclusive? My video card (Radeon 7950) is more than capable of running windowed D3D11 and the transition from windowed to exclusive is annoying.
Any recommendations to change the default upscale settings?
Should I run Reclock with madVR?


Any input appreciated. :) Only if you are experiencing problems. Do not use it otherwise.
Windowed mode has lower performance and has the potential to suffer from screen tearing. If neither of these are an issue for you, there's no need for exclusive.
I've done a lot of testing on this. For chroma upsampling you should be using Mitchell-Netravali. The only other alternative is Bicubic 75. Any other scaling algorithm is detrimental to image quality when used for chroma upsampling, either because it softens the chroma too much causing colour bleeding, or sharpens it too much and adding ringing.

All upsampling is a balance between sharpness, ringing and aliasing, and the best algorithms in madVR only do two of the three well, so you have to decide what trade-off is most acceptable to you for Luma upsampling.
SoftCubic 80 is my personal preference, as it avoids aliasing and ringing in most cases, but does so at the cost of low sharpness.
Mitchell-Netravali is the best balance between sharpness and ringing, but does so at the cost of showing aliasing with some sources. This is probably what most people would be happiest with.
Bicubic 75 is the best balance between sharpness and aliasing, but does so at the cost of introducing ringing into the image.

If you want sharper than that, Spline 3 is the only other choice, but there's far too much ringing for it to be a realistic option in my opinion.

See this post (http://forum.doom9.org/showpost.php?p=1560193&postcount=12242) for some example images.
Absolutely. madVR does its best to avoid frame drops, but your refresh rate is never going to exactly match the video's framerate, and film should be played back at 24Hz, not 24/1.001 (23.976...) It also offers other features such as WASAPI audio output, audio upsampling, ReClocking 25p PAL content to 24p.

[mzso]

Please post a sample.


OK. Here you are. http://min.us/mbghHPtmA4/1f

I wrote down the settings and others in the description of the images

[fairchild]

@6233638 using Bicubic 60 instead of Bicubic 75 based on just the values in MadVR you lose a teensy weensy amount of sharpness but you gain less ringing and a tad more aliasing (still less than using Mitchell for luma). Any observations that you've noted that would make my choice of using Bicubic 60 for luma instead of Bicubic 75 a bad one?

[Keiyakusha]

Absolutely. madVR does its best to avoid frame drops, but your refresh rate is never going to exactly match the video's framerate, and film should be played back at 24Hz, not 24/1.001 (23.976...) It also offers other features such as WASAPI audio output, audio upsampling, ReClocking 25p PAL content to 24p.[/list]

But as I understand and from my experience this is true if we watch lets say 24p content on 24, 48 etc. hz screen. If we watch it on 60 hz screen it does nothing good.
And who says film should be played at 24hz? Film should e played at the correct rate, not in some selected. If it was made into 24/1.001 - that's how it should stay. By speeding it up to 24hz reclock will drop some audio samples to not to create desync. More harm than good.
In other words, I'd say if you don't see some sync problems, reclock is not needed. And I doubt many people would notice the actual difference.

[Xello]

But as I understand and from my experience this is true if we watch lets say 24p content on 24, 48 etc. hz screen. If we watch it on 60 hz screen it does nothing good.
And who says film should be played at 24hz? Film should e played at the correct rate, not in some selected. If it was made into 24/1.001 - that's how it should stay. By speeding it up to 24hz reclock will drop some audio samples to not to create desync. More harm than good.

24 is the standard for film (ie 35mm), it's slowed to 23.~ for NTSC and sped up to 25 for pal.

[Keiyakusha]

24 is the standard for film (ie 35mm), it's slowed to 23.~ for NTSC and sped up to 25 for pal.
I know that this is standard but audio that we have matches 23.976 rate, not 24. So who cares what are the actual film rate when it was on tape? we have 23.976 with matching audio. No need to touch anything. Not to mention that now everything gets created in 23.976, cause there is no tape anymore.
If standard was somehing like 27fps but on digital media you was getting 18fps (made up numbers), you would speed it up to 27 too and adjust audio? what for?

EDIT: pal is a bit different case, the only reason to slow it down back I see is because 25fps is not how it was intended to watch. I never dealing with pal stuff but before slowing it down i would want to know perfectly how was processed audio. was pitch adjusted during conversion or not? But since I probably won't have this info, I won't slow down anything cause no guarantee that it will be closer to how it was intended. Plus any conversion degrades quality.

[6233638]

@6233638 using Bicubic 60 instead of Bicubic 75 based on just the values in MadVR you lose a teensy weensy amount of sharpness but you gain less ringing and a tad more aliasing (still less than using Mitchell for luma). Any observations that you've noted that would make my choice of using Bicubic 60 for luma instead of Bicubic 75 a bad one?Because with Bicubic 60 you get noticeable ringing and aliasing, whereas with Bicubic 75 it's mostly just ringing. I selected those options because they are the best balance between having two positives and one negative, vs one strong positive, and two negatives. It's not necessarily a bad choice, I just think that those are your three best options depending on what your preference for ringing/aliasing/sharpness is.


OK. Here you are. http://min.us/mbghHPtmA4/1f

I wrote down the settings and others in the description of the images
Can you actually post a sample video rather than screenshots? There's no way for Madshi to try and reproduce the issue (if it is an issue) without it.

I've definitely not seen aliasing like that with madVR when it hasn't been a problem with the source file. From your screenshots, it looks like chroma is shifted over to the right with Potplayer's filter compared to madVR.

I know that this is standard but audio that we have matches 23.976 rate, not 24. So who cares what are the actual film rate when it was on tape? we have 23.976 with matching audio. No need to touch anything. Not to mention that now everything gets created in 23.976, cause there is no tape anymore.Many discs are actually at the original 24fps. In fact, I would say that this seems to be a more common occurrence now. A random pick from one of my 2012 discs: (We Need to Talk About Kevin)
http://img99.imageshack.us/img99/1958/24fps.png

pal is a bit different case, the only reason to slow it down back I see is because 25fps is not how it was intended to watch. I never dealing with pal stuff but before slowing it down i would want to know perfectly how was processed audio. was pitch adjusted during conversion or not? But since I probably won't have this info, I won't slow down anything cause no guarantee that it will be closer to how it was intended. Plus any conversion degrades quality.It is exactly the same situation. 24fps is slowed down to 24/1.001 for legacy NTSC compatibility with 3:2 pulldown at 60/1.001, just like PAL was sped up to 25fps for compatibility with 50Hz displays.

The difference is that very few Blu-rays are sped up to 25fps any more (especially because they need to be 1080i50 rather than 1080p25) because legacy compatibility is almost irrelevant with Blu-ray, and so they're 24fps on the disc.

And I can say with certainty that PAL films are not pitch-adjusted when sped up to 25fps, just like they aren't pitch-adjusted when slowed down to 24/1.001fps. If you want the correct speed for movement in films and the correct audio pitch, you need to be watching them at 24fps, not 24/1.001 or 25. It's very easy to tell if you are familiar with some of the music in the soundtrack, compare the DVD with the Blu-ray etc.


And even if the disc you're watching is 24/1.001, and your display is set to "23Hz" it is not going to be exactly 24/1.001, so you will need ReClock to avoid dropped frames over the course of a two hour film anyway. You might as well go all the way and watch it at the correct speed while you're at it. Having the ability to upsample audio and ReClock the video is one of the major advantages that an HTPC-based system has over a standalone player.

[Keiyakusha]

Many discs are actually at the original 24fps. In fact, I would say that this seems to be a more common occurrence now. A random pick from one of my 2012 discs: (We Need to Talk About Kevin)
Good, means we don't need to touch anything anyway. this has no relevance to the topic since discussion is about reclock.

It is exactly the same situation. 24fps is slowed down to 24/1.001 for legacy NTSC compatibility with 3:2 pulldown at 60/1.001, just like PAL was sped up to 25fps for compatibility with 50Hz displays.
1) you cant be sure was it slowed from 24 or was created at 23.976
2) in whatever case if you already have 23.976, your audio matches it. no reason to touch it. It is impossible to notice speed difference, but possible to notice when audio samples gets dropped or if we chose to resample audio instead - overall quality will degrade.

To sum up things. If you have some problems like dropped frames or jerky playback, it is good idea to match video fps with refresh rate. But if you have no problems like that or if you just casual user like many others and use 60hz display to watch everything you throw into it - you don't need reclock.

Question was if user should use reclock. You answered - absolutely yes. Which is wrong, this is not more than your opinion. He should use it only if it is needed.
Maybe I misunderstood your post but I got the idea that you say you always need to use reclock to speedup/slowdown. This is not true. You need to do it only if you really need.

P.S.
And I can say with certainty that PAL films are not pitch-adjusted when sped up to 25fps
Really? All titles from any company? I have nothing more to discuss with you.

[nevcairiel]

And I can say with certainty that PAL films are not pitch-adjusted when sped up to 25fps
There certainly is content where the PAL audio is pitch adjusted. Just a blanket statement won't hold up. :p

[6233638]

There certainly is content where the PAL audio is pitch adjusted. Just a blanket statement won't hold up. :pI won't exclude the possibility, but I have never encountered it, and it should be obvious if it's happening.

Good, means we don't need to touch anything anyway. this has no relevance to the topic since discussion is about reclock.It is definitely relevant if you are running your display at 24/1.001 and not 24Hz.

1) you cant be sure was it slowed from 24 or was created at 23.976Is it a commercially released film? 24fps.

Is there the possibility of a rare exception to this rule? Yes, as with all rules, but the number of films created at 24/1.001fps is significantly less than those created at 24fps and sped down.

2) in whatever case if you already have 23.976, your audio matches it. no reason to touch it. It is impossible to notice speed difference, but possible to notice when audio samples gets dropped or if we chose to resample audio instead - overall quality will degrade.Dropped frames are very obvious during playback, and ReClock is essential to eliminate them. By the time the audio gets to your ears, I guarantee you would not be able to tell the difference in a double-blind test.

To sum up things. If you have some problems like dropped frames or jerky playback, it is good idea to match video fps with refresh rate. But if you have no problems like that or if you just casual user like many others and use 60hz display to watch everything you throw into it - you don't need reclock.So for anyone that is in this topic looking for the best video quality possible, ReClock is required.

Question was if user should use reclock. You answered - absolutely yes. Which is wrong, this is not more than your opinion. He should use it only if it is needed.It is necessary if your display's refresh rate does not match the source framerate exactly, which is always the case.

[dansrfe]

I would just like to say that many of the music videos on YouTube that are 25fps are shot at 25fps and don't seem to be pitch adjusted or sped up. This is probably the exception to PAL content. Example: http://www.youtube.com/watch?v=8UVNT4wvIGY

[nevcairiel]

Any PAL content actually produced in a PAL country is usually real 25p.

[dansrfe]

Any PAL content actually produced in a PAL country is usually real 25p.

Yep, that's what i was thinking. I can tell mainly because i have to go into ReClock and uncheck "PAL 25fps -> 24fps slowdown" and manually set Media adaption speed to "Original" since otherwise it speeds it up to 25.714 or something like that. Kind of annoying actually.

[Keiyakusha]

Is it a commercially released film? 24fps.

Is there the possibility of a rare exception to this rule? Yes, as with all rules, but the number of films created at 24/1.001fps is significantly less than those created at 24fps and sped down.
You so wrong about that. You shouldn't assume that American movies is the only 24p content and America is the only country that uses NTSC. This is may be true that all American movies are 24fps, I have no idea I don't watch them.

Dropped frames are very obvious during playback, and ReClock is essential to eliminate them.
So for anyone that is in this topic looking for the best video quality possible, ReClock is required.
Why you stating the obvious? If you have dropped frames, reclock may help you, I said so too. But even when you have 60hz display it is possible that you'll be able to watch 23,976, 29.97, 30, 24 etc content with 0 drops. Don't assume that everyone have dropped frames during playback.

It is necessary if your display's refresh rate does not match the source framerate exactly, which is always the case.
If it doesn't match exactly why would we want to match it? Software+hardware should be able to match everything by making adjustments to when frame presented on the screen. Things like reclock needed when everything else fails.

[mzso]

Can you actually post a sample video rather than screenshots? There's no way for Madshi to try and reproduce the issue (if it is an issue) without it.

I've definitely not seen aliasing like that with madVR when it hasn't been a problem with the source file. From your screenshots, it looks like chroma is shifted over to the right with Potplayer's filter compared to madVR.

It would be pointless. Its not video specific. It happens with all.

Something's definitley not working right. Even the .bat to restore default preferences didn't work, nor did unregistering and removing madvr. (The preferences were remembered after reinstall somehow). Changing driver versions also didn't have any affect on the jaggedyness.

[6233638]

Any PAL content actually produced in a PAL country is usually real 25p.Television content, absolutely. Film is 24p regardless of its country of origin. It's designed for theatres, not televisions. (possible exception being made-for-TV films, but even then I would expect 24p as it's cheaper to shoot)

And an increasing number of modern television shows are shot at 24p today, not 25/30p. (buy the Blu-ray for just about any show and you will see it's 24p) The only real exception I've seen to this are some BBC-produced shows, but some of the more recent discs I've had from them have been 24p as well now.

Why you stating the obvious? If you have dropped frames, reclock may help you, I said so too. But even when you have 60hz display it is possible that you'll be able to watch 23,976, 29.97, 30, 24 etc content with 0 drops. Don't assume that everyone have dropped frames during playback.You have these options: Perfect sync between the audio, video and display clocks. This is virtually impossible with HTPC-based playback.
Syncing the video to the refresh rate without resampling the audio. Sound is "bit-perfect" but drifts out of sync over time. (lip-sync errors)
Dropping/duplicating video frames to keep video in sync with the audio. Results in non-smooth playback.
Syncing the framerate of the video to the refresh rate and dropping/duplicating audio frames to keep it in sync with the video. Results in noticeable audio problems.
Syncing the framerate of the video to the refresh rate, and resampling the audio to keep it in sync. This is the only option which provides both smooth video playback and keeps audio in sync without audible problems. For this, you need to use ReClock.
Option 5 is the only acceptable choice, and is required whether you're playing 24/1.001fps content at "23Hz" on your display or not, because the "23Hz" option is not exactly 24/1.001.

If it doesn't match exactly why would we want to match it? Software+hardware should be able to match everything by making adjustments to when frame presented on the screen. Things like reclock needed when everything else fails.Sorry, I don't think you have an understanding of how the Audio & Video clocks work on a PC. If it were possible to sync up the clocks in hardware, there would be no need for ReClock.

It would be pointless. Its not video specific. It happens with all.

Something's definitley not working right. Even the .bat to restore default preferences didn't work, nor did unregistering and removing madvr. (The preferences were remembered after reinstall somehow). Changing driver versions also didn't have any affect on the jaggedyness.
I think it's important because that is definitely not happening here.

Best examples I could find of something red which might show aliasing:
http://www.abload.de/thumb/foxl2ygd.jpg (http://www.abload.de/img/foxl2ygd.jpg) http://www.abload.de/thumb/texta6l9z.jpg (http://www.abload.de/img/texta6l9z.jpg)

[Keiyakusha]

Sorry, I don't think you have an understanding of how the Audio & Video clocks work on a PC. If it were possible to sync up the clocks in hardware, there would be no need for ReClock.
I say results may be different depending on what hardware user have. We have one renderer but with one hardware everything is fine, with another we have drops. Maybe bugged drivers or many other things. But when everithing works like it should, there is no drops.

You have these options:
...
So basically you say that sync can be achieved only by messing with number of frames or samples whatever video, audio or both before they presented on the screen? You definitely know a lot. Bye. ^_-

[leeperry]

Any PAL content actually produced in a PAL country is usually real 25p.
It might be hard to detect pitchshifted 24p@25, but slown down 25p@24 sounds like everyone is really tired and/or has been taking a lot of testosterone pills lately :D

[InsulinJunkie]

Many discs are actually at the original 24fps. In fact, I would say that this seems to be a more common occurrence now.

The vast majority of US discs I've seen that are left at 24 fps are mostly European-owned-and-transferred properties (Studio Canal properties spring to mind).

[vivan]

OK. Here you are. http://min.us/mbghHPtmA4/1f

I wrote down the settings and others in the description of the imagesCan you post screenshot with madVR OSD (Ctrl+J) with that video?

[King Kong]

I've done a lot of testing on this. For chroma upsampling you should be using Mitchell-Netravali. The only other alternative is Bicubic 75. Any other scaling algorithm is detrimental to image quality when used for chroma upsampling, either because it softens the chroma too much causing colour bleeding, or sharpens it too much and adding ringing.

All upsampling is a balance between sharpness, ringing and aliasing, and the best algorithms in madVR only do two of the three well, so you have to decide what trade-off is most acceptable to you for Luma upsampling.
SoftCubic 80 is my personal preference, as it avoids aliasing and ringing in most cases, but does so at the cost of low sharpness.
Mitchell-Netravali is the best balance between sharpness and ringing, but does so at the cost of showing aliasing with some sources. This is probably what most people would be happiest with.
Bicubic 75 is the best balance between sharpness and aliasing, but does so at the cost of introducing ringing into the image.

If you want sharper than that, Spline 3 is the only other choice, but there's far too much ringing for it to be a realistic option in my opinion.

See this post (http://forum.doom9.org/showpost.php?p=1560193&postcount=12242) for some example images.
Very helpful suggestions imo! :goodpost:
Right now I use Mitchell-Netravali for chroma upsampling, SoftCubic 80 for luma upsampling and Bicubic 50 for luma downsampling.
As I usually watch Full HD Content (99%) would you suggest to stick with it or should I change (at least) the luma upsampling routine?

Thank your for your guidance in advance! :thanks:

[nevcairiel]

If you watch mostly Full HD material on a Full HD screen, the Luma scaling is not of much use for you, so the setting isn't all that important.

[6233638]

So basically you say that sync can be achieved only by messing with number of frames or samples whatever video, audio or both before they presented on the screen? You definitely know a lot. Bye. ^_-If you know otherwise, please share it with the world. I'm sure a lot of people would be interested in knowing how to genlock the video and audio clocks on a PC.

Otherwise, the video and audio clocks on a PC are totally separate, and you need to make a compromise somewhere in order to achieve perfect sync over the entire length of a full 90–120 minute film.

The least perceptible option we have today, is upsampling the audio with ReClock.

[mzso]

I think it's important because that is definitely not happening here.

Best examples I could find of something red which might show aliasing:
http://www.abload.de/thumb/foxl2ygd.jpg (http://www.abload.de/img/foxl2ygd.jpg) http://www.abload.de/thumb/texta6l9z.jpg (http://www.abload.de/img/texta6l9z.jpg)

Yeah, well I'm open to ideas what could be the cause of the problem(s). I tried anything I could think of...

[mzso]

Can you post screenshot with madVR OSD (Ctrl+J) with that video?

Sure. ( :mad: Minus is screwed ATM awesome )
http://www.abload.de/thumb/madvr-with-info5jjsz.png (http://www.abload.de/image.php?img=madvr-with-info5jjsz.png)

[vivan]

h264, 8 bit, 4:2:0 -> YUY, 8 bit, 4:2:2means that decoder performs vertical chroma upsampling probably using NN (nearest neighbor) algorithm. Than EVR makes the same in horizontal direction, using the same algorithm, than shader fixes (smooths) the result.
MadVR does the same (chroma upsampling in horizontal direction), but using better (selected) algorithm, but it can't fix what decoder have done.

So, you just need to make decoder to output video without chroma upsampling (in YV12/NV12 format).

[mzso]

means that decoder performs vertical chroma upsampling probably using NN (nearest neighbor) algorithm. Than EVR makes the same in horizontal direction, using the same algorithm, than shader fixes (smooths) the result.
MadVR does the same (chroma upsampling in horizontal direction), but using better (selected) algorithm, but it can't fix what decoder have done.

So, you just need to make decoder to output video without chroma upsampling (in YV12/NV12 format).

I don't think there's such option in LAV. (And I didn't think that decoders should do upsampling)

[vivan]

I don't think there's such option in LAVNV12 and/or YV12 should be checked in "Output Formats", that's all. However LAV performs proper upsampling (better than NN).
If you are using LAV decoder, than check of there is anything between it and renderer (like ffdshow raw filter or vobSub).

(And I didn't think that decoders should do upsampling)It shouldn't. But it can, if it forced to.

[mzso]

NV12 and/or YV12 should be checked in "Output Formats", that's all. However LAV performs proper upsampling (better than NN).
If you are using LAV decoder, than check of there is anything between it and renderer (like ffdshow raw filter or vobSub).

It shouldn't. But it can, if it forced to.

Well, it was the internal video processing filter....
But I didn't set it to do anything. I mostly used it to change lightness, when the blacks were poor in the video, and occasionally for some processing effects, like mirroring.
The problem is fairly recent, so something might have gone wrong in potplayer... Don't know why does it even touch upsampling...

Also I wonder how EVR-CP knew that the file was 16-235 when it was badly encoded like nevcairiel said.

Examining further enabling the YUY2/RGB24/32 High Definition Conversion Also fixes the blockiness. But it still manipulates the image before madVR, which I guess is unwanted, right?

Also it makes the image look a lot better with EVR-CP without the YV12 Chroma Upsampling shader. Although not quite as good. Too bad you can't set the player to don't do any colorspace/upscaling manipulation while keeping the other features: contrast, lightness, saturation, mirroring and such

[Andy o]

Good, means we don't need to touch anything anyway. this has no relevance to the topic since discussion is about reclock.ReClock's main function is specifically to address when this mismatch is present. This discussion is exactly what Reclock is about.

1) you cant be sure was it slowed from 24 or was created at 23.976
I will disagree with 6233638 here. I don't think this matters, if you're using ReClock to adjust speed (i.e. not bitstreaming undecoded audio). I have been able to barely ABX a 1000 and 1001 Hz sine wave (0.1% difference), but I'm very skeptical that without an immediate switch A/B test one would be able to tell if what you're hearing has been sped up, let alone complex audio like movie soundtracks or even music. Worse, in the real world, you're not comparing. 24-25p is big enough though, it's about a half-tone difference.

2) in whatever case if you already have 23.976, your audio matches it. no reason to touch it. It is impossible to notice speed difference, but possible to notice when audio samples gets dropped or if we chose to resample audio instead - overall quality will degrade.

To sum up things. If you have some problems like dropped frames or jerky playback, it is good idea to match video fps with refresh rate. But if you have no problems like that or if you just casual user like many others and use 60hz display to watch everything you throw into it - you don't need reclock.I agree in the 60Hz case, unless you wanna slow down 25p to the original 24 for movies, but ReClock can be useful for people who can't get exact 23.976 Hz or 24.000 Hz in their cards, and that's most people.


Maybe I misunderstood your post but I got the idea that you say you always need to use reclock to speedup/slowdown. This is not true. You need to do it only if you really need.For me, the more useful use of ReClock is that I can just not worry about if the content is encoded in 24.000 or 23.976. I just play anything around 24 into 23.976 (or 24.977 like my card apparently does).

[Keiyakusha]

I agree in the 60Hz case, unless you wanna slow down 25p to the original 24 for movies, but ReClock can be useful for people who can't get exact 23.976 Hz or 24.000 Hz in their cards, and that's most people.
The whole discussion is because I don't understand one thing and so far didn't found an answer: why would you want to get exact value of any framerate/refresh rate if you don't have dropped frames, what if the work that renderer does alone is enough? Just because matching some numbers and thinking that playback is "perfect" makes piece in your mind?

[namaiki]

At a low refresh rate of ~24Hz, the video and audio can slowly go out of sync during playback depending on when the video renderer decides to drop or repeat frames to get the video back into sync.

(I think you would have to have something terrible like 23.5Hz to make it noticeable though I'm not sure... I run my screen at ~71.92973Hz and don't use ReClock)

[Keiyakusha]

At a low refresh rate of ~24Hz, the video and audio can slowly go out of sync during playback depending on when the video renderer decides to drop or repeat frames to get the video back into sync.

(I think you would have to have something terrible like 23.5Hz to make it noticeable though I'm not sure... I run my screen at ~71.92973Hz and don't use ReClock)

I see, this makes sense! But so far this is not contradicts to what I was saying at the beginning. Reclock needed to solve something. User shouldn't be forced to use it unless he figures out he needs it.

[kalston]

Reclock isn't necessary for everyone, end of story.

Take me (I always use madVR + LAV btw):

-I have a 71.923hz display, 23.976 or 24fps content is flawless even without Reclock. No dropped or repeated frames with madVR. My display also does 75hz or 60hz if needed.

-I use JRiver MC17 and its audio renderer running in ASIO for bit exact playback and never had any A/V sync issue (only time I did was because I messed around with audio buffers like an idiot). MPC-HC with Reclock/WASAPI works fine too using the reference clock thing (the difference between my actual clock and the reference clock is too small to matter it seems).

I'm not using JRiver Videoclock thing (a Reclock like feature) because again, I just don't need it for flawless playback. I rarely ever watch PAL content so I never bother speeding it up (but I can understand why you would do it since the difference is audible).

However I think speeding 23.976 to 24 is a terrible idea with blu-rays if you don't have dropped/repeated frames to start with, you're butchering the lossless audio for no good reason by processing it (no matter how good Reclock is at it I can't stand the idea, I have good ears and bad eyes so for me audio has to be as good as it can possibly be on my setup).

[Andy o]

I've never had audio drift out of sync because of framerate/refresh mismatch, but I can't speak for others.

It boils down to this. You do need perfect sync of clocks if you don't want repeated/dropped frames. Or, you can use ReClock and forget about it. The more your hardware is close to perfection, the less you'll need ReClock. There is a point where you might not care if you get one dropped/repeated frame, it may be every few minutes, or even an hour or more, it depends on what you find annoying. There is also the fact that if you have one frame error in a long enough time, then you might just not notice it. If you don't feel like you wanna use it, then just don't, especially if you're bitstreaming undecoded audio, but then you'll have to make sure that your refresh rate is close enough to the frame rate not to be annoying, and you can't fix 24-25p speedup.

What I was saying was that for me, ReClock just lets me run everything at 23.977 or whatever my display refreshes at. I don't have to switch between "23" and "24" in my GPU (though madVR can do that automatically for the most part), I just use one refresh rate for everything that's close.

Also, I'm decoding audio, and using ReClock is the only free way to get WASAPI exclusive output, which allows the receiver to switch number of channels automatically depending on the source.

[Budtz]

I have a question that is not specifically related to madvr but I’m at my wit’s end and dont know what else to do than to ask you fine people at this forum.

I have a Denon dbp-2010 bluray player and a Samsung d6500 tv.

I have confirmed both tv and player are set to 24p/hz

About every 15 minutes i get a small stutter.

It seems like the player and the TV cannot sync up probably and thus the TV need to ether skip or repeat a frame. Could this be right and can any1 shed some light on how TV's and standalone players work together and would I could do? I have tried all the settings with motion-settings and stuff setting the tv in game mode. It dosn’t help

[SamKook]

Is there a way to disable the "Windowed" and "Exclusive" OSD message when switching from them?

I have the OSD option off in MPC-HC 1.6.2.4902 and I couldn't find any OSD options in MadVR 0.82.5

BTW, I still want to use the exclusive mode so disabling it isn't a solution.

[Mikey2]

It's not handling this though, here is my testing with the file:

Played with EVR-CP = too bright (can see it in the black borders immediately)
Played with EVR-CP with the 16-235 -> 0-255 [SD] [HD] shader option enabled = perfect
Played with madVR set at normal 0-255 = too bright, identical to EVR-CP
Played with madVR set at tv 16-235 = even brighter the previous "too bright" result

Any suggestions ?

This is odd, most people have problems the other way around (i.e. the infamous "black crush") Luckily you do not have that since if it is too dark and you compensate the wrong way, then you may lose some data (namely 0-15 and 236-255...)

Actually when you are using the EVR shading, you are doing what's called "double-expansion," which seems to work for you, but is a "no-no" for most people.

First off, gamma shouldn't have anything to do with it. Sure you can mask the incorrect colorspace with a higher gamma setting, but that is not really fixing the problem.

Are you using madVR as your video decoder? If you are using something like LAV or ffdshow and converting to RGB colorspace, there is another 0-25/16-235 setting there.

What kind of graphics card are you using? For NVidia cards, under "Adjust Video Color Settings" -> Advanced, you can set the Dynamic Range there (0-255 or 16-235.)

Finally, what kind of display are you using? Can you adjust the colorspace on the display? Most monitors default to 0-255, while older TV's default to 16-235 (if they are not using YUV 4:2:0/4:2:2 colorspaces.)

To test your native display, do you see all the colors in this image and is black fully black and white fully white?

http://i155.photobucket.com/albums/s297/mikeyurkus/0-255BWDVD.png



Videos are a little trickier, but there are many test-images out there. I usually find the m2ts file with all the black-bars on a Pixar Blu-Ray...

Good luck!

[Xello]

Good luck!

Thanks for the info, I was told about this double expansion thing by another person too.

I'm using Lav video decoder and I did try the options in there relating to colour space. I also tried the different settings in Nvidia control panel, from RGB / YCbCr444 to manually switching between full and limited range in the Video Colour Settings. My display is a plasma (panasonic VT30) and is fine with the image you have, i've used similar clipping patterns and greyscale stepping patterns to set up the TV and it's working fine in that regard. I'm in the middle of sourcing the US version of this film to see if it was just a problem with the japanese one.

[turbojet]

For those looking for a sharpener you might want to try FineSharp (http://forum.doom9.org/showthread.php?p=1569035#post1569035) It's the best I've used but ffdshow is required.

Madshi, is there any chance of adding negative 'b' values for bicubicresize?
I use 2 different resolution screens and swapping the width to set in the autoresize script I found in ffdshow is inconvenient. Also would be nice to offload it to the gpu leaving the 5-10% CPU for other things. Here's a few examples of a 1280x720 video played on 1680x1050 LCD:
MadVR bicubic 75 chroma/luma (http://i48.tinypic.com/352rgg5.png)
bicubic (-0.5,0.75) in ffdshow (http://i47.tinypic.com/987eao.png)
bicubic (-1,0.75) (http://i46.tinypic.com/o0qtkn.png)

Instead of adding more types of bicubicresizers is it possible to use a variable for b and c parameters instead?
B could be named blur with -100, -75, -50, -25, 0, 25, 50, 75, 100
C could be named sharpness with 25, 50, 60(?), 75, 100
Default to 0, 50 = Catmull-Rom and maybe it could replace SoftCubic as well.
It would be nice if the strengths were saved when switching away from Bicubic, they currently don't.

I agree with 6233638 on the resizers on an nvidia gpu but on an intel igpu bicubic 100 seems to work the best without adding artifacts although it's still a little softer than the nvidia with bicubic 75/mitchell-netravali. Maybe it's gpu dependent?

[nevcairiel]

I agree with 6233638 on the resizers on an nvidia gpu but on an intel igpu bicubic 100 seems to work the best without adding artifacts although it's still a little softer than the nvidia with bicubic 75/mitchell-netravali. Maybe it's gpu dependent?

The resizers are implemented with pixel shaders, they should look the same on every GPU, unless Intel did something odd to their implementation, or for some reason post-processing is being triggered.

[ryrynz]

Madshi, I know you weren't particularly interested in cosmetics at this stage, but the "minimized" window from MPC that occasionally springs up before playing the next file is the only thing that's affecting otherwise perfect playback.
I hope you can patch that up in your next cycle.

[zhengwy]

Is there any plan to make translations of the Madvr release?

[chros]

If you watch mostly Full HD material on a Full HD screen, the Luma scaling is not of much use for you, so the setting isn't all that important.
And, is Chroma Upsampling performed when playing 1080p material on 1080p display?
Thanks

[nevcairiel]

And, is Chroma Upsampling performed when playing 1080p material on 1080p display?
Thanks

On most content, yes.
Chroma is usually only quarter the resolution, so it needs to be upscaled.

[mzso]

On most content, yes.
Chroma is usually only quarter the resolution, so it needs to be upscaled.

Woulnd't it be better if the decoder would always output. RGB 4:4:4 and encoders/decoders would handle palette reduction and whatnot internally?

This color subsampling thing feels like a really cheap way to save bandwidth/storage space. (Just like interlacing)

[G_M_C]

Woulnd't it be better if the decoder would always output. RGB 4:4:4 and encoders/decoders would handle palette reduction and whatnot internally?

This color subsampling thing feels like a really cheap way to save bandwidth/storage space. (Just like interlacing)

The decoder decodes, the renderer does rendering to match display/output requirements. Thats the way it's done. And each component does it's best.

Source (4:2:0) -> Decoder (4:2:0) -> Renderer (4:2:0 -> 4:2:2/4:4:4/RGB24 etc.) <--> HDMI.

[nevcairiel]

Woulnd't it be better if the decoder would always output. RGB 4:4:4 and encoders/decoders would handle palette reduction and whatnot internally?

A decoder should output the data as it was stored in the file. If the file is encoded as 4:2:0 (meaning quarter chroma), the decoder should output this. The renderer will usually do a much better job upsampling the chroma.

[mzso]

A decoder should output the data as it was stored in the file. If the file is encoded as 4:2:0 (meaning quarter chroma), the decoder should output this. The renderer will usually do a much better job upsampling the chroma.

I meant something a bit different. By not using chroma subsampling at all. Instead the encoder would reduce/compress the chroma information, more "intelligently" adaptively instead. What I think is that bitrate relative quality would be better. And in this occasion the decoding would naturally always output rgb 4:4:4, and no upsampling be done by anything.

(I'm not sure but it might require a video compression format that doesn't exist at the moment)

This subsampling thing feels like to me as if for audio, they'd just decrease the resolution and/or sampling rate instead of using complicated algorithms for lossy compression, and try to restore as much as possible with upsampling.

[Keiyakusha]

mzso
This is not related to encoder at all, this is standard colorspace. All things created for mass distribution are already 4:2:0. It is possible to initially compress video as 4:4:4, but noone interested in it for mass distribution. You won't see the difference anyway (if upscaling algorithm is decent enough). Human's eye sensitive to luma, not chroma. Compression algorithms are advanced enough. Just 4:4:4 is a twice more data to compress for no reason. For the same file size 4:4:4 will look worse than 4:2:0. And having it RGB and not YUV is totally useless.