madVR - high quality video renderer (GPU assisted)

[glc650]

LAV Audio can do lossless decoding of DTS-HD and TrueHD. (DTS requires the Arcsoft dtsdecoderdll.dll copied into LAV's directory)What happens if you don't have this .dll?

When playing a video, there is an additional preference dialogue you can bring up either from the tray icon (if enabled) or the filter list in your player.

http://www.abload.de/thumb/reclock-4yoyms.png (http://www.abload.de/img/reclock-4yoyms.png)

Under no circumstances should you enable the "slave reference clock to audio" option, as that effectively disables ReClock's resampling/sync corrections.What does "Locked" (Media adaption section) do?

[njfoses]

I don't think it launches directly into fullscreen exclusive mode, but if you open MPC-HC's preferences, you can go to Playback > Fullscreen and enable the "Launch files in fullscreen" option.

Yea i have that checked but it doesnt launch into fullscreen exclusive mode.

[agustin9]

Still love my Xp:)
It works great for me, for most things..
I have a Win7 partition on my main computer also but use it only for modern games.
I dont like all the extra fanzy stuff that came along with win7,
I have tweaked it down to look like Xp though:)
Its a taste thing I guess hehe

But does anybody know what it could be that makes it so MadVr doesent work on my main computer?
It is certainly powerful enough..
And it works on the other Xp machines I have tried it on.
But you have 8gb of ram, and 32bit OSs only recognize around 3. You're wasting lots of memory.

On your problem, try different driver versions, older ones. That's other problem with XP, Nvidia and AMD don't give a f*ck about XP anymore and the drivers are shit.

[Dodgexander]

Has anyone noticed a problem with deinterlacing using dxva?

I am running the latest catalyst drivers with my Radeon 6800, but for some reason the only way I can get deinterlacing to work is by forcing it using lav videos option.

If i have ffdshow dxva or the windows decoder, deinterlacing just doesn't turn on despite the control panel settings. This only happens with h264. Not mpeg2.

Seems like Madvr isnt passing the info on that the video needs to be deinterlaced.

How can i troubleshoot? I already reset madvr settings and uninstalled/reinstalled. I have also changed the deinterlacing settings in madvr, even when assuming a file is interlaced, it does not work unless forced in lav.

EDIT* With EVR as renderer, deinterlacing works fine with other decoders.

[6233638]

What happens if you don't have this .dll?A few years ago now, the resampler in ReClock was changed from SSRC to Libresample, which uses a lot less CPU power, but does not sound as good in my opinion. For many (most?) people there is probably no difference though.

The file I linked to is the older SSRC resampler, compiled to work with the current version of ReClock.

Back when dual core CPUs were common and you had to use CPU decoding with madVR, ReClock's CPU usage could be a problem. Now that both madVR and LAV Video support DXVA decoding, and much faster quad core CPUs are common, the CPU usage of SSRC resampling is less of an issue.

What does "Locked" (Media adaption section) do?If you change it away from "Auto (best)" and don't lock it, it will reset back to Auto when you play another file. So it probably doesn't make a difference whether you lock it or not, if you are using Auto anyway.

Yea i have that checked but it doesnt launch into fullscreen exclusive mode.I'm not sure that there's a way to do that then.

Has anyone noticed a problem with deinterlacing using dxva?

I am running the latest catalyst drivers with my Radeon 6800, but for some reason the only way I can get deinterlacing to work is by forcing it using lav videos option.Is this with DXVA Native or Copy-Back in LAV? I seem to recall having issues with DXVA Native.

[Seegs108]

Does anyone have trouble getting any version of MadVR to work on a monitor in portrait mode? Landscape works like a charm but when dragging MPC-HC over to the portrait screen or even starting the video in the portrait screen causes the video to freeze. Every other renderer works fine.

[huhn]

But resampling is a mathematically precise process and resampling in Reclock is done to a suitably high precision. Provided the adjustments Reclock is needing to make are low (<=0.1% if adjusting up to what is needed for NTSC slow down) it should have no noticeable impact for 99.9% of user, except for almost certainly noticeably smoother video. Of course there will always be audiophiles who do not want the audio touched at all and they will need to accept less smooth video. But for most of the population I think getting video smooth is the better trade off.

yeah and that's the most important thing the human ear can't hear the difference. it is not perfect but as perfect as it could be. at least for pc used.

[jmone]

Hello there.
I've checked some pulldown detection patterns: http://www.avsforum.com/t/1276064/1920x1080-24-25-30fps-test-patterns-detect-pulldown-issues

and i do have same issue as the person that posted these patterns ("Pulldown / Frame Interpolation: One thing I've noticed that is interesting with the test files, it shows you what pull down and frame interpolation is being applied by the PC and/or TV. Each of the test files is only encoded with only 1 red square per frame. So when playing say the NTSC test file (29.97) you see two red squares when played at 60hz...in this case (I guess) the video renderer or TV displays each frame twice but overlaps them with the following frame. Now the interesting thing is when I play the Blu-ray test file (23.976) at 24hz I see 3 red boxs on my setup. As this is not being done in the PC, it looks like it must be the TV doing a 3:3 frame interpolation (though the documentation on the PIO LX608 is pretty vague on it's settings and options here)."

i do have screen forced to 72hz,and using reclock,but it looks like something in the video chain is putting some frame interpolation and i see 3(or 2,its really hard to see) red boxes. ive tried to force vsync,and play with pulldown detection in catalyst control center but in vain. Im using madvr+newest lav filters+reclock

im using xp 32bit,hd4870 and dell 2209wa (the 3squares also appear when i run 30fps test on 60hz,so i dont think its a screen issue,no one else reported that too)

That would be me and why I created these test patterns as it was driving me nuts. I too have a Pio and under Options --> Drive Mode I have selected option "2" as it is the most natural to my eyes. These three options change how the Pio does frame rate interpolation and as you have seen with these test patters you can actually see what the Pio is doing. ...and you are right - it is very poorly documented what the settings do on the Pio!

[glc650]

A few years ago now, the resampler in ReClock was changed from SSRC to Libresample, which uses a lot less CPU power, but does not sound as good in my opinion. For many (most?) people there is probably no difference though.

The file I linked to is the older SSRC resampler, compiled to work with the current version of ReClock.

Back when dual core CPUs were common and you had to use CPU decoding with madVR, ReClock's CPU usage could be a problem. Now that both madVR and LAV Video support DXVA decoding, and much faster quad core CPUs are common, the CPU usage of SSRC resampling is less of an issue.I was actually asking about the Arcsoft dtsdecoderdll.dll. What happens if you don't have it?

[jmone]

You should buy Arcsoft TMT. Others may have found that downloading the trial of TMT will also provide access to that file (without a timelimit) but the ongoing use of it is up to your moral stance on such things.

[glc650]

You should buy Arcsoft TMT. Others may have found that downloading the trial of TMT will also provide access to that file (without a timelimit) but the ongoing use of it is up to your moral stance on such things.I didn't ask why I should pay for it I asked what it does. My reclock setup seems to work fine without it with DTS-HD tracks so why do I need this .dll?

[Jong]

Without it LAV will decode the old- style, lossy, DTS core (included for compatibility as part of all DTS-HD soundtracks), so you are not getting HD, lossless audio.

[jmone]

Appologies glc650 - I misread what you asked. As Jong said, there is currently no way of decoding the full DTS-HD MA with feeware software. The Arcsoft dtsdecoderdll.dll file however will be called on my LAV if it exists to solve this issue.

[FlygeSoppen]

But you have 8gb of ram, and 32bit OSs only recognize around 3. You're wasting lots of memory.

On your problem, try different driver versions, older ones. That's other problem with XP, Nvidia and AMD don't give a f*ck about XP anymore and the drivers are shit.

I know.. win7 is alot faster and all that, but I run older programs and games that does not work as good in win7.
So my main operating system is still winxp.
And I allso love old things the best:)

By the way, I have tried all of the older drivers i can use with the 670gtx graphics card. But I cant go any lower than 300 series of drivers since my graphics card is to new.

But the same drivers works good with the Geforce620. So I dont know... Its not a WinXp issue and not a drivers issue, it must be something else i think hmmmm.

The screen is completely black and MPC-HC hangs for about 1 minute then laggy pictures with choppy sound, just like my pc was some old pentium or something.

Thanks for the answers:) But I dont think I can find the solution by myself, think I have tried it all.

[Devrim]

Hi,

I have 2 monitors which are connected to different GPU. My first screen is connected to a Nvidia card and the second one to an Intel card. Whenever I switch my MPC-HC window to my second screen with the Intel GPU the 1080i clip I'm watching is stuttering. Is it possible to make madVR use the Nvidia card even if it is on the second monitor with Intel GPU? Thanks! (This only seems to happen with 1080i content btw!)

[MSL_DK]

http://www.abload.de/thumb/reclock-2wzotd.png (http://www.abload.de/img/reclock-2wzotd.png)

Page 2: Set the "assumed framerate when not found" to what is most common with the formats you watch, or just leave it at <unknown>.

What about Hardware Interface? Are the correct setting not DirectDraw with madVR?

[konakona]

That would be me and why I created these test patterns as it was driving me nuts. I too have a Pio and under Options --> Drive Mode I have selected option "2" as it is the most natural to my eyes. These three options change how the Pio does frame rate interpolation and as you have seen with these test patters you can actually see what the Pio is doing. ...and you are right - it is very poorly documented what the settings do on the Pio!

well i only used your pio as an example-im using dell 2209wa

[6233638]

I was actually asking about the Arcsoft dtsdecoderdll.dll. What happens if you don't have it?If you're bitstreaming, nothing. If you're decoding, you get the DTS-core track rather than the DTS-HD track.

Now the interesting thing is when I play the Blu-ray test file (23.976) at 24hz I see 3 red boxs on my setup.If you are at 72Hz, you will see three images. If you were at 48Hz, you would have a double image etc. That's just how display technology works. It's not interpolation.

That would be me and why I created these test patterns as it was driving me nuts. I too have a Pio and under Options --> Drive Mode I have selected option "2" as it is the most natural to my eyes. These three options change how the Pio does frame rate interpolation and as you have seen with these test patters you can actually see what the Pio is doing. ...and you are right - it is very poorly documented what the settings do on the Pio!Any interpolation the Pioneer sets do is controlled by the PureCinema Film Mode options, not the Drive mode.

Drive mode sets the refresh rate used with PAL (50Hz) signals.
Drive Mode 1: 75Hz - only suitable with films.
Drive Mode 2: 100Hz - suitable for film and video content
Drive Mode 3: 60Hz - Stupid

The reason for this is the Plasma image quality gets worse the higher the refresh rate goes, and running at 50Hz natively flickers too much. Realistically though, this should always be set at 2, unless you never watch anything other than film sources.


To enable 72Hz with 24p playback, you need Film Mode set to Advance.

What about Hardware Interface? Are the correct setting not DirectDraw with madVR?I thought madVR was using Direct3D. Best to leave it on Auto.

[Dodgexander]

Is this with DXVA Native or Copy-Back in LAV? I seem to recall having issues with DXVA Native.

The problem isn't only with Lav. I just used lav as an example, since unlike other decoders, it lets you "force" deinterlacing.

Using the Microsoft or FFDSHOW dxva decoders, you can't "force" deinterlacing. Only in the control panel settings can you change the deinterlacing setting and this does not work with madvr


Strange huh? Never had this issue before with madvr

[6233638]

The problem isn't only with Lav. I just used lav as an example, since unlike other decoders, it lets you "force" deinterlacing.

Using the Microsoft or FFDSHOW dxva decoders, you can't "force" deinterlacing. Only in the control panel settings can you change the deinterlacing setting and this does not work with madvr


Strange huh? Never had this issue before with madvrHave you tried using LAV's DXVA Copy-Back option with madVR?

[Dodgexander]

Thanks 6233638 and yes, it happens both with copy-back and native. The only way I can get hardware accelerated deinterlacing to work with madvr is by setting lav to aggressive or forced and as I said, other decoders, deinterlacing doesn't work at all with madvr.

Edit*

I have tried now:

Reinstalling Latest AMD drivers for my graphics card.
Reinstalling madvr
Reinstalling lav
Reinstalling mpc
Changing deinterlacing preferences in madvr "if in doubt activate deinterlacing".

None have worked.

[pururin]

With Intel iGPU (HD3000/4000), the good scaling algo is DXVA2? Or bicubic75+lanczos3AR should be better?

And when select luma scaling to DXVA2, chroma can be set to anything and it'll still use intel's internal algo right?

[glc650]

If you're bitstreaming, nothing. If you're decoding, you get the DTS-core track rather than the DTS-HD track.Is there a way to verify the dtsdecoderdll.dll is being used?

[sneaker_ger]

LAV Audio's status tab will say "dts-hd ma" under input "codec". But this really belongs in the LAV Filters thread - not the madVR thread.

[njfoses]

Question regarding madvr deinterlacing. I very rarely watched interlaced content. When i do it is generally ntsc 1080i tv sitcoms. Should i have madvr autodetect for deinterlacing or force film mode? It seems film mode would be used for most content but im unaware of any way to know when to switch to video which makes me think auto detection would be best.

[ryrynz]

With Intel iGPU (HD3000/4000), the good scaling algo is DXVA2? Or bicubic75+lanczos3AR should be better?

Use Lanczos 3 AR, it's much better. In some cases on an HD4000 you can use Jinc 3 AR.


And when select luma scaling to DXVA2, chroma can be set to anything and it'll still use intel's internal algo right?

Correct.

[ice25]

* added double/triple expanded TV range to "source levels" toggle


Could you please add separate keyboard shortcuts for those? Thanks.

[DragonQ]

Question regarding madvr deinterlacing. I very rarely watched interlaced content. When i do it is generally ntsc 1080i tv sitcoms. Should i have madvr autodetect for deinterlacing or force film mode? It seems film mode would be used for most content but im unaware of any way to know when to switch to video which makes me think auto detection would be best.
If you force film mode it won't deinterlace interlaced content. In most cases, deinterlacing being turned on when it's not needed won't have any negative effects (particularly with hardware deinterlacing), whereas not deinterlacing an interlaced clip will make it look like arse.

Therefore, leaving it on auto is usually the best thing to do:
http://www.aotplaza.com/Files/HTPC/Screengrabs/MadVR/madVR%20Settings.png
You can check whether deinterlacing has been turned on for a particular video by pressing CTRL + ALT + SHIFT + D.

[Dodgexander]

I tried Cyberlinks decoder and deinterlacing works fine, so for LAV it needs to be aggressive, for microsoft or ffdshow dxva deinterlacing doesn't work at all.

Basically any decoder using the controls from the AMD Driver does not work.

[njfoses]

If you force film mode it won't deinterlace interlaced content. In most cases, deinterlacing being turned on when it's not needed won't have any negative effects (particularly with hardware deinterlacing), whereas not deinterlacing an interlaced clip will make it look like arse.

Therefore, leaving it on auto is usually the best thing to do:
http://www.aotplaza.com/Files/HTPC/Screengrabs/MadVR/madVR%20Settings.png
You can check whether deinterlacing has been turned on for a particular video by pressing CTRL + ALT + SHIFT + D.

If film mode doesn't deinterlace what would it be used for? Seems many people say film mode is the "correct" deinterlace option for most content. Like i said i watch very little interlaced content but i just want to make sure when i do i have the best possible setting.

[Qotscha]

Film mode removes pulldown (http://en.wikipedia.org/wiki/Telecine#Reverse_telecine_.28a.k.a._inverse_telecine_.28IVTC.29.2C_reverse_pulldown.29) from telecined content.

[madshi]

I presume that LAV can't IVTC? I often play 29.97@59.94 1080i with LAV using CUVID, and Reclock sees 59.94 sees as expected.
I think LAV can remove soft-telecine, but not hard-telecine. Not sure, though, you'd have to ask nevcairiel. At least it can't do full IVTC in the same way that DScaler IVTC Mod and madVR can do, AFAIK.

I'm using LAV Splitters/Decoders in PotPlayer and MPC's default splitters and decoders in MPC; both give the same freeze. I also performed a clean PotPlayer install, that is to say it was using its internal decoders/splitters, when selecting madVR the freeze occurred again.

The freeze actually occurs with all files equally, whether AVC1, XVID or WMV but if you still want a sample of a file let me know and I'll upload one.
Weird. Could you please double check with LAV Splitter/Decoders in MPC-HC? Are you using software or hardware decoding? Try software decoding. If the problem then still occurs, please upload a sample where it's especially obvious, and please tell me what you're doing exactly in MPC-HC to reproduce the problem. Thanks.

Thanks, I've found the "settings.bin" file in C:\Users\Peuj\AppData\Local\VirtualStore\Program Files (x86)\LAV Filters\x86\madVR

I've looked in the bat file and the way it's written, the "settings.bin" file cannot found.
Argh, sometimes Windows is too clever for its own good. I'd prefer if Windows simply blocked the creation of files if the folder is not writable, instead of storing them in some wild folder nobody ever looks into.

Okay, after installing another player, disabling hardware decoding, uninstalling my mouse drivers, and changing to using Aero from not using Aero, I figured out what it is. Pointer trails. If you have show pointer trails on in the mouse config, you cannot see the mouse cursor in exclusive mode. (At least on my system) So I'd like to report a bug :sly:
That's got nothing to do with madVR, though. I don't know why this is as it is, and there's nothing I can do about it.

is anyone else having this problem?
http://truexfan81.webs.com/madvr.jpg

seems to only happen when playing xvid files in fullscreen mode using madvr, tried evr, the vid looks normal

i use madvr 0.85.7
chroma: Bicubic 75 AR
luma: SoftCubic 100

thanks

update: activating scale in linear light or AR filter on luma seems to fix it
Nobody else reported this problem. Which GPU and OS are you using? What happens if you change the zoom factor a bit (e.g. NumPad 1 in MPC-HC)? Does the problem go away then or is it still there?

What is the recommendation for "use a separate device for presentation" on Windows 7 x64?
Enable it if it makes no problems.

Madshi, where is the postprocessing sharpness algorithm Gpu based
Some time in the future maybe. Not now.

1- frame step seems buggy if i want to step many frames MPC get stuck for 1-2 second :(
Can you tell me how exactly I can reproduce this problem on my PC?

2- colors look darker and washed out :(
The question is whether madVR is too dark or whether EVR is too bright. Please make sure the media player color controls are set to neutral. Then please double check the GPU driver control panel options for video. They should usually be set to "application decides" (or something like that). Finally, try switching madVR output levels to see whether switching to TV or PC levels makes the image look correctly. And of course make sure that what EVR gives you is *really* the correct brightness.

1. In an ideal world, with the perfect graphics card and CPU, what resizer settings give the best results ?
Chroma and Luma upscaling: Jinc3 with anti-ringing turned on and linear light turned off.
Luma downscaling: Catmull-Rom with anti-ringing and linear light turned on.

when I open the video and I use any of the shortcuts in mpchc for fullscreen (there are 2, one that says fullscreen and the other that says fullscreen without resolution change) for a split second the video gets stretched out (covering the black bars) and goes back to it's normal fullscreen/resolution.
Maybe I can improve this in a future build, but it's not high priority for now. I consider this a minor cosmetical problem, and there are important key features still missing. So those have priority for now...

Hi, when I use subtitles with 720p videos in ZoomPlayer with madVR, it uses 100% of my CPU and drops frames like crazy with VobSub and ffdshow subtitle filter.
Well, this seems to be an issue with the subtitle renderers. Not sure what this has to do with madVR.

Instead of cropping to mod4 what about padding to mod4 instead? But can this only be done when it really needs it, deinterlacing?
Possible, but not as easy as it may sound.

Is there any way to control gpu ram in dxva2?
Not that I know of. Of course using a lower GPU queue size in madVR saves GPU RAM.

Madshi, do you know why this MBAFF AVC video (http://www.mediafire.com/?lutsoixr0lcagub) isn't deinterlaced by MadVR in auto mode (using LAV & MPC-HC)? EVR plays it back fine (25 fps during first ~10 s, 50 fps for last ~20 s) but MadVR never turns deinterlacing on.
It looks like the first few seconds are actually progressive, and this information also ends up in the media type, which triggers madVR to believe the file is actually progressive.
LAV also thinks this at first, but once it switches to MBAFF, LAV notices and informs madVR with a new media type (switches from interlace flags 0x0 to 0x81), but this is not enough, apparently. ;)
Hmmmm... I guess I'm only checking whether deinterlacing should be turned on/off at the start of playback. Maybe I can improve that in case the decoder updates the media type.

Interestingly, although the "Aggressive Deinterlacing" workaround works for the ~hour of interlaced video after the progressive title sequence, it doesn't work for the pre-titles section, which is definitely interlaced video but not being detected as such. Tis possible the metadata is wrong in this section. Sample here. (http://www.mediafire.com/?yb995bwznf8ney7)
This sample looks like 25p, from the point of view of the decoder. It doesn't get deinterlaced by EVR, either.

At some point in the future I plan to add an automatic progressive vs. film vs. video detection to madVR. That should then detect such cases, too, but it could be a while until I get to that...

IRC channel anyone? Maybe we could start one on NewNet?
I've no problem at all if you guys want to start an IRC channel. But please don't expect me to join.

just wondering if I will get frame drops no matter what I do unless i get reclock? I managed to get my display refresh to a 23.976xx something - madVR displays that, but still I get a 1 frame repeat every 12-13 minutes - or so the OSD shows.
This information in the OSD is just an estimate. It isn't always accurate. Trust your eyes. Is playback smooth?

It should be possible to get smooth playback without Reclock.

I honestly don't think matching rates can stop frame drops repeats. I was able to match my playback rate to my display rate reported by madvr down to like 0.0001hz using svp interpolation.

I still get drops. Madshi suggested it could be due to the clock deviation so your display will never exactly be 23.976. For those brief moments I guess you could drop a frame. That's my understanding
I actually made adjustments to my playback rate upwards in tiny increments using svp until I got 0 frame drops but can't really be sure that im not getting frame repeats. I think there's always a need for some software solution to sort out this type of problem I guess. Matching rates is good but not enough.
Actually matching rates *should* be good enough. Audio clock deviation should be constant, so it's not like at some point in time suddenly the clock deviates and madVR has to drop a frame. madVR drops frames to keep audio and video in sync when there's a constant drift between audio and video clocks. If you match clocks reasonably well, there should not be frame drops/repeats for several hours at least.

Please remind, do I need to set Catalyst video to 16-235 for DXVA2 de-interlacing no matter I use 0-255 for output? Does it affect DXVA decoding or anything else somehow?
You should set all options so that the application decides. For "dynamic range" (I guess that's what you mean) in the Catalyst driver, simply disable it.

Is it known problem that madVR can't get the EDID from TV after TV turn off/turn on? I can get the picture but "device" is greyed in madVR's properties and any adjustments are not possible untill the whole PC reboot.
No, that is definitely not a known problem. Does restarting the media player not help, either? Sounds like a driver/OS bug to me. These things work just fine here. And nobody else reported this problem yet.

In fact targetting a perfect 23.976024 isn't a good idea.
I managed once to get 23.97602 but still dropped frames... it's because that number in madvr isn't exact.
Not sure what you mean. The refresh rate measurement in madVR is very exact - but it's measured with the system clock, not with the audio hardware clock. You're right that trying to make the measured refresh rate be exactly 24.000/1.001 isn't always the correct solution. Deviation between system and audio hardware clocks need to be taken into account...

I though madVR displays an accurate refresh rate
It does. But it's measured with the system clock. Which is usually not perfectly identical to the audio hardware clock. Which is the cause of all the problems...

madshi> Are your own calibration solution still on your to do list?
It's still on my list of things to look into.

[madshi]

Start clip, move MPC-HC to second screen, wait for video to start playing again, immediately close MPC-HC.

madVR log:http://www.sendspace.com/file/54c6u4
Creation of the Direct3D device seems to fail after switching the monitor, with Direct3D returning: "access denied". I don't know why this happens. Maybe it has to do with DXVA scaling being turned on? Try if turning it off fixes the issue?

P5K Premium mainboard, QX6850 Quad core 3ghz(overclocked to 3.6ghz), 8gb RAM, Geforce 670 card, Xfi soundcard and WinXP.
It did not work out that well. I have tryed to follow different guides on the net without any success.

When all is installed and configured just like the other computers, all I get is about 1minute of hanging blackness
with choppy sound and then some laggy film afterwards.. Its the same in window and fullscreen.
Please make sure you set things like Anti-Aliasing, Anisotropic filtering etc to "let application decide" (or whatever it is named). If that doesn't help, maybe I can see what's going on if you upload a madVR debug log (please zip it!).

I've checked some pulldown detection patterns: http://www.avsforum.com/t/1276064/1920x1080-24-25-30fps-test-patterns-detect-pulldown-issues

and i do have same issue as the person that posted these patterns(long quote here) "(Pulldown / Frame Interpolation: One thing I've noticed that is interesting with the test files, it shows you what pull down and frame interpolation is being applied by the PC and/or TV. Each of the test files is only encoded with only 1 red square per frame. So when playing say the NTSC test file (29.97) you see two red squares when played at 60hz...in this case (I guess) the video renderer or TV displays each frame twice but overlaps them with the following frame. Now the interesting thing is when I play the Blu-ray test file (23.976) at 24hz I see 3 red boxs on my setup. As this is not being done in the PC, it looks like it must be the TV doing a 3:3 frame interpolation (though the documentation on the PIO LX608 is pretty vague on it's settings and options here)."

i do have screen forced to 72hz,and using reclock,but it looks like something in the video chain is putting some frame interpolation and i see 3(or 2,its really hard to see) red boxes. ive tried to force vsync,and play with pulldown detection in catalyst control center but in vain. Im using madvr+newest lav filters+reclock,playing using mpc-hc,screen is dvi-d connected with custom-reduced blanking times to allow 72hz on single-link using powerstrip.

im using xp 32bit,hd4870 and dell 2209wa (the 3squares also appear when i run 30fps test on 60hz,so i dont think its a screen issue,no one else reported that too)
Not sure if there's a question hidden somewhere in your post? madVR does not do frame interpolation, if that's what you want to know...

But resampling is a mathematically precise process
You make it sound as if there's exactly only one right result of doing resampling. But that's far from the truth. For audio resampling the same problems apply as for video resampling: Depending on resampling filter characteristics, you get aliasing, ringing or loss of high frequency response. Or all of that.

Provided the adjustments Reclock is needing to make are low (<=0.1% if adjusting up to what is needed for NTSC slow down) it should have no noticeable impact for 99.9% of user
Another misconception. Whether you adjust by 0.1% or 5% doesn't make much of a difference. Even running the audio through the resampler with exactly 0% sampling rate change will already introduce all the artifacts like ringing etc. Actually, if you do have to use resampling, your best bet is to upsample to as large a sampling rate as you can. This way you get some positive changes in addition to the artifacts, too, at least. But upsampling to high sample rates comes with its own problems. DACs might not work as well with high sampling rates as they work for lower rates. So whether resampling to high sampling rates actually sounds better will depend on the actual hardware you're using.

ReClock has to resample audio to work, but it is high quality, and any theoretical loss of quality should be imperceptible.
ReClock also gives you the option to upsample the audio if your hardware supports it, in which case there should be no loss in quality at all.
You'll get ringing in any case. Audio upsampling isn't so very different from video upsampling.

Under no circumstances should you enable the "slave reference clock to audio" option, as that effectively disables ReClock's resampling/sync corrections.
I disagree. If you have managed to get the refresh rate close enough, there shouldn't be frame drops/repeats long enough to play the longest movie out there, rendering Reclock's resampling/sync corrections useless. In that case it does make a lot of sense to enable the "slave reference clock to audio" option, so that you can get kernel mode streaming without having to resample audio.

Has anyone noticed a problem with deinterlacing using dxva?

I am running the latest catalyst drivers with my Radeon 6800, but for some reason the only way I can get deinterlacing to work is by forcing it using lav videos option.

If i have ffdshow dxva or the windows decoder, deinterlacing just doesn't turn on despite the control panel settings. This only happens with h264. Not mpeg2.
What does the OSD say about deinterlacing for those h264 files? If it's disabled, what happens if you press Ctrl+Alt+Shift+D twice to manually enable deinterlacing in madVR? Does that work? Can you upload a small sample with which I can reproduce the problem on my PC?

Does anyone have trouble getting any version of MadVR to work on a monitor in portrait mode?
See "limitations" in the first post of this thread.

I have 2 monitors which are connected to different GPU. My first screen is connected to a Nvidia card and the second one to an Intel card. Whenever I switch my MPC-HC window to my second screen with the Intel GPU the 1080i clip I'm watching is stuttering. Is it possible to make madVR use the Nvidia card even if it is on the second monitor with Intel GPU? Thanks! (This only seems to happen with 1080i content btw!)
madVR by default renders on the GPU which is connected to the target monitor. It would be theoretically possible to render on a different GPU, then transfer the rendered pictures to system RAM, then copy them to the other GPU, then display them there. But that would be a lot of extra work. It would also make the configuration dialog more complicated. So I don't plan to add such a feature in the near future.

Question regarding madvr deinterlacing. I very rarely watched interlaced content. When i do it is generally ntsc 1080i tv sitcoms. Should i have madvr autodetect for deinterlacing or force film mode? It seems film mode would be used for most content but im unaware of any way to know when to switch to video which makes me think auto detection would be best.
Currently madVR can not find out itself which mode it should use. So the "auto" mode always uses DXVA deinterlacing. If your sitcoms are native film content (24p), forcing madVR into "film" mode would allow you to output true 24p to your display, which would result in smoother motion. This is not possible when using DXVA deinterlacing. However, if your sitcoms are native interlaced recordings, film mode would produce visible combing artifacts. If your display can't do 24p, anyway, I'd stick with "auto". If you want 24p output then you can try film mode. I guess most newer sitcoms are 24p based. But some older sitcoms are native interlaced, I think.

Could you please add separate keyboard shortcuts for those? Thanks.
I'm sorry, I don't plan to. I mean I could add separate shortcuts for pretty much *everything*, but I'd end up with 2 million shortcuts, and I'd like to avoid that. I think the toggle shortcut is good enough, isn't it?

If you force film mode it won't deinterlace interlaced content.
If you say it like this you might confuse some users. Many people consider telecined content as "interlaced content", too. It's important to be clear. If you say "film mode won't properly deinterlace *native* interlaced content" then I fully agree with you.

If film mode doesn't deinterlace what would it be used for? Seems many people say film mode is the "correct" deinterlace option for most content. Like i said i watch very little interlaced content but i just want to make sure when i do i have the best possible setting.
There are 2 different kinds of "interlaced content": Truely native interlaced content, and telecined content. The latter was originally 24p (or 25p) content and was just converted to 60i (or 50i). Film mode can perfectly undo this and restore the original 24p/25p content. Film mode does not work properly with truely native interlaced content, though.

[DragonQ]

Hmmmm... I guess I'm only checking whether deinterlacing should be turned on/off at the start of playback. Maybe I can improve that in case the decoder updates the media type.
Sounds like a good idea. :)

If you say it like this you might confuse some users. Many people consider telecined content as "interlaced content", too. It's important to be clear. If you say "film mode won't properly deinterlace *native* interlaced content" then I fully agree with you.
Fair enough; I rarely consider "telecined content" (2:3 pulldown) since it doesn't exist in "PAL regions", thus I never encounter it.

Also, wouldn't that technically be detelecining/IVTC, not deinterlacing? ;)

[njfoses]

Currently madVR can not find out itself which mode it should use. So the "auto" mode always uses DXVA deinterlacing. If your sitcoms are native film content (24p), forcing madVR into "film" mode would allow you to output true 24p to your display, which would result in smoother motion. This is not possible when using DXVA deinterlacing. However, if your sitcoms are native interlaced recordings, film mode would produce visible combing artifacts. If your display can't do 24p, anyway, I'd stick with "auto". If you want 24p output then you can try film mode. I guess most newer sitcoms are 24p based. But some older sitcoms are native interlaced, I think.


If you say it like this you might confuse some users. Many people consider telecined content as "interlaced content", too. It's important to be clear. If you say "film mode won't properly deinterlace *native* interlaced content" then I fully agree with you.


There are 2 different kinds of "interlaced content": Truely native interlaced content, and telecined content. The latter was originally 24p (or 25p) content and was just converted to 60i (or 50i). Film mode can perfectly undo this and restore the original 24p/25p content. Film mode does not work properly with truely native interlaced content, though.



Can mediainfo or a similar program determine natively interlaced vs telecined?

[madshi]

Fair enough; I rarely consider "telecined content" (2:3 pulldown) since it doesn't exist in "PAL regions", thus I never encounter it.
Telecined PAL content is 50i. It absolutely exists. E.g. I have samples of a PAL DVD where each encoded frame contains the bottom field of the previous frame and the top field of the next frame. Without proper IVTC you'd either get combing or 50p output instead of 25p.

Also, wouldn't that technically be detelecining/IVTC, not deinterlacing? ;)
I consider "video mode deinterlacing" and "IVTC" as being different "deinterlacing" modes.

Can mediainfo or a similar program determine natively interlaced vs telecined?
No. At least not for hard-telecined content.

[Jong]

You make it sound as if there's exactly only one right result of doing resampling. But that's far from the truth. For audio resampling the same problems apply as for video resampling: Depending on resampling filter characteristics, you get aliasing, ringing or loss of high frequency response. Or all of that.

Another misconception. Whether you adjust by 0.1% or 5% doesn't make much of a difference. Even running the audio through the resampler with exactly 0% sampling rate change will already introduce all the artifacts like ringing etc. Actually, if you do have to use resampling, your best bet is to upsample to as large a sampling rate as you can. This way you get some positive changes in addition to the artifacts, too, at least. But upsampling to high sample rates comes with its own problems. DACs might not work as well with high sampling rates as they work for lower rates. So whether resampling to high sampling rates actually sounds better will depend on the actual hardware you're using.

You'll get ringing in any case. Audio upsampling isn't so very different from video upsampling.
Its a fair cop and my language was sloppy, but what I was trying to do was compare the Reclock resampler, which is at least a defined formula, calculated to high internal accuracy, with the option to upsample, with pitch correction, which involves chopping audio up into time slices and the processing those 2-channels at a time with loads of potential for both audible artefacts at slice boundaries and loss of phase coherence between each channel pair, e.g. Fronts and rears. IMO the resampling in Reclock has the potential to be inaudible, if not precise(!). Pitch correction, not so much.

And the point about 0.1% was that the tempo change of this small adjustment should be inaudible (possible artefacts aside). Not that the artefacts, such as they exist, would be less because the adjustment was small.

But yeah, talking about precision is dangerous:o

[DragonQ]

Telecined PAL content is 50i. It absolutely exists. E.g. I have samples of a PAL DVD where each encoded frame contains the bottom field of the previous frame and the top field of the next frame. Without proper IVTC you'd either get combing or 50p output instead of 25p.
If that was deinterlaced in "video mode", as long as it was correctly detected as BFF, it'd just be output as 50p with every other frame being a repeat of the previous one, surely? In the same way that playing 25p content on a normal TV would actually be 50p/75p/100p with repeated frames. If not, then I've never seen that.

[Qaq]

Is it known problem that madVR can't get the EDID from TV after TV turn off/turn on? I can get the picture but "device" is greyed in madVR's properties and any adjustments are not possible untill the whole PC reboot.No, that is definitely not a known problem. Does restarting the media player not help, either? Sounds like a driver/OS bug to me. These things work just fine here. And nobody else reported this problem yet.
Restarting the media player does not help of course. OK, so I guess its just a driver (XP) bug then. I'll let you know if that also happen in win7.

[nevcairiel]

If that was deinterlaced in "video mode", as long as it was correctly detected as BFF, it'd just be output as 50p with every other frame being a repeat of the previous one, surely? In the same way that playing 25p content on a normal TV would actually be 50p/75p/100p with repeated frames. If not, then I've never seen that.


Video Mode deinterlacing of telecined PAL just results in 50p, even if it would be supposed to only be 25p, so usually we never notice any such problems.

[madshi]

what I was trying to do was compare the Reclcock resampler, which is at least a defined formula, calculated to high internal accuracy, with the option to upsample, with pitch correction, which involves chopping audio up into time slices and the processing those 2-channels at a time with loads of potential for both audible artefacts at slice boundaries and loss of phase coherence between each channel pair, e.g. Fronts and rears. IMO the resampling in Reclock has the potential to be inaudible, if not precise(!). Pitch correction, not so much.

And the point about 0.01% was that the temp change of this small adjustment should be in audible (possible artefacts aside). Not that the artefacts, such as they exist, would be less because the adjustment was small.
Ok, fair points, I have to agree.

If that was deinterlaced in "video mode", as long as it was correctly detected as BFF, it'd just be output as 50p with every other frame being a repeat of the previous one, surely? In the same way that playing 25p content on a normal TV would actually be 50p/75p/100p with repeated frames. If not, then I've never seen that.
Video Mode deinterlacing of telecined PAL just results in 50p, even if it would be supposed to only be 25p, so usually we never notice any such problems.
Yes, running this through a video mode deinterlacer (e.g. YADIF) should produce acceptable image quality with 50 different progressive frames being the result, but image quality would not be as good as properly done IVTC, of course.

[nevcairiel]

Hardware Deinterlacers can detect the cadence if you're lucky and produce proper quality, just with duplicate frames.

[madshi]

Yes, but at least my AMD card doesn't handle PAL film content well at all. It very often switches to video mode deinterlacing.

[glc650]

I disagree. If you have managed to get the refresh rate close enough, there shouldn't be frame drops/repeats long enough to play the longest movie out there, rendering Reclock's resampling/sync corrections useless. In that case it does make a lot of sense to enable the "slave reference clock to audio" option, so that you can get kernel mode streaming without having to resample audio.What's the threshold for "close enough"?

Currently madVR can not find out itself which mode it should use. So the "auto" mode always uses DXVA deinterlacing. If your sitcoms are native film content (24p), forcing madVR into "film" mode would allow you to output true 24p to your display, which would result in smoother motion. This is not possible when using DXVA deinterlacing. However, if your sitcoms are native interlaced recordings, film mode would produce visible combing artifacts. If your display can't do 24p, anyway, I'd stick with "auto". If you want 24p output then you can try film mode. I guess most newer sitcoms are 24p based. But some older sitcoms are native interlaced, I think.And if you force video mode it also uses DXVA deinterlacing? What does it does when you force film mode?

No. At least not for hard-telecined content.What's the best way to determine if content is natively interlaced vs telecined?

[madshi]

What's the threshold for "close enough"?
When you can playback a full long movie without frame drops/repeats.

And if you force video mode it also uses DXVA deinterlacing? What does it does when you force film mode?
Yes, force video = auto = DXVA deinterlacing at the moment. Film mode uses madVR's own IVTC algorithm, which currently runs on the CPU, which means it isn't compatible to native DXVA decoding at the moment.

What's the best way to determine if content is natively interlaced vs telecined?
There are more variations than just natively interlaced and telecined. E.g. there's content which was telecined and then overlayed with natively interlaced content. Or there's field blended film content. This is a really complicated topic. What you could do is enable madVR's film mode and then check if playback quality is good. If it is, then you very likely have a telecined film source. If you see combing then you have one of the other variants, for which at the moment you should better use DXVA deinterlacing.

[truexfan81]

Nobody else reported this problem. Which GPU and OS are you using? What happens if you change the zoom factor a bit (e.g. NumPad 1 in MPC-HC)? Does the problem go away then or is it still there?

win7 64bit and nvidia gt430 gpu
yes NumPad 1 does make the problem go away

[madshi]

win7 64bit and nvidia gt430 gpu
yes NumPad 1 does make the problem go away
Then it seems to be a problem with the relatively new fixed zoom factor resamplers. Which movie resolution and target rectangle does the madVR OSD (Ctrl+J) show when the problem occurs? Are you totally sure you're really using v0.85.7? I'm asking because there was a problem with some older builds. So if you're still using an older build that may eventually explain the problem.

[Jong]

When you can playback a full long movie without frame drops/repeats.Surely that's the problem. Even 23.977 vs 23.976 (far more accurate than most GPUs out of the box) results in a drop every 17mins. To guarantee no drops in a movie you would need almost 10 times that accuracy, which is pretty hard even with custom timings. I know because I tried really hard about 5-6 years ago. I'd say it is close to impossible for most people! Now it's true most people, in most movies, might not notice one or two frames drops depending on when they happen (especially because of all the good work you have done in MadVR to eliminate repeated judder when a frame needs to be dropped), but if that is your standard something like Reclock (or Videoclock in JRiver MC) is just about essential.

[glc650]

When you can playback a full long movie without frame drops/repeats.But since the madVR counters are not that accurate, just how much deviation between refresh/frame rates can there be before having to resort to resampling audio with reclock to prevent frame drops/repeats?

Film mode uses madVR's own IVTC algorithm, which currently runs on the CPU, which means it isn't compatible to native DXVA decoding at the moment.
So what is the alternative here, implementing DXVA IVTC? Isn't madVR's own IVTC the best way to go (for quality) anyway?

[truexfan81]

Then it seems to be a problem with the relatively new fixed zoom factor resamplers. Which movie resolution and target rectangle does the madVR OSD (Ctrl+J) show when the problem occurs? Are you totally sure you're really using v0.85.7? I'm asking because there was a problem with some older builds. So if you're still using an older build that may eventually explain the problem.

movie resolution 640,360
target rectangle 0, 0, 1920, 1080

and madVR reports v0.85.7
tomorrow i can post a sample vid if needed