madVR - high quality video renderer (GPU assisted)

[Bandito]

Do we need to use uninstall.bat before upgrading to a newer version of madVR?

[ryrynz]

Originally Posted by Bandito

Do we need to use uninstall.bat before upgrading to a newer version of madVR?

No. If it's already installed all you need to do is copy it over the old files.

[BluesFanUK]

How does the 4.4.4 test image work?
"If you want to test whether your display supports RGB in 4:2:0, 4:2:2 or 4:4:4, you can use this test image. Make sure you display it with 1:1 pixel mapping, otherwise it won't work."

[huhn]

open the picture unscale it and look at the numbers you can see.

[madshi]

Originally Posted by hubblec4

I have posted in the LAV thread an issue.

With the new madVR0.89.17 play the DVD Aliens(test dvd without the big-vobs) with 4:3 format (menu and video), but it should be 16:9. With EVR it plays fine with 16:9 format.

Thanks, will have a look at this when I find some time.

Originally Posted by KoD

Hi madshi, you were right, the settings are reset, both the file and the registry entries get deleted, so that's not the reason why the subtitle positioning got fixed. In fact, after doing what you asked me, I can now say with certainty that the subtitle positioning issue happens only when DXVA is used for the scaling of the luma and of the chroma. If DXVA is not used for scaling, then the subtitles are at their proper location in the image. It's as if DXVA is used for scaling, then the subtitles are kept at their original position in the frame in absolute values, which is now somewhere in the middle-left part of the image. And yes, using software decoding + DXVA scaling also shows this issue; it's not related to QuickSync.

In order to reproduce this, could you send me 2 "settings.bin" files (from the madVR folder, or if madVR doesn't have write access to that folder, it's in "HKEY_CURRENT_USER\Software\madshi\madVR\Settings"), one where subtitle positioning works and one where it doesn't work - together with a small sample?

Originally Posted by KoD

I confirm there is an AR issue for DVD playback myself, with a DVD of my own. But this is not new, it's here since at least 0.89.15 (did not test further back).

The copyright notice is 4:3 and is displayed properly, the studio intro is also 4:3 and no issues, and then the menu appears, which is a static image, and should be displayed as 16:9, but madVR displays it as 4:3. In EVR it's 16:9. No sample from me, but I think there needs to be the two videos included as well to reproduce this, and I don't know how to cut the vobs with the intro videos in.

Sample?

Originally Posted by Shiandow

I think preventing this 'edge filling' entirely will require a different algorithm, you'll need to detect edges in a somewhat more intelligent manner. Although, if you don't mind doing an extra processing pass, I did manage to get some nice results if I combine a somewhat modified Bilateral scaler it with a softened SuperRes pass, that's still highly experimental though.

There are too many changes atm in various places in madVR. It's not a good time to add experimental algo tests on top of that. Maybe later. Or when you have something near "final".

Thanks!

Originally Posted by fredgido

Love the thin edges enhancement! On anime it makes the outlines look so sharp (I set it to maxium)! Shapen and crispen looks slightly better on very moderate amounts. On anime, 1080p BD encodes it improves more consistently a lot more than on 480p dvd rips from what I tested. Looks like it is more hard to detect edges on low resolution blurry anime. Enhance detail only makes encoding artifacts more visible.

I will use the enchantments and try to tweak when watching anime and try different quality and art animes for some time then post my experience.

Glad to hear you like it!

Originally Posted by djsolidsnake86

madshi have you found how fix the stuttering when open a file and go to fullscreen?
no one experienced this? i'm on windows 10, with 2 monitors cloned

You could try using extended instead of cloned. Cloned is difficult for madVR to handle properly. I'm not even sure if madVR is getting the scanline information from the proper monitor in that setup.

Originally Posted by Siso

Is it me or the latest madvr versions take screenshots in 16-235 instead of 0-255? EVR cp is ok.

That probably greatly depends on the exact settings you're using. There are a million different render paths in madVR, e.g. DXVA scaling on/off etc, downscaling vs. upscaling etc. It's possible that there's a problem in some situations, but I can only fix it if I know your exact settings (e.g. upload "settings.bin"). Also a screenshot of the debug OSD (Ctrl+J) would be useful.

Originally Posted by nlnl

That is another log. Thanks!
https://drive.google.com/file/d/0B7a...ew?usp=sharing

Will have to delay this once more. Sorry for taking so long.

Originally Posted by firewater

I like the fullscreen exclusive seekbar of madvr because it only appears in fullscreen mode, but I also dislike the purple color, the font, and background (ideally I would love to only have the time and the bar without background). My question is, any chance you could add an option in the future to modify this bar via images or something like that?

Maybe in some far away future. Atm adding missing features is much more important than doing some cosmetical eye candy tweaks.

Originally Posted by dansrfe

What exactly does enhance detail do?

It increases the contrast between brighter and darker pixels in image areas where there are no edges nearby. Doing so makes image detail stand out more, but it also makes noise, grain and artifacts more visible. So it's probably more useful for clean/good sources.

Originally Posted by Nullack

When using the current madVR on a 4K UHD display with high DPI say 300%, the madVR CTRL J renderer stats are too small to read.

True. Will be fixed in some future version, but not a priority atm.

Originally Posted by truexfan81

v0.89.17 according to the settings SS attached

already using Ordered dithering

edit even with latest driver, its the same, cannot use the new algos, gpu usage jumps from 45% up to 100% as soon as i activate any of them. Adaptive Sharpen does not have that problem.

Something is weird there. Which media player are you using? Is it possible that you have different madVR versions installed on your PC? E.g. J.River MC comes with its own madVR version.

Also make sure your GPU settings don't force anisotropic filtering or anti-aliasing. Doing that slows down your GPU without visual benefit in madVR.

Originally Posted by RyuzakiL

Does Madvr now, runs properly on Windows 10 TH2 + AMD GPU?

AFAIK generally yes, as long as you don't have CrossFire, but there's always the odd user who has problems, so there's no guarantee.

Originally Posted by retrue

I think I have found a bug. I am using the last development version of MPC-HC with the last version of XYSubFilter and the last version of MadVR
The bug happens in MadVR plus XYSubFilter when displaying upscaled VobSub subtitles.
It is an anime episode, 848x480 internal resolution, MPEG4 (H264) video, softcoded VobSub subtitles.
When displaying the vid at 1x resolution the subtitles are correctly displayed at the bottom of the screen.
When the vid is at 50% resolution (in window mode) it doesn't show subtitles at all.
If the vid is upscaled, resolution 2x (window mode) or it is upscaled fullscreen at 1920x1080 the subtitles appear on the left and in the middle of the screen and the subtitles themselves are not upscaled (they appear with the same size than in the resolution 848x480).
MadVR with other subtitles filters (VSFilter and Internal MPCHC filter) display correctly the subtitles upscaled and the bottom of the screen.
The MPCHC Personalized Enhanced Video Renderer with XYSubFilter shows the subtitles correctly too.
You can see screenshots of the vid here with the subtitles. In all these screenshots I am using MPCHC, MadVR and XYSubFilter.
If you need more info, please, tell me.
http://www72.zippyshare.com/v/8bgIjJfZ/file.html

Thanks, I'll have a look at this later, when I find some time.

Originally Posted by Francois76l

I have one demand/question:
Many of us are experiencing the same problem: i1d3 (or low end colorimeters) measure our display with a perfect black (accuracy problem) and the result is always a linear gamma from 0 to 100%. The reason is simple: perfect black = infinite contrast ratio but we don't have an oled tv!
As a consequence the blacks are crushed between 0 to approximately 2.5%.
The 17-18 blanking bars are always invisibles after 3dlut creation.

I tried to decrease the gamma @1.25% with videoequalizer v2.06. It works very good (17-18 became visibles again) but it's not compatible with a 3dlut and I can't use this solution.
Do you think it could be possible to create a tool in madvr to modify the gamma curve in realtime after the 3dlut?
The goal is not to make a new gamma curve but only to correct the grayscale with our eyes at very low ire because the i1d3 can't.
A 100 points gamma editor would be really amazing for us. Something like videoequalizer but in madvr, after the 3dlut...
The idea is just to lower the gamma between 0 to 5%. It would be so cool...

Maybe in a future version, but I've many other things to do first.

Originally Posted by clsid

@madshi
Maybe the settings tool could detect if certain driver functionality (e.g. OpenCL) is missing and notify the user when enabling any options that depend on that functionality?

That might be a good idea if there is only one GPU/monitor. On my PC I have 3 different GPUs installed at the same time, though, so such a check would be pretty useless here.

I'll revisit the whole settings dialog, and its user-friendliness (or lack of) before I reach v1, but atm it's not a priority for me.

Originally Posted by 70MM

Im about to start using madvr with Jriver. Im only needing it for up to 1080p (no higher scaling) and wonder what the best settings I should be using for 1080p. I have a Sony 4k projector yet I don't want to scale to 4k as Im about to sell the projector for a new JVC which is still only 1080p. Im only using the system for ripped BDs on MKV 1:1 copy and a number of video files. Im using the Nvidia GeForce GTX 970 with 4GB ram card.

The new JVCs seem to handle 4K input relatively well, so you *may* benefit from upscaling in madVR, even though the JVC panels are only 1080p. Might at least be worth a try. I'd wait until the JVC has arrived and then simply play with the various options to see which produces the best overall result.

BTW, in case you have a CIH/CIA/CIW setup, madVR can automate all that for you, including setting automatically activating lens memories for cinemascope movies etc...

Originally Posted by aufkrawall

I'm using D3D11 in windowed fullscreen on Windows 10 for a few days now and so far there hasn't been a problem with playback jittering/tearing.
Either something has been fixed in the current Nvidia drivers (using 359.00) or it was the big Windows 10 November update.

Good to hear!

Originally Posted by Georgel

I get random crashes (very often) when changing from normal to fullscreen, using right click and full screen, or alt+enter. The mpc-hc + madvr crashes, but windows does not crash. I wanted to ask, why this might happen, I will post a log file, if anyone can give me a hang on how to read the log file, or how to generate it.

Originally Posted by truexfan81

i get a crash 1 out of every 3 times i launch a video lol and its only when using madvr, funny thing is the crash reports always point to the nvidia driver happens with absolutely every version of the nvidia driver too.

I'll need the crash report. If you don't get a crash report file on your desktop automatically, click on the "show bug report" button, then press Ctrl+C, after that you have the bug report in your clipboard.

Originally Posted by Ver Greeneyes

madshi, could you add support for either arithmetic operators, or some more measures of throughput in the profile select rules? I ask because I'm pretty sure the amount of time it takes my GPU to apply a particular filter scales linearly with the amount of pixels it has to process. So if I know it can handle a particular profile for 720p30, I'd like to use that as an upper limit, so it uses the same profile up to 1280*720*30 = 27648000 pixels per second.

In a profile select rule, I imagine this would work like (srcWidth * srcHeight * deintFps <= 27648000), or (srcPixels * deintFps <= 27648000), or (srcThroughput <= 27648000), depending on what is available. Adding support for arithmetic operators (or at least multiplication) would generalize the rule selection nicely. In my case, I'd probably combine this with (srcWidth == targetWidth or srcHeight == targetHeight) to know when I only have to worry about chroma scaling.

Makes sense, I'll add it to my to do list.

Originally Posted by TheElix

What might be preventing Fullscreen Exclusive mode from activating on my computer? I have fresh-installed MPC-HC and madVR and it doesn't help.

Windows 10 x64
MPC-HC 1.7.10
madVR 0.89.17

Edit: Madshi, I uploaded log, bug report and DxDiag information in one archive: https://yadi.sk/d/8Y5cl6AAkyp5t Also, in this instance, the player also crashed. As it happens every single time now.

The crash is when doing error diffusion. You could try updating your GPU driver, maybe that helps?

In order to find the failure to enter FSE mode, I'll need another log with the debug OSD (Ctrl+J) active while recording the log, because having the debug OSD active adds some more information to the log that I need in this case.

Originally Posted by xabregas

So i have tested the new xysubfilter and new madvr and i was almost amazed when i saw that in 2.35 movies the subtitlees now can go to the black bars outside the movie window. But it was a matter of minutes until they went back to the front of the movie while playing. Anyone knows how to fix it, i.e making the subtitles stay in the black bars??

There is a known bug that I need to work on. It also requires a new XySubFilter patch, which is why I've not fixed it yet. Will be fixed sooner or later, but might take some time...

Originally Posted by adhara

Question, if I put "activate deinterlacing" off on MadVr, is the display reponsible for doing this stuff ?

No. An HTPC is not good with outputting interlaced images. You do need to deinterlace on your HTPC.

Originally Posted by Aktan

Hi All, I have a problem that hopefully someone can help me fix. I'm thinking I'm just missing something silly in some settings. I'm using MPC-HC Version 1.7.9 opening an analog capture device. I use ffdshow tryouts to change the output colorspace (to avoid the forced load of MPC-HC internal Deinterlacer filter on 4:2:2 color spaces) and to deinterlace the source. I use madVR Version 0.89.9 as the renderer. The problem I have is if I use ffdshow to double the frame rate for deinterlacing, madVR seems to drop all those extra frames. I see this in stats and it only happens if I double the frame rate. This problem doesn't exist with other renderers like EVR as it is a lot smoother when the frame rate is doubled. I didn't use madVR's own deinterlacing becuase it wasn't deinterlacing well. I'm guessing this is probably due to the fact that the queues are stuck at 0-1 in stats. I'm also guessing this is why madVR keeps dropping frames. The combination of ffdshow deinterlacing and madVR renderer does not have a problem if I use a MPEG2 source instead of an analog capture source probably since the queues can be filled properly then. Does anyone have an idea of what I can do to fix this?

madVR only drops frames if either the GPU is too slow or if the refresh rate it too low. Please make a screenshot of the Ctrl+J madVR menu when you get those frame drops, then maybe we can help.

Generally you should get better quality when letting madVR deinterlace. So it might make sense to try that again and post a screenshot of Ctrl+J with that setup, too.

[madshi]

Originally Posted by gmail123

my englishe is poor..
madVR v0.89.14---madVR v0.89.17 can not pass through the DVD .ifo file, the AR video stream couldnt be read，the scale of the images is not proportional；if the .vob files are opened directly, it can be recongnised accordingly. this problem would only occur if PAL DVD，NTSC DVD are running and functioning normally。

Can you upload a small sample for me with which I can reproduce this issue?

Originally Posted by 70MM

Does anyone know if you can do a non linier stretch with madvr?
Can you stretch a 16:9 image to scope 2.35 screen. Where the centre is correct shape but the sides slightly stretched to fill the scope ratio?

Do you *REALLY* want this?? I consider non-linear stretch very ugly. It looks ok as long as the camera doesn't move. But for any horizontal pans it looks totally psychadelic to my eyes, because the distortion becomes totally obvious in that situation.

I might add NLS in some far away future version, but it's very low priority because I consider NLS to be bad.

Originally Posted by MysteryX

Lately I'm getting regular crashes with madVR. Attached is a crash report.

Don't ever attached anything to this forum, takes ages to get it approved, and sometimes it never gets approved at all. Upload somewhere else, please.

Originally Posted by Uoppi

1) Progressive segmented frame, i.e. 25 progressive frames divided/segmented into 50 fields. Meaning the video ought to be weaved into 25p. Does simply "deint=Off" do it? The auto deinterlacing seems to erroneously detect the material as interlaced, as per the container tag.

Most of these are flagged correctly, so most of the time you can either force the deinterlacing completely off, or you can force madVR into film mode, if you want to be extra safe. DXVA deinterlacing (which is the default deinterlacing mode) will usually handle these "correctly", too, meaning you should get proper image quality, but DXVA deint outputs 50 frames instead of 25, so you'll have double the GPU load.

Originally Posted by Uoppi

2) Progressive film stored on DVD as interlaced. Simply "deint=Off"?

Same as above for PAL DVDs. NTSC DVDs are quite different, though. Those often produce problems when playing with deinterlacing turned off. Usually forced film mode works best there. There are a couple rare field blended NTSC DVDs out there which require DXVA video mode deinterlacing, but that's really rare.

Originally Posted by Uoppi

3) Pal speedup removal of 25p content or film stored as interlaced on DVD (using Reclock). I'm guessing simply "frameRate=24" or "frameRate=23.976" (which one?)

I'm not an expert on how this is done using Reclock so I can't really answer this one.

Originally Posted by Uoppi

if I use deint=Film, I get several ms worth of performance hit and the OSD says "IVTC", even though, as far as I know, only the repeat flags of the DVD are being ignored. For example in Handbrake, it's "safe" to simply leave deinterlacing off and set fps to 25 when ripping progressive material DVDs. So I'm just struggling to grasp the real world difference between the tags I mentioned

For most PAL DVDs turning deinterlacing off should be fine, as mentioned above. There are a few odd DVDs out there which have wrong telecine flags. For those you need to use madVR's forced film mode to get proper progressive output.

Originally Posted by Uoppi

2) In the case of progressive segmented frame material, two consecutive fields should be merged/weaved (whatever the correct term is) into a progressive frame. Again, when using "deint=Film", the OSD reads "IVTC" and 2:2 cadence - does that mean madVR is properly merging the fields? I would have thought the term "IVTC" isn't applicable in this case. At a quick glance at least, the playback appears identical whether I use deint=Film or frameRate=25 (the content within the container is progressive 25 fps).

My "definition" of IVTC is that madVR converts telecined content (progressive segmented frame material is some sort of telecined content, too) back into progressive. "2:2" means that always 2 consecutive fields are weaved together. *However*, these doesn't have to be top+bottom of the same encoded frame.

Originally Posted by Uoppi

- Using madVR's auto detection on any of its settings, progressive segmented frame videos (from a Canon camcorder) are detected as interlaced and are deinterlaced instead of merging the consecutive fields into progressive frames
- Content I've verified as progressive (going frame by frame) sometimes still gets deinterlaced with madVR's auto detection, regardless of the detection setting

madVR doesn't really have a good auto-detection yet. Auto-detection simply means DXVA deinterlacing is used to do the job. DXVA deint internally auto-detects whether the content is video or film and then applies the proper algorithms. However, DXVA deint always outputs double frame rate. So it's all not ideal yet. Tagging is of course a great solution, if you're willing to invest the time.

Originally Posted by Uoppi

So, summa summarum, from what I've gathered here, it doesn't hurt anything to always use deint=Film for all progressive PAL content stored in an interlaced-flagged container. I'll be using that then instead of deint=Off.

It shouldn't hurt (other than raising CPU load a bit), but if the content is cleanly stored/encoded, it also doesn't help. It does have the benefit of auto-detecting those rare PAL DVDs which have wrong flags.

Originally Posted by Uoppi

How "intelligent" is the processing? Let's say, for example, if I have by accident placed a Blu-Ray rip in a folder that has the deint=Film tag, will madVR "know" what to do with the picture, i.e. to leave the source untouched (because it's native progressive)?

The quality/reliability of madVR's IVTC algorithm is pretty good. It's extremely unlikely that running it on a progressive Blu-Ray rip will do any harm. It should simply detect that the video is a 2:2 cadence and not do anything other than analyzing the frames and detecting/confirming the 2:2 cadence.

It's impossible for me to know for sure whether a source is progressive or not. For all I know, the video stream can claim it's progressive but it would still be telecined. Because of that, if forced film mode is activated, IVTC is always performed, even if the source appears to be progressive. But as mentioned above, due to the very high reliability of the madVR IVTC algorithm it's not a problem if you activate it accidently - other than raising CPU load a bit.

Originally Posted by Georgel

I managed to get a crash report from a crash mpc-hc+madvr that has happened. I thank in advance if someone could explain to me what is the problem.

I uploaded the .txt file to my dropbox account.

https://www.dropbox.com/s/0fxakdk93w...eport.txt?dl=0

Another thing, I always send my crash reports to mpc-hc automatically, but the crashes only happen when using mpc-hc with madvr, so I really hope that someone can offer me a helping hand. Thanks!

The crash occurred when trying to create a D3D11 device. You could updating your GPU driver. Maybe that helps? If not, disabling error diffusion and/or NNEDI3 might be necessary to get rid of the crash.

Originally Posted by mogli

madVR v0.89.17

[...]
* fixed: repeated frames were reported although smooth motion FRC was on

I checked thoroughly and no, all it did was reporting actual repeated frames which are just omitted from the OSD now. (But from my current understanding this is to be expected with FRC, if frame rate and display rate are essentially identical.)

Hmmmm... Just to clarify: How do you define "repeated frames"? E.g. when playing a 1fps movie on a 60Hz display, madVR is repeating frames all the time, even with smooth motion turned on. Only about 1-2 of those 60 frames will be blended/unique. This is the way smooth motion frc works, so listing such repeated frames in the Ctrl+J menu doesn't make much sense, IMHO.

Or are you saying that every time the old version reported a repeated frame, you got a visual stutter? That would be something completely different, of course!

Originally Posted by garson

I'm having strange issue I guess with MadVR.
I have latest MPC-HC 1.7.10, latest MadVR, Lav filders, running on XP. I have a lot of dropped frames on dark/black scenes with not much going on.
So, for example I'm watching Hobbit 720p (8GB mkv file) and everything is pretty much smooth but when movie finishes and Cast of characters scene starts there is a lot of stuttering and dropped frames.
Also, I watched movie Circle (2015) which takes place in dark room with several persons in it, not much going on there, but I got a lot of dropped frames again, movie is barely watchable.

Does anybody now what can be the issue with dark (and slow) scenes?

I can think of 2 possible reasons:

1) Do you have a 3dlut active? If so, some old GPUs may run into problems if certain areas of the 3dlut are accessed. I've had that reported by one user some years ago. But it was a *really* old GPU.

2) Maybe it's some sort of deinterlacing issue? Are you using forced film mode? Film credits are sometimes video mode and require proper deinterlacing.

I suppose with EVR this problem does not occur? Can you post a screenshot of the Ctrl+J debug menu in the moment when those frame drops occur?

Originally Posted by GCRaistlin

madshi
Last example in http://forum.doom9.org/showthread.ph...17#post1271417 contains "srcAR" value name while there's no such value name in syntax description above.

Fixed, thanks.

Originally Posted by m-v-w

Could Madvr support MPVfuture?

Not sure what you mean with "support"? Are you asking whether a future mpv version might allow to play videos by using madVR? If so, I think that's very unlikely because mpv doesn't use DirectShow and is multi-platform. So the whole architecture is very different from what madVR is built for.

Or are you asking whether I'd be willing to help mpv future development? If that's your question then my answer is a simple no. I don't have the time to help other projects. Furthermore madVR is supported by a lot of different media players. I can't help one of them but not the others. That wouldn't go down well with the other media player devs/companies.

Originally Posted by BluesFanUK

Hey madshi, is there any plans to create a dedicated MadVR Video Player?

Not really. One reason is that I simply don't have the time/resources. Another reason is that doing so would annoy all those media player devs/companies that support madVR. They don't want competition from me.

Originally Posted by THEAST

I spent the past 3-4 hours troubleshooting this issue; I am positive this is an incompatibility between madVR and AMD's crossfire; I don't think using the 32-bit version of madVR/MPC-HC will fix this. Anyway, here is the results of my findings:

- MPC-HC will crash only if I pause it, switch my active window to another application, work for a few minutes, then go back to MPC-HC, go full screen (exclusive or windowed) and resume the video. If I go back to MPC-HC quickly, it doesn't crash.

- If I disable crossfire globally, it will not crash anymore.

- If I enable crossfire globally but only for applications with a profile, it will still crash. Even if I make a profile for mpc-hc64.exe and madHcCtrl.exe and disable crossfire for both, it will still crash. Based on GPU clock and utilization, crossfire is NOT disabled regardless of what I do and both GPUs are used anyway when playing videos.

- If I disable "use Direct3D 11 for presentation" and "use a separate device for presentation", video playback will hang instead of crashing; i.e. MPC works fine but refuses to resume video playback. I can use the same MPC window to re-open the file just fine, in which case it will play properly.

- If I resume playback before going full screen, MPC will not crash but I will get a black screen while audio plays fine. Both GPUs stay at idle clocks in this case.

Needless to say, I can easily blame AMD for this problem but I highly doubt they would even care or try to resolve this issue, no matter how many times I report it.

Any feedback from other people with Crossfire setups is highly appreciated.

P.S. Disabling crossfire in games seems to work correctly. I am going to report this to AMD anyway but I won't expect them to fix it any time soon, if at all.

Yeah, Crossfire has repeatedly being reported as causing problems with madVR, unfortunately.

Originally Posted by RyuzakiL

Win 7 + Madvr = Crossfire always on but stable with no dropped frams

Win 8 + Madvr = Crossfire disabled but you have to use SXBR with no dropped frames (Best Combination)

Win 10 + Madvr = Crossfire always on Unstable, Madvr always crashes (Piece of Crap)

this was running on 2X HD7850 with Latest AMD Crimson Driver (Beta Hotfix.) All are on tested on a clean installed machine.

I suggest you stay away from Windows 10+AMD Crimson Driver combination till they sort this shit out.

Now going back to Windows 8.

Ouch, very sad to hear that.

Originally Posted by lanzorg

Madshi could you add supersampling or a way to use image doubling for any videos?

Maybe at some point in the future. For now you can simply upscale a tiny bit (the smallest amount your media player supports, like 0.5% or so).

Originally Posted by har3inger

Whenever you set up an image upscale that results in a final step of upscale in one dimension but downscale simultaneously in another dimension, lanczos 3 is used for the upscale. Is this because Jinc cannot be used for 1 dimensional upscaling?

Exactly.

It's not a limitation of the Jinc algorithm itself, it's just that I haven't implemented Jinc downscaling in madVR (yet?).

Originally Posted by XMonarchY

Will DirectX 12 allow the use of multi-GPU in madVR?

No.

[madshi]

madVR v0.89.18 released

http://madshi.net/madVR.zip

* added tone mapping support for HDR content
* added gamut mapping support for HDR BT.2020/DCI-P3 content
* added support for SMPTE 2084 transfer function decoding
* added support for receiving HDR metadata from LAV Video Decoder
* added "maximum display luminance" option to display properties page
* added "HDR" profile variable
* SuperRes now supports being run after every ~2x upscaling step (again)
* improved JVC projector ip control connection reliability

The big keyword for this build is HDR. I suppose many users might not know how it works exactly, so some explanations:

1) First of all, this madVR build does *not* aim to send HDR content untouched to the display, to let a "HDR compatible" display do all the processing. Doing that might be supported by some future version (as an option). But for now what madVR does is convert the HDR content in such a way that it looks well on *any* display (new and old), regardless of whether the display officially supports HDR or not.

2) The key idea of HDR is to increase the peak white luminance, so that we can get brighter highlights, and more texture detail in bright image areas. HDR content uses 10bit (or more) instead of the usual 8bit, and it also uses an improved transfer (gamma) function, which means we also get better shadow detail at the same time.

3) HDR content is encoded using a totally different transfer ("gamma") curve. So playing HDR content with a media player and display which don't understand the new transfer function won't look good/correct. It will look totally washed out.

4) The new transfer function directly maps each pixel to a specific desired luminance value (in "nits" or cd/m²). Possible values are 0 nits up to 10,000 nits. However, videos don't have to use the full 0 - 10,000 range, they often top out at either 4,000 nits or 1,200 nits (from what I've seen).

5) UHD Blu-Ray will usually be encoded in BT.2020 color space, which is extremely wide. However, videos don't have to use the full BT.2020 space, they usually only use DCI-P3, inside of the encoded BT.2020 container.

6) Today's displays (even the best of the best) cannot really properly reproduce HDR content. E.g. many display's don't reach full DCI-P3 colors yet (let alone BT.2020), and no available display today comes even close to be able to output 10,000 nits! Which means that someone somewhere has to convert the HDR content down to something the display can handle. The consumer electronics world plans that the display will do this conversion (only new models, obviously) - and every TV manufacturer will write their own conversion algorithms because there's no standard for that! From what I've been told those conversion algorithms (at least for the first few HDR display generations) are probably not going to be very high quality.

7) madVR is able to compress both the luminance and the colors down to what your display can handle. It uses reasonably high quality algorithms for that. I might find even better algorithms in the future, but for now these algorithms should be a good starting point.

8) UHD Blu-Rays will come with an updated copy protection. So madVR will not be able to play them - unless the copy protection gets broken at some point.

If you want to test HDR playback, please update to a new LAV nightly build because only the latest LAV nightlies are able to pass some HDR metadata information to madVR!

(P.S: There's also a new "madTestPatternSource" version available with some HDR test patterns. Link see first post in this thread.)

[nevcairiel]

Are there any smart ideas how to properly configure a display for HDR? I could set it to its maximum brightness and measure this, and tell madVR about it.
However, the maximum brightness is far above what I find "pleasing" for SDR movies, or even the desktop/player UI, so its really not "good" to configure it to that.

Of course I could tell madVR about the brightness of my average setup, but then there is nothing to gain from HDR at all.

[THEAST]

Originally Posted by madshi

Yeah, Crossfire has repeatedly being reported as causing problems with madVR, unfortunately.

So, there is no way to prevent the crashes without disabling Crossfire globally? I'm not trying to accelerate madVR's computation with crossfire, mind you, but rather, I am trying to find out why the crashes happen even though crossfire is not enabled for MPC-HC/madVR and is only enabled for applications that specifically have a profile.

[mogli]

Originally Posted by madshi

Or are you saying that every time the old version reported a repeated frame, you got a visual stutter? That would be something completely different, of course!

Yes, before the current version repeated frames were reported and visual stutter occured at that very moment. Now there's only the latter.

Note that this only seems to happen when fps is almost the same as the display rate. It took me some time to remember that the interlaced NTSC DVDs (actual video, not film) will have there frames doubled by DXVA deinterlacing, therefore running 59 fps at 60 Hz. This seems to be hard for FRC. (No way around starting to use ReClock, it seems?)

Originally Posted by madshi

Only about 1-2 of those 60 frames will be blended/unique. This is the way smooth motion frc works, so listing such repeated frames in the ctrl+j menu doesn't make much sense, imho.

madVR never did report such repeated frames, at least not for me. When viewing a 24 fps movie at 60 Hz with FRC on I never got even one repeated frame reported in any version. Or do you want to say, this was by accident, because each and every frame was unique or blended, so that there never was any actual repeated frame?

[Bandito]

Originally Posted by ryrynz

No. If it's already installed all you need to do is copy it over the old files.

k thx

[madshi]

Originally Posted by nevcairiel

Are there any smart ideas how to properly configure a display for HDR? I could set it to its maximum brightness and measure this, and tell madVR about it.
However, the maximum brightness is far above what I find "pleasing" for SDR movies, or even the desktop/player UI, so its really not "good" to configure it to that.

Of course I could tell madVR about the brightness of my average setup, but then there is nothing to gain from HDR at all.

That's a pretty good question. I suppose once HDR content is around, many users will have the same problem. To be honest, I've no simple solution available right now. I suppose one option would be to define a peak nits level for SDR content. In that case madVR would simply render SDR videos darker. The disadvantage of that solution would be that a part of the 16-235 data range would go unused, which would be a shame. But then, I suppose dithering is good enough to handle that.

Another option would be to use different presets in the TV, but only some TVs support that, and the lack of auto switching could be quickly become annoying.

Edit: My idea above would only work for video playback, obviously, it would not do anything for desktop or player GUI. If you need a solution for that, too, maybe one option would be to misuse the GPU gamma ramps to make everything darker? That should work for desktop and media player GUI. However, GPU gamma ramps don't use dithering, I think, so it would not be a good solution for image quality.

Originally Posted by THEAST

So, there is no way to prevent the crashes without disabling Crossfire globally? I'm not trying to accelerate madVR's computation with crossfire, mind you, but rather, I am trying to find out why the crashes happen even though crossfire is not enabled for MPC-HC/madVR and is only enabled for applications that specifically have a profile.

That's a question you'd have to ask AMD. I don't think there's anything I can do.

Originally Posted by mogli

Yes, before the current version repeated frames were reported and visual stutter occured at that very moment. Now there's only the latter.

Note that this only seems to happen when fps is almost the same as the display rate. It took me some time to remember that the interlaced NTSC DVDs (actual video, not film) will have there frames doubled by DXVA deinterlacing, therefore running 59 fps at 60 Hz. This seems to be hard for FRC. (No way around starting to use ReClock, it seems?)

madVR never did report such repeated frames, at least not for me. When viewing a 24 fps movie at 60 Hz with FRC on I never got even one repeated frame reported in any version. Or do you want to say, this was by accident, because each and every frame was unique or blended, so that there never was any actual repeated frame?

Ok, so I misunderstood your original report. I didn't think there was visual stutter. So I suppose I should revert the OSD change.

What fill state do the queues have when that stuttering occurs? How many frames are you presenting in advance? Or did you disable the "present several frames in advance" option? FWIW, it's recommended to use that option because it increases presentation reliability.

[huhn]

Originally Posted by madshi

That's a pretty good question. I suppose once HDR content is around, many users will have the same problem. To be honest, I've no simple solution available right now. I suppose one option would be to define a peak nits level for SDR content. In that case madVR would simply render SDR videos darker. The disadvantage of that solution would be that a part of the 16-235 data range would go unused, which would be a shame. But then, I suppose dithering is good enough to handle that.

double HDMI connection should be possible in the near future.

switching the HDMI output on a TV doesn't sound too terrible.

and DP 1.3 to HDMI 2.0 should work with an passive adapter.
you are simply way ahead of time with that feature.

but i don't think leaving the TV at a high brightness for SDR content is a good idea. none OLED screens will lose a lot of CR this way. and local dimming has a enough of problems already without this high brightness.

[Thunderbolt8]

Originally Posted by madshi

3) HDR content is encoded using a totally different transfer ("gamma") curve. So playing HDR content with a media player and display which don't understand the new transfer function won't look good/correct. It will look totally washed out.

So does this mean even when using madVR for watching an UHD Blu-ray on my current plasma TV (VT60) that it will look bad/not right?

[nevcairiel]

Originally Posted by Thunderbolt8

So does this mean even when using madVR for watching an UHD Blu-ray on my current plasma TV (VT60) that it will look bad/not right?

No, madVR will correct that.
Note however that not all UHD Blu-rays will use HDR, in fact most will not. There may even be two releases of movies, one in HDR and one SDR.

[fairchild]

Thanks for new version Madshi, I have a quick question though. How do we know what to set the display peak luminance to? I have an older 1080p 32" EX400 Sony LCD and a 1080p 55" VT60 Panasonic Plasma. Both by default are set to 600 nits. Not sure how or where to find the correct setting. I know one thing is for certain, the plasma is a much higher end display which may affect this setting?

[nevcairiel]

Originally Posted by fairchild

Thanks for new version Madshi, I have a quick question though. How do we know what to set the display peak luminance to?

Sometimes its in the documentation, otherwise you have to measure it.
Note that documented peak luminance will always refer to 100% backlight setting, which you may not find good for general watching purposes, like myself.

[James Freeman]

Originally Posted by madshi

4) The new transfer function directly maps each pixel to a specific desired luminance value (in "nits" or cd/m²). Possible values are 0 nits up to 10,000 nits. However, videos don't have to use the full 0 - 10,000 range, they often top out at either 4,000 nits or 1,200 nits (from what I've seen).

First, Thank you for this pioneer build!

Questions:
How can you really know in what range the studio mastered their movie? Seen where?

Judging by what I see with madTestPattern, how did you know what region is more important to compress and what to clip?
With 1200nit mastered content (presumably), the whole range is compressed.
With 4000nit, again practically the whole range is compressed.
Doesn't that mean that the movie will be too dark overall just so that we will be able to see the super bright parts?

[TheElix]

Originally Posted by madshi

The crash is when doing error diffusion. You could try updating your GPU driver, maybe that helps?

In order to find the failure to enter FSE mode, I'll need another log with the debug OSD (Ctrl+J) active while recording the log, because having the debug OSD active adds some more information to the log that I need in this case.

Madshi, thank you for your patience and for finding time to look into my problem. I have switched to Random Dithering and it seems to fix crashes. But at around the same time when it crashed previously now the video hangs up. It appears like the video goes in an endless loop of several frames while the audio keeps playing.

Here's the log with the debug OSD: https://yadi.sk/d/QDP2CRqHmDvVi Maybe you can tell why it doesn't go into FSE?

I've downloaded several test videos and tried the new HDR feature. I am currently using a professional NEC P241W monitor which is true 10 bit through DisplayPort cable and is capable of >Adobe RGB coverage and has a preset for DCI color space. I've set the maximum brightness which is 350 cd/m2 for this display and set the white point to 6500K. I used 400 cd/m2 in madVR settings. Well, it works. But I have horrible amount of dropped frames even though rendering is well under 10 ms. It looks like I cannot truly appreciate the difference now since I'm not getting 10 bit without FSE.

Also, you said that we may choose the maximum luminance between 0-10000 cd/m2, but the lowest I can choose now is 400 cd/m2. Is it intentional? Can we choose luminance in increments other than 100-200?

[James Freeman]

I think the new "display peak luminance" feature only works if your display has a static and unchangable minimum light level (MLL) of the display.
Meaning, it only truly works if you have a FALD (full-array local dimming) display.
If your display can do 4,000nits but keeps the contrast ratio at 1000:1 it simply does not differ from any other SDR TV.

OR, I might be very wrong!
..and madVR actually compress the HDR content to 1000:1 and you should turn your SDR display brightness to max (400nit) to utilize the new feature...?

I need some HDR demo clips, anyone?