madVR - high quality video renderer (GPU assisted)

[Uoppi]

this shows the color boosting from reconstruction and the ringing i reported some weeks ago.

And this occurred with Reconstruction sharp, not soft, if I recall correctly?

[nevcairiel]

I have a question. I see that there is an option to move the subtitles. Would it be possible to allow the user to change the color and the font too? I know it's easy with .srt files but I can't find a way to do it with DVD and Blu-Ray subtitles. I found that your option to move the subtitles works with Blu-Ray so I immediately thought to ask you if it would be possible.

DVD and Blu-ray subtitles are bitmaps (images), to change their font you would need to OCR and re-render them, nothing I would ever expect to happen.

[l0rdraiden]

Are you, by any chance, running your screen at 120Hz or something? If yes try to use 60 Hz and see if that works...

MadVR never used to play nice with 120Hz in my case.

Greetz, Unr3aL67

Thanks a lot, as you pointed out at 60 Hz works fine, my monitor was at 144 Hz.

Is this a bug or known issue of Madvr? is there any plan or any way to fix it?

@madshi ?

[DragonQ]

DVD and Blu-ray subtitles are bitmaps (images), to change their font you would need to OCR and re-render them, nothing I would ever expect to happen.
Seriously? Why would you ever use bitmaps for such a thing?

[leandronb]

I have the same problem if I use MPC-BE with internal filters.
But if I use MPC-BE with LAV filters the problem goes away.

Just tested and it's the same problem also with the latest madVR version.

Updated madvr and have the same problems too. I use LAV filters+madvr+ac3filter with MPC.
Black screen after skiping a video file to another in the same folder.

[nevcairiel]

Seriously? Why would you ever use bitmaps for such a thing?

Because font rendering is an annoying topic, and it greatly simplifies the required playback capabilities to just show images. Blu-ray subtitles can even include small images for commentary tracks or interactive elements.
While you can't do some of the transformations that you could with text subtitles, bitmap subtitles instead generally just work out of the box without having to worry about having the proper font, supporting the language and encoding the text is in, etc etc.

Blu-rays can actually include text subtitles, but its an extremely rarely used feature. I don't think I have a single disc that uses it, out of hundreds.

[XMonarchY]

After 345 posts, I have a hard time believing any of your questions are serious? Do you really want an answer or are you trying to flood this forum with useless posts?

You must know what NNEDI3 is -- image doubling, not image upscaling. And of course not all of the downscaling options are Bicubic; they are no different than the image upscaling algorithms. Many probably use very similar approaches to image scaling, with similar math -- but they are not the same.

Yet I have it in Chroma Upscaling... So Chroma Upscaling is similar to Image Doubling? You're confusing me further. I visited this thread enough times to understand the most important aspect - the best image quality settings with flawless performance (no dropped/skipped/delay frames and/or no presentation glitches) on my rig and, generally speaking, how these settings compare against lower quality settings. Dithering, SuperRes, Sharpening, 3DLUT, color depth, 0-255/16-235 range, de-blocking/artifact reduction, smooth motion, Image Downscaling and Image Doubling/Quadrupling.

I know what Chroma is, what Luma is, what SuperRes does, but full understanding of Chroma Upscaling and Image Upscaling and the differences between them, as well as, differences between different types of processing for those functions is not yet present in my brain. I know what processing looks best (IMHO) and runs great on my rig.

Aside from my day-time job search, I get paid building, setting up, and optimizing Home-Theater Gaming PC's (which obviously includes settings up madVR!), as well as, calibrating displays/projectors for those interested in having a decent Home Theater with the best image quality their Home-Theater Gaming PC's can pull. The completely understanding of Chroma Upscaling and Image Upscaling is not necessary to perform the work I described above...

Do you really think most people visiting this thread have complete understanding of Chroma Upscaling and Image Upscaling and why there's NNEDI3 available for Chroma Upscaling, but not Image Upscaling, even though NNEDI3 is available for Image Doubling/Quadrupling??? I am sorry if my ignorance aggravated your mood....

I also asked about whether different processing names were equivalent to some form of Bicubic because some members reported little-to-no difference between SSIM and Bicubic 125/150 (whatever value). Thus, it possibly implies other processing types could also be equivalent to some Bicubic value.

[Ver Greeneyes]

Yet I have it in Chroma Upscaling... So Chroma Upscaling is similar to Image Doubling?Yes, it could just as well be called Chroma Doubling. It's only used for videos where the chroma channels are stored at a lower resolution than the luma channel (most videos, but not all, are stored with the chroma channels half as wide and half as tall as the luma channel). The chroma channel is doubled first (using data from the higher resolution luma channel in the case of Bilateral and Reconstruction), and then the other upscaling and downscaling options operate in (linear) RGB.

[huhn]

And this occurred with Reconstruction sharp, not soft, if I recall correctly?

yeah that should be the case not sure anymore.

How does superXBR75 compare to reconstruction?

it's not nearly as sharp but it doesn't really mess up.

[huhn]

Hmm, something's wrong with DXVA Upscale colors

DXVA vs Spline
http://screenshotcomparison.com/comparison/162815


DXVA vs Jinc
http://screenshotcomparison.com/comparison/162814


It seems to only affect upscale, not DXVA downscale.
Anyone else who experienced this?


Using madvr 0.90.10
Nvidia 361.91

i guess checking your GPU settings is a start.

DXVA scaling can be affected by the GPU driver settings. which is a real problem.

[JarrettH]

If I had to visually judge between jinc, sxbr75, and reconstruction soft...

Reconstruction soft: softest of three, least artifacts

Jinc AR: middle sharpness, most aliasing

sxbr75: sharpest, low aliasing, most artifacts from ringing

Madshi said reconstruction sharp isn't something to rely on and soft is more performance optimized.

[James Freeman]

James can you share your videos? Thanks for testing it BTW, I wouldn't have considered it useful for chroma if you hadn't mentioned it.

I can not share, but I can tell you the name of the greatest movie for all chroma upscaling tests and share some pictures from it (under fair use). :)
The movie called Samsara (2011), it's one of the most beautiful Bluray discs every picture quality fan should own.

This movie is my "go to" for anything madVR testing.

Cheers

http://www.mediafire.com/convkey/d88e/jq4t239otcrv1cuzg.jpg
http://www.mediafire.com/convkey/0872/n75ipbgmbdapxqtzg.jpg
http://www.mediafire.com/convkey/b05e/6z7cmc26xw2qd2czg.jpg
http://www.mediafire.com/convkey/6cd2/fjs1u67j4u64c44zg.jpg

[huhn]

pls change the pictures to links. the image host is way to slow anyway.

[madshi]

Madshi, anything useful in the ChromaReconstructor 3.0 update that can be applied to madVR?
Also, any chance of other options improving soft? Placebo soft? Soft AR? The sharp options seem pretty unusable as far as accuracy goes.
Would like something that's a mix of what Jinc AR and Reconstruction gives me essentially.
There is one new algo in 3.0 which might be useful, but probably rather soft. Maybe it would be an alternative to the current soft variant, I don't know. At this point I have too many other things to do to look into it. Chroma upscaling is not that important compared to some other things.

Will there be an option to customize the color of the seekbar in FSE mode when we arrive madvr 1.0? Are there plans to support chaptermarks? That would be really cool, so one vote from me! :-D
No plans right now. Maybe some time in the future.

Yes, this time the memleak is gone completely. Big thanks for all your efforts, madshi!
Good to hear, thanks!

NNEDI3 won't ever work with newer nvidia drivers with CUVID ?
I don't know. Ask NVidia. nevcairiel and I have reported the problem to them.

I have the same problem if I use MPC-BE with internal filters.
But if I use MPC-BE with LAV filters the problem goes away.
Might be a bug in the internal filters, or a bug in madVR. Can't say much more with the limited information I have. I'd recommend using LAV, in any case.

madshi,
MPC-HC menu > View > Statistics shows the wrong framerate when using madVR. With all other renderers it displays the playback framerate of the video (video framerate * playbackspeed). In case of madVR it shows the screen refresh rate.

According to U78 the info is gathered using "get_AvgFrameRate()" method provided by the "IQualProp" interface of the filter graph.
Actually it shows the actually rendered frame rate on other renderers. I think madVR does the same thing, except that in some of the rendering modes its going to present a new frame for every VSYNC, hence showing screen refresh.

So I wouldn't necessarily call it wrong. If you use a mode that only presents every frame once, ie. D3D9 or Overlay, it shows what you might expect.
I use D3D9. Disabling "present several frames in advance" indeed gives the framerate value I expected.
@clsid, please read the section "frame rates" here:

https://msdn.microsoft.com/en-us/library/windows/desktop/dd407325(v=vs.85).aspx

It says that IQualProp described the "actual frame rate achieved during playback". What that means exactly is somewhat open for interpretation. From my point of view it's the number of frames per second that madVR is sending to the GPU for presentation. Some of those are duplicates, though, depending on the presentation mode. I suppose I could exclude those duplicates from the IQualProp stats. But to be honest, I don't find it very important, and as nevcairiel said, I don't really think what I'm doing right now is "wrong".

Is there a practical reason why you need this? Or is it just a "cosmetical" problem?

madshi, will you be extending the GPU queue to maybe 32 or 64 eventually? Same question for frames to present in advance.

I've noticed higher values produce smoother motion with smooth motion turned on. I'm guessing it has to do with how many frames are blended together.
Present in advance can't be increased, thats a limit of the Direct3D API.
And personally I seriously doubt that increasing the queues increases the smoothness, as long as the value is high enough for the algorithm to work. It won't blend infinitely many frames, just the exactly right amount.
^ Agree with nevcairiel here. Higher queue sizes usually help against frame drops and presentation glitches, but that's the only real use for them. Do you get frame drops or presentation glitches? If not then I don't think higher queues would help in any way.

Unfortunately their OLEDs cost an arm and a leg.
True. But I fear in 2-3 years everything LCD/LED will be yesterday's tech. Also, if you're interested in HDR support, even 2016 models are still considered early adopter (meaning: throw away next year) models. Personally, I'll wait for 2017 or 2018 models. Hopefully OLED might be cheaper by then.

I have the same issue in PotPlayer and Windows 10. It doesn't make a difference if I use the latest LAV nightlies or PotPlayer's internal decoders. The only way I've been able to work around this issue is to uncheck D3D11 in madvr. Has anyone found a way to avoid the issue while keeping D3D11?
Not sure why that happens. Works fine for me. Have you tried different GPU drivers, maybe? Windows 10 is still somewhat problematic for HTPC use, unfortunately... :(

Madshi there is a way to reduce the perfomance hit on reconstrution placebo and sharp?
I don't think so, at least not without changing image quality. I've already written compute shaders to get the best performance possible. These algos are by design VERY power hungry.

I painted a test pattern with thin and thick lines of blue and red circles in 4:4:4.
Then video captured the screen with CamStudio (screen recording software).
Then compressed the RGB (4:4:4) video through x264 to YCbCr 4:2:0 lossless.

I simply compared the original 4:4:4 video with the 4:2:0 video and chose a chrome upscaler which looks closest to the original 4:4:4 video.

To my eyes Super-XBR 150 almost recreates the thin chroma data without touching much the thick circles.
Be very careful with such tests, though, because the chroma position in a properly encoded video file must be offset in one direction by 0.5 pixels, but not both. Often with manually created test RGB images you end up with a video file which has the chroma channel in the wrong position, which would pretty much render all tests with that video invalid.

I've noticed that when anti-ringing is enabled with super-xbr and to a lesser degree Jinc (both with AR enabled and not along with some other scalers) can bring green tinges and possibly other colors if the source white contains some inside, which isn't ideal. This is only noticeable upon close inspection however.

The below screenshot comparison shows how bad this can be but you'd be hard pressed to notice it while watching.

super-xbr 150 vs super-xbr 150 AR (http://screenshotcomparison.com/comparison/162790)
Yes, in this specific test image super-xbr 150 without AR looks better. But there are other tests images where it's the opposite, e.g. see aufkrawall's favorite Anime test video.

Hmm, something's wrong with DXVA Upscale colors

DXVA vs Spline
http://screenshotcomparison.com/comparison/162815

DXVA vs Jinc
http://screenshotcomparison.com/comparison/162814

It seems to only affect upscale, not DXVA downscale.
Anyone else who experienced this?

Using madvr 0.90.10
Nvidia 361.91
Hmmmm... Could be a bug in madVR, or the NVidia driver doing weird stuff. In any case, the NVidia DXVA scaling implementation wasn't very good, last time I checked. Is your GPU so slow that you have to use it?

I spent a long time playing with your upscaling refinement options but I can't find a correct setting for every contents. I admit that I'm not a fan of sharpening filters, it always sharpens the flaws in the picture (even if your edge enhancement is the best I've seen!) and the only one I can use is the Reality Creation from Sony. The thin edges option is quite good too on DVD anime but I don't want it on movies so it's still not a set and forget setting.
If my edge enhancement is the best you've seen then why can't you live with it, but you can live with Reality Creation? Doesn't that mean you think Reality Creation is the best edge enhancement you've seen? Just wondering...

I have a question. I see that there is an option to move the subtitles. Would it be possible to allow the user to change the color and the font too?
DVD and Blu-ray subtitles are bitmaps (images), to change their font you would need to OCR and re-render them, nothing I would ever expect to happen.
For text subtitles, this is the job of the subtitle renderer which usually has such options available. For bitmap subtitles see nevcairiel's reply.

Thanks a lot, as you pointed out at 60 Hz works fine, my monitor was at 144 Hz.

Is this a bug or known issue of Madvr? is there any plan or any way to fix it?
When running your monitor at 144Hz, please turn on the Ctrl+J OSD and check if the refresh rate is properly detected. If it is, are we talking about fullscreen exclusive mode or windowed mode? One thing you could try is to make the GPU queue bigger and increase the number of prepresented frames. That may help.

Updated madvr and have the same problems too. I use LAV filters+madvr+ac3filter with MPC.
Black screen after skiping a video file to another in the same folder.
Every time? Or just sometimes? Are you using D3D11 presentation? What happens if you try D3D9 instead?

Yes, it could just as well be called Chroma Doubling. It's only used for videos where the chroma channels are stored at a lower resolution than the luma channel (most videos, but not all, are stored with the chroma channels half as wide and half as tall as the luma channel). The chroma channel is doubled first (using data from the higher resolution luma channel in the case of Bilateral and Reconstruction), and then the other upscaling and downscaling options operate in (linear) RGB.
FWIW, it used to be this way. And still is, if the video isn't downscaled by noticeable amount. For downscales madVR may decide to scale the chroma channel directly to the target resolution. Because of that it's not always "doubling", but it can also be just "chroma upscaling", or in extreme downscaling cases even "chroma downscaling".

[James Freeman]

True. But I fear in 2-3 years everything LCD/LED will be yesterday's tech. Also, if you're interested in HDR support, even 2016 models are still considered early adopter (meaning: throw away next year) models. Personally, I'll wait for 2017 or 2018 models. Hopefully OLED might be cheaper by then.

We live in an exiting time indeed!

[nevcairiel]

When running your monitor at 144Hz, please turn on the Ctrl+J OSD and check if the refresh rate is properly detected. If it is, are we talking about fullscreen exclusive mode or windowed mode? One thing you could try is to make the GPU queue bigger and increase the number of prepresented frames. That may help.

The only mode I found which worked decently on a 144Hz screen is D3D9 Overlay. All others have the present queue running nearly empty. Not sure why this is tied to the refresh rate though.

[Sunset1982]

Quote:
Originally Posted by Sunset1982 View Post
Will there be an option to customize the color of the seekbar in FSE mode when we arrive madvr 1.0? Are there plans to support chaptermarks? That would be really cool, so one vote from me! :-D


No plans right now. Maybe some time in the future.

Maybe you could add it to your todo list, even with low priority. Thx madshi!

[leandronb]

Every time? Or just sometimes? Are you using D3D11 presentation? What happens if you try D3D9 instead?


Just sometimes, after 2 or 3 episodes and is not episode related because when is not fullscreen i cann skip fine all the episodes.
I am using d3d9.
Right now an episode ended and auto skipped to the next, the screen went black, i quit fullscreen and return to previous episode then back to the next and worked fine. Is something related to fullscreen or refresh rate switch i think.

[FDisk80]

Anyone else seeing strange jitter when using MPC-BE + LAV + madVR with Exclusive Mode when seek bar is on screen?
I checked for frame drops or glitches but they are not registering.

When I reveal the seek bar I'm seeing Present queue dropping from "7-8 / 8" to "1-2 / 2".

Maybe this is the reason? Any way to fix this?

Strangely this jitter does not happen with internal MPE-BE filters instead of LAV.

[Stereodude]

Hmm, something's wrong with DXVA Upscale colors

DXVA vs Spline
http://screenshotcomparison.com/comparison/162815


DXVA vs Jinc
http://screenshotcomparison.com/comparison/162814


It seems to only affect upscale, not DXVA downscale.
Anyone else who experienced this?


Using madvr 0.90.10
Nvidia 361.91
Check the settings in the video card control panel. Some of the color, brightness, or contrast setting have probably been changed.

Edit: I see several others have beaten me to the suggestion.

[mogli]

When I reveal the seek bar I'm seeing Present queue dropping from "7-8 / 8" to "1-2 / 2".
That's madVR's low latency mode to react more quickly to your clicks.

[Stereodude]

Anyone else seeing strange jitter when using MPC-BE + LAV + madVR with Exclusive Mode when seek bar is on screen?
I checked for frame drops or glitches but they are not registering.

When I reveal the seek bar I'm seeing Present queue dropping from "7-8 / 8" to "1-2 / 2".

Maybe this is the reason? Any way to fix this?

Strangely this jitter does not happen with internal MPE-BE filters instead of LAV.
I'm pretty sure it's intentional. The queue's drop so there's not all the extra frames of lag from your clicking on it.

[madshi]

Hey madashi, I sometimes get an "exclusive mode failed" error. Sometimes it works sometimes not. I have attached a log file. Maybe you could look into it if you find some time. Thanks!

https://www.dropbox.com/s/aw95mu8hj35ui04/madVR%20-%20log.rar?dl=0
Interesting, haven't seen this one yet. It means that D3D11 believes that there's some other window occluding the madVR rendering window. Do you have some 3rd party application running which may be in the foreground? Try closing down 3rd party apps, maybe you can find the guilty one.

Okay, here is a log. I hope I did it right. I went FSE, activated the OSD, started playback, and turned off the OSD a few seconds in after about 12 dropped frames had occurred.

http://stereodude.net/madVR_Stereodude_frame_drops.7z
Hmmmm... In that log a fade wasn't even detected at all. Are you sure you uploaded the right log? At least from what I can see in this log there are no frame drops caused by the "rerender frames when fade in/out is detected" functionality.

The only mode I found which worked decently on a 144Hz screen is D3D9 Overlay. All others have the present queue running nearly empty. Not sure why this is tied to the refresh rate though.
Strange. Not really sure what to say...

Just sometimes, after 2 or 3 episodes and is not episode related because when is not fullscreen i cann skip fine all the episodes.
I am using d3d9.
Right now an episode ended and auto skipped to the next, the screen went black, i quit fullscreen and return to previous episode then back to the next and worked fine. Is something related to fullscreen or refresh rate switch i think.
Ok, maybe I can see something if you create a debug log.

Anyone else seeing strange jitter when using MPC-BE + LAV + madVR with Exclusive Mode when seek bar is on screen?
I checked for frame drops or glitches but they are not registering.

When I reveal the seek bar I'm seeing Present queue dropping from "7-8 / 8" to "1-2 / 2".

Maybe this is the reason? Any way to fix this?

Strangely this jitter does not happen with internal MPE-BE filters instead of LAV.
The present queue behaviour is as intended, as already explained by mogli and Stereodude. The jitter is probably a driver issue. Most users don't have this issue, some have. There's not much I can do about it, except disabling the "low latency" mode, which however is very nice because it greatly increases the responsiveness of GUIs.

[leandronb]

Ok, maybe I can see something if you create a debug log.



How can i do that? i searched online and read that i have to run the bat present in the folder of madvr, but i couldn't find this file.

[l0rdraiden]

When running your monitor at 144Hz, please turn on the Ctrl+J OSD and check if the refresh rate is properly detected. If it is, are we talking about fullscreen exclusive mode or windowed mode? One thing you could try is to make the GPU queue bigger and increase the number of prepresented frames. That may help.


The refresh rate is properly detected
It happens in window mode fullscreen mode, and fullscreen exclusive mode.
I have increased the GPU queue to 16 and now it works fine. It's weird because the GPU queue was and is now full all the time, no drops

I hope this feedback can help you to fix or improve something or maybe increase the GPU queue default.

Thanks a lot

[madshi]

How can i do that? i searched online and read that i have to run the bat present in the folder of madvr, but i couldn't find this file.
It should be there, at least it's in the download from my homepage. Do you at least have a "madVR [debug].ax" file? Are you using a 32bit or 64bit media player?

The refresh rate is properly detected
It happens in window mode fullscreen mode, and fullscreen exclusive mode.
I have increased the GPU queue to 16 and now it works fine.

I hope this feedback can help you to fix or improve something or maybe increase the GPU queue default.
Ok, good to know. Increasing the default GPU size is dangerous because it might result in some users running out of GPU RAM.

[madshi]

madVR v0.90.11 released

http://madshi.net/madVR.zip

* added support for receiving 3D subtitle depth from splitter
* added support for rendering subtitles and OSD at a specific 3D depth
* added ISubRenderCallback4 interface to allow ISR to render 3D subtitles
* downscaling anti-ringing filter is now much more agressive
* removed integrated anti-ringing filter from "crispen/sharpen edges"
* added optional external anti-ringing filter for ringing sharpening algos
* re-added experimental "linear light" option for "crispen edges"
* image doubling can now be forced on (if no scaling is done otherwise)
* added 25%, 50%, 75%, 100% strength option to SSIM downscaling
* fixed: MPC-HC ISR subtitles didn't work in 3D mode
* fixed: 3D rendering with Bilinear downscaling wasn't really using Bilinear
Notes:

1) 3D subtitle rendering needs the latest LAV nightly build from last night. In theory different subtitle tracks could have different depth information. Currently I'm ignoring that. Proper support for using the right depth information for the currently selected subtitle track will be added to the next build, but will also require a new LAV build. For now the current solution should work ok.

2) 3D subtitle rendering currently only works with XySubFilter. Making the MPC-HC/BE internal sub renderers support that, too, needs work from the MPC devs, but it should be pretty easy to implement, and madVR already supports new interfaces for that.

3) The downscaling anti-ringing filter is now nearly "perfect" in that it should remove pretty much every bit of ringing caused by downscaling. However, there's a certain danger that this might introduce aliasing or other artifacts. So I need you to test this and let me know if you find any problems.

4) "crispen edges" and "sharpen edges" have a stronger effect now because their internal anti-ringing filters were removed. "crispen edges" now looks much nearer to the original FineSharp algorithm. The new sharpening anti-ringing filter should help remove most ringing artifacts caused by "crispen edges", "sharpen edges" and "LumaSharpen", without lessening the sharpening effect much. Because the new AR filter seems to work pretty well, I've re-introduced a "linear light" option for "crispen edges" (formerly known as FineSharp). Would be great to get your feedback about whether you prefer linear light on/off, and whether there are still the same artifacts that made us remove the FineSharp linear light option in the past (at that time we didn't have a good AR filter). Also please let me know if the AR filter works as expected.

5) You can now force doubling even if no scaling is needed (e.g. Blu-Ray playback on a 1080p display). When doing that, the upscaling refinement options are applied on the doubled (2160p) resolution. This practically allows you to run all the sharpening algos in supersampling mode. Extra performance hungry, but allows quite nice results, especially in combination with Bicubic150 or SSIM downscaling.

I've put the things I need feedback for in bold, just to be safe. Hope I don't overexert you guys.

[leandronb]

It should be there, at least it's in the download from my homepage. Do you at least have a "madVR [debug].ax" file? Are you using a 32bit or 64bit media player?


I think that is the problem, i use klite codec pack with only lav filters and madvr. Now i installed the madvr version that you just released and it came with much more files.
The interesting thing is that the problem is gone at least for now, one thing i notice is that sometimes when i open a video file the screen goes black for a while, and is random.
But i had this for a while and do not bother me at all.
If the other problem happen again i will post here with the log.
thanks.

[foozoor]

madVR v0.90.11 released

http://madshi.net/madVR.zip

* added support for receiving 3D subtitle depth from splitter
* added support for rendering subtitles and OSD at a specific 3D depth
* added ISubRenderCallback4 interface to allow ISR to render 3D subtitles
* downscaling anti-ringing filter is now much more agressive
* removed integrated anti-ringing filter from "crispen/sharpen edges"
* added optional external anti-ringing filter for ringing sharpening algos
* re-added experimental "linear light" option for "crispen edges"
* image doubling can now be forced on (if no scaling is done otherwise)
* added 25%, 50%, 75%, 100% strength option to SSIM downscaling
* fixed: MPC-HC ISR subtitles didn't work in 3D mode
* fixed: 3D rendering with Bilinear downscaling wasn't really using Bilinear
Notes:

1) 3D subtitle rendering needs the latest LAV nightly build from last night. In theory different subtitle tracks could have different depth information. Currently I'm ignoring that. Proper support for using the right depth information for the currently selected subtitle track will be added to the next build, but will also require a new LAV build. For now the current solution should work ok.

2) 3D subtitle rendering currently only works with XySubFilter. Making the MPC-HC/BE internal sub renderers support that, too, needs work from the MPC devs, but it should be pretty easy to implement, and madVR already supports new interfaces for that.

3) The downscaling anti-ringing filter is now nearly "perfect" in that it should remove pretty much every bit of ringing caused by downscaling. However, there's a certain danger that this might introduce aliasing or other artifacts. So I need you to test this and let me know if you find any problems.

4) "crispen edges" and "sharpen edges" have a stronger effect now because their internal anti-ringing filters were removed. "crispen edges" now looks much nearer to the original FineSharp algorithm. The new sharpening anti-ringing filter should help remove most ringing artifacts caused by "crispen edges", "sharpen edges" and "LumaSharpen", without lessening the sharpening effect much. Because the new AR filter seems to work pretty well, I've re-introduced a "linear light" option for "crispen edges" (formerly known as FineSharp). Would be great to get your feedback about whether you prefer linear light on/off, and whether there are still the same artifacts that made us remove the FineSharp linear light option in the past (at that time we didn't have a good AR filter). Also please let me know if the AR filter works as expected.

5) You can now force doubling even if no scaling is needed (e.g. Blu-Ray playback on a 1080p display). When doing that, the upscaling refinement options are applied on the doubled (2160p) resolution. This practically allows you to run all the sharpening algos in supersampling mode. Extra performance hungry, but allows quite nice results, especially in combination with Bicubic150 or SSIM downscaling.

I've put the things I need feedback for in bold, just to be safe. Hope I don't overexert you guys.

Thank you so much madshi, that's awesome. ;)

[leeperry]

:thanks: for the new build, eventually you've come back with a zillion new options to try but I was already quite amazed this afternoon by 720p30@1080p video footage using FRC/RS+SR@1 chroma/Jinc AR upscaling+SR@1 luma and Jinc non-AR downscaling, both very sharp and yet not digital looking wow oh wow.....but then again you just added strength sub-options to SSIM so let's roll the dice once more shall we.

Jinc never disappoints IME except that it's very sharp when downscaling, would strength options for Jinc downscaling also be possible please? I rest my case that softer SR would also be very useful with such sharp downscalers but it's your call. At some point you gotta find the best upscaling/downscaling sharpness compromise as the goal is to neither oversharp one or the other IMO.

Would forcing SR also be possible please? I love the look on 1080p movies and I'm currently forced to upscale one notch :o

[nevcairiel]

Would forcing SR also be possible please? I love the look on 1080p movies and I'm currently forced to upscale one notch :o

SR relies on comparing the upscaled and original image, I don't think it can really work without any scaling. You could use the forced doubling if you want SR to always be active.

[leeperry]

SR relies on comparing the upscaled and original image, I don't think it can really work without any scaling. You could use the forced doubling if you want SR to always be active.
Sure thing, a 0.001 pixel upscale would work for me. The smallest upscale technically possible to allow SR to kick in basically as 1080p movies don't usually require doubling anyway.

[FDisk80]

madsi, the stats text turned back after v0.90.11 update. Very hard to read. It only turns white if I move the mouse to the seek bar in fullscreen or if a video is paused in windowed or fullscreen mode.

[yok833]

Thanks Madshi!! Awesome as always ! :)
With the last build I have the subtitles written in black and also the text of the information when I press Ctrl+J... Have I pressed a wrong button or is it due to the last build??

[madshi]

Huh? Sounds weird, and doesn't happen on my PC. Can I see a screenshot, please? (Please don't attach it to this forum, but upload elsewhere and just post a link).

[ryrynz]

madVR v0.90.11 released
5) You can now force doubling even if no scaling is needed (e.g. Blu-Ray playback on a 1080p display). When doing that, the upscaling refinement options are applied on the doubled (2160p) resolution. This practically allows you to run all the sharpening algos in supersampling mode. Extra performance hungry, but allows quite nice results, especially in combination with Bicubic150 or SSIM downscaling.


Great stuff, was wondering whether this would be supported. Wonder if it's worth it... We'll see soon enough.

[yok833]

Huh? Sounds weird, and doesn't happen on my PC. Can I see a screenshot, please? (Please don't attach it to this forum, but upload elsewhere and just post a link).

Here it is
https://goo.gl/photos/mZ1VnWQAghWik7ct9

when I press stop or an arrow to move forward the information on the left side and the subtitles turn white again???

[FDisk80]

Huh? Sounds weird, and doesn't happen on my PC. Can I see a screenshot, please? (Please don't attach it to this forum, but upload elsewhere and just post a link).

Here you go
http://i.imgur.com/T2DEK3u.jpg

[FDisk80]

Here it is
https://goo.gl/photos/mZ1VnWQAghWik7ct9

when I press stop or an arrow to move forward the information on the left side and the subtitles turn white again???

Yes, same issue.

[Stereodude]

Hmmmm... In that log a fade wasn't even detected at all. Are you sure you uploaded the right log? At least from what I can see in this log there are no frame drops caused by the "rerender frames when fade in/out is detected" functionality.
I only made one log, so it can't be the wrong one. I can repeat the test and make a log each way if that would help.

What are the dropped frames from? It drops a few frames at the very start of playback (not unexpectedly), but the drops after that are a mystery to me since the queues don't empty and the max rendering time is still way less than 1/refresh.

[FDisk80]

Also volume OSD is not visible anymore. It just shows the background.

Edit: Looks like the white text comes back in "low latency" mode for some reason. As soon as the Present queue goes up it turns black again and for volume OSD turns transparent.

[Unr3aL]

The only mode I found which worked decently on a 144Hz screen is D3D9 Overlay. All others have the present queue running nearly empty. Not sure why this is tied to the refresh rate though.

I think this was discussed a long time ago. It's hard to say now, but wasn't the issue tied to multi-monitor setups alone?
I thought it had something to do with timings not playing well if one of the monitors had a high (120, 144Hz) refresh rate and happened to be the one the movie was played on...
It could've also just been a wild guess at that time.

I myself happen to have a 120Hz monitor and I use a projector for movies.
Playing @120Hz on the monitor never really worked for me without a lot of visible jerking and judder as well as tons of presentation glitches.
The only thing that helped was going back to old path and playing around with the number of backbuffers used.
That issue (for me) dates back to when the new render path was introduced.

I've never tried upping the GPU queue though, so far I am auto-changing to 60Hz - minor annoyance...

Greetz, Unr3aL67

[Manni]

Here it is
https://goo.gl/photos/mZ1VnWQAghWik7ct9

when I press stop or an arrow to move forward the information on the left side and the subtitles turn white again???

Same issue here with OSD in black since 0.90.11.

[RyuzakiL]

SOrry for the noob question but for what use is the scale in sigmoidal light in Image Upscaling?

I don't see any noticeable changes in PQ with it on but I notice some slight decrease in performance.

And also I'm using the latest version which is v0.90.11 and I experience black colored fonts in the subs and ctrl+j.

and also ctrl+j reports that I'm still using Lanzcos3AR even if I selected JincAR as my Image Upscaler so what gives??

[huhn]

4) "crispen edges" and "sharpen edges" have a stronger effect now because their internal anti-ringing filters were removed. "crispen edges" now looks much nearer to the original FineSharp algorithm. The new sharpening anti-ringing filter should help remove most ringing artifacts caused by "crispen edges", "sharpen edges" and "LumaSharpen", without lessening the sharpening effect much. Because the new AR filter seems to work pretty well, I've re-introduced a "linear light" option for "crispen edges" (formerly known as FineSharp). Would be great to get your feedback about whether you prefer linear light on/off, and whether there are still the same artifacts that made us remove the FineSharp linear light option in the past (at that time we didn't have a good AR filter). Also please let me know if the AR filter works as expected.

even 4.0 sharp edges or crispen edge has nearly no effect when scaling 1080p to 2160p even with disabled AR filter.

[CarlosCaco]

Hey madshi i tried to test supersampling on 720p content on my 1360x768 screen, the doubling is working but upscaling refinements not, no matter what sharpen algo i check dont do any diference on image and on osd don't affect the render times... Same with super resolution!
I even tried to check all the upscaling refinements but has no changes on picture and render times...

Remember that on last month i asked why upscaling refinements not working on 720p on 1360x768 screens.
You said that because upscaling was very little so upscaling refinements are not activated

I was thinking if somewow this can be relate to that (supersampling upscaling refinements not activated because the Target resolution is the same or very near the movie original resolution)

By the way thank you again for madvr, every week we get a better picture quality! Its very impressive!

[Thunderbolt8]

This practically allows you to run all the sharpening algos in supersampling mode.
what does this mean? that normal BD playback on a 1080p screen now looks better with all this applied than when just viewing the movie normally?

[leandronb]

Updated to the newer version and after a while with no problems it gave me the same error.
I used the debug mode to create the log, it created a 69MB file and i compressed it to 3mb and uploaded to mega.
Steps i did, opened an episode, went fullscreen, skipped 1, skipped 2 and it gave me the black screen i said earlier, hit esc to exit fullscreen and closed mpc.
this is the link:
https://mega.nz/#!tAIVEIoL!YjThiXXYoD9DNx9BRXmbgsAn2dYTS2yv0isJBuGLuGU

thanks in advance.

[ryrynz]

what does this mean? that normal BD playback on a 1080p screen now looks better with all this applied than when just viewing the movie normally?

Potentially yes. The algorithms have a larger image to work on that gets downscaled so the end result should look neater than them being applied to a straight 1080 image.. Then you get the visual benefits of downscaling from one of madVR's HQ downscalers. Essentially you'll be getting as close to 4k as you can get on your full HD screen with today's readily available full HD content.

[huhn]

what does this mean? that normal BD playback on a 1080p screen now looks better with all this applied than when just viewing the movie normally?

"better" is clearly not the term i would use here "different" is more like it.

it is clearly against the creators intent and a good source doesn't need sharpening.