/mvsm/ - mpv vs madvr

File: mpv_vulkan.png (193KB, 512x512px) Image search: [Google]

193KB, 512x512px

Snif, no more commits on the vulkan branch.

>>56875024
Haasn was quoted a few threads ago where he said something like he's going into hibernation again now and listed some stuff others can do when they want to continue it.

>>56875739
>hibernation
what is he, a fucking bear?

how to do stereo width on mpv now?
# Extra Stereo filter
af=lavfi=[extrastereo=1.5]

is deprecated now ;_;

File: questionss.jpg (43KB, 716x720px) Image search: [Google]

43KB, 716x720px

how do I get rid of these libva messages?

 $ mpv https://www.youtube.com/watch\?v\=3wC1dD-JwWc
libva info: VA-API version 0.39.2
libva info: va_getDriverName() returns 0
libva info: Trying to open /usr/lib/x86_64-linux-gnu/dri/i965_drv_video.so
libva info: Found init function __vaDriverInit_0_39
libva info: va_openDriver() returns 0
libva info: VA-API version 0.39.2
libva info: va_getDriverName() returns 0
libva info: Trying to open /usr/lib/x86_64-linux-gnu/dri/i965_drv_video.so
libva info: Found init function __vaDriverInit_0_39
libva info: va_openDriver() returns 0
libva info: VA-API version 0.39.2
libva info: va_getDriverName() returns 0
libva info: Trying to open /usr/lib/x86_64-linux-gnu/dri/i965_drv_video.so
libva info: Found init function __vaDriverInit_0_39
libva info: va_openDriver() returns 0
$ 

also, in my mpv.conf I have this:

ao=alsa:resample=no  

but now I'm getting this message:

    ao/alsa: Using suboptions is deprecated. Use the global '--alsa-resample' option instead of 'resample' suboption.

not sure how to fix this. changing it to:

1 ao=resample=no 

results in NO sound :(

>>56878314

ao=alsa
also-resample=no

>>56878559
TY anon!

now I only have libva messages to deal with.. they're so fucking annoying.

Why doesn't Fedora add MPV to their yum software repository?

After I upgraded mpv, I've been seeing this

   cplayer: --vo=opengl-hq is deprecated! Use --profile=opengl-hq instead.

My config file

[hardware-auto]
profile-desc="iGPU low energy defaults"
vo=opengl-hq
hwdec=auto-copy
video-sync=display-resample

[default]                                                                                                             profile=hardware-auto 

wat do?

>>56878698

vo=opengl
profile=opengl-hq

>>56874625
>MPV
>can't press a shortcut to automatically normalize audio
DROPPED

>>56878710
Werks! I've left hwdec & video-sync options. That ok? thanks!

what's the mpv equivalent for ctrl+j in madvr?

Anyone has a sane input.conf file?

>>56878940
stats.lua

>>56878735

N af toggle lavfi=dynaudnorm

next?

>>56878949
thanks

>current year
>watch video of some people talking
>most of the frame is wasted
>the two people it cuts between don't populate the same portion of the frame, so cropping won't work
>tfw can't tell MPV to track faces and crop in on them
it hurts to live

>>56879139
vf_opencv when?

"60fps" videos on mpv?

File: 60fps.jpg (403KB, 2048x1080px) Image search: [Google]

403KB, 2048x1080px

>>56879576
Works fine for me?

>>56874625
How do I change mousewheel to volume control and not video skip?

>>56879603
I meant software tweaking like SVP

>>56879659
>SVP
You mean “24 fps with shittons of artifacts on top”?

>>56879676
Exactly

>>56879712
Well if that's your cup of tea, try MVTools (original) or SVP (shitty russian fork)

>>56879726
Thanks anon, always heard about mpv here. so i decided to finally give it a try after always using MPC.

>>56879643

MOUSE_BTN3 add volume 2
MOUSE_BTN4 add volume -2

>>56879873
i prefer this:

    MOUSE_BTN3 add volume -2
   MOUSE_BTN4 add volume 2

makes a lot more sense if you use smartphones etc.

>>56879893
>if you use smartphones etc.
kill yourself

File: Old Man Yells at Cloud.jpg (47KB, 800x450px) Image search: [Google]

47KB, 800x450px

>>56879912

File: weird_bears.gif (2MB, 480x270px) Image search: [Google]

2MB, 480x270px

>>56875854
Probably...

Is there a deblocking shader somewhere?
Deband sometimes helps but it's not enough.

>>56882050
What kind of deblocking shader would you be interested in?

that aside, deband with a high radius can work well enough

How do you restart the player while retaining the same file and position of playback with a keybind?

>>56874625
>low
>medium
>high
I literally don't get it. Its not like video games where the special effects are created as you play.
You are literally viewing a bunch of flat pictures. How can there be quality settings?

>>56878667
MPV uses nonfree softwares.

How do I get MPV to not resize in windowed mode? I want it to always be at 1280 pixel width.

>>56883206
They distort the frames of video with effects much like using photo editing software to "enhance" images, you need power to make big enhancement with so many frames in real time.

When I'm using the nnedi3 luma prescaler opengl-shader, everything seems to shift a little bit to the lower right compared to using just ewa_lanczossharp for scale. See http://screenshotcomparison.com/comparison/186421
What gives? Which one is fucking up, because I can't see it being possible that one scaler would be so much more "correct" that it would shift the image that much.

>>56883206
higher profile - less distortions

>>56883259
yes mpv's nnedi3 shifts image

I use mpc and madvr as an external player for kodi, how do i do this with mpv on linux?

>>56883275
That has to be a bug though, right? You can't shift the image without destroying and adding some information, which is basically artifacting. Did I get memed on when people said the prescaler was "good"?

>>56883297
>Did I get memed on when people said the prescaler was "good"?

Nobody who actually develops mpv thinks NNEDI3 is good. You were just likely told so by dumb windows users.

>>56883297
>//!OFFSET 0.000000 -0.500000
this is why I guess
mpv is the only player that shifts image with prescalers I guess.
>Did I get memed on when people said the prescaler was "good"?
Don't know who told you that, I was always saying that prescalers are shit.

>>56883337
>>56883346
There were plenty of configs posted here that included those, even when it was still included in mpv itself. But I can't say that those people weren't retarded. Oh well, at least I learned that I shouldn't use that shit then today.

How are the shaders that the OP posted? Also shit?

>>56883372
>How are the shaders that the OP posted?
One is chroma scaler, another is antiringing with sharpening, I use second one myself.

Is prescaling only good for supersampling (so with downscaling)?
Why not prescaling with linear scalers like spline36 or jinc if nnedi3 shifts the picture?

>>56883396
Sound superfluous to me then. I don't notice any ringing, and I believe ewa_lanczossharp is sharp enough. Chroma also seems like it's better than with crossbilateral.

File: Olga Katysheva insta - Olgachocolate 1473974739670.jpg (102KB, 640x640px) Image search: [Google]

102KB, 640x640px

Any wangblows user here? I've been using last year build, new one giving few trouble.
Any takers for spoonfeeding? I am on chinkpad x201, win7.

1.need to send screenshot to same directory as video. this is sending on desktop. old build worked fine

screenshot-template="~/screenshots/mpv_ss_%F_%tY.%tm.%td_%tH-%tM-%tS.%P"

2.video is not rotating with

Ctrl+r cycle_values video-rotate "90" "180" "270" "0"

my vf are to blame?

#vo=opengl:backend=dxinterop
vo=direct3d_shaders
#vo=opengl:backend=angle
video-sync=display-resample
#hwdec=auto
#hwdec=dxva2-copy
hwdec=no

3.autocrop errors saying changing filters failed.
https://github.com/mpv-player/mpv/blob/master/TOOLS/lua/autocrop.lua

>>56883372
>There were plenty of configs posted here that included those, even when it was still included in mpv itself.

And? Do you just blindly copy other people's configs? Try reading the manual and figuring things out for yourself for a change. Even then those configs were likely posted by dumb windows users like yourself I suppose.

>>56883460
>vo=direct3d_shaders

The fuck are you smoking?

>>56883291
bump

>>56883487
>chinkpad x201,win7
that's best i could do m8

>>56883471
>Do you just blindly copy other people's configs?
No, they said they were better, and most of the time these changes are very obvious. I did try it out though, also there wasn't really any documentation about how "shit" this is. So it wasn't just a blind copy and paste, I did research it.

>>56883505
nigga pls i use an HP house heater on low madvr settings and opengl-hq for mpv
your chinkpad must be absolute trash

>>56883554
i was getting artifacts in fullscreen, brightness/contrast/saturation disabled. wtf was i supposed to do faggot?

>>56883596
scam sell your chink trash to some other /g/ autist and get something better

>>56883456
>I believe ewa_lanczossharp is sharp enough
How it can be sharp if it's blurier than source image?

>>56883622
tell me something with half as good keyboard and i'll buy it.

>>56883684
By being sharper than ewa_lanczos.

>>56883729
By how much?

>>56883746
About 3 sharp units.

>>56883817
TIL: 3 sharp units are completely unnoticeable.

>>56874625
I love this picture, especially the "be brillant" part in the background.

This doesn't work, why? :(
z cycle-values profile "sharp" "soft"

Why madshi doesn't want to share the madVR source exactly?

>>56884916
because it's shit and unoptomized

his dithering implementations are completely mongoloid

>>56884944
>because it's shit and unoptomized
So why are many people still using it?

>his dithering implementations are completely mongoloid
There are two dx11 only experimental dithering options.
Nothing about mongoloid here.

>>56884944
Please report these here:
http://bugs.madshi.net/view_all_bug_page.php

>>56883259
>>56883275
>>56883297
Sounds like an issue with the NNEDI3 script you're using

>>56883275
Yes, but if the script annotates the correct OFFSET then mpv's scaler would correct the shift

>>56883297
>You can't shift the image without destroying and adding some information, which is basically artifacting
Good job describing what NNEDI3 does

>Did I get memed on when people said the prescaler was "good"?
Yes

>>56884916
how many windows only programs share their source code?

>>56885119
>Sounds like an issue with the NNEDI3 script you're using
show me nnedi3 script that doesn't shift image.

>>56883746
>>56883817
>>56883846
By exactly 1.9107694051635482%

>>56884464
‘profile’ is not a normal setting. You could toggle the options directly though

>>56885157
Just tested this one and there's no shift for me

https://raw.githubusercontent.com/bjin/mpv-prescalers/master/nnedi3-nns16-win8x4-all.hook

>>56885263
Actually, it seems like a very recent change broke this. Let me bisect it.

>>56885275
The one I used was changed 4 months ago, so I doubt it.

>>56885275
Both nnedi3 and superxbr were shifting image from the very beginning, one of the reasons I disliked them.

>>56885275
>>56885289
Oops, never mind - I'm stupid, disregard that. It still works fine on current master.

>>56885292
Well it works fine for me

>>56884464
What are you even trying to do?

>>56884916
>Why madshi doesn't want to share the madVR source exactly?
He is scared others might "steal" stuff from him and make a better product than he did.
That's not even a joke as he said so in the doom9 forums. He also said he doesn't want to talk about details of future plans because he fears others (he mentioned mpv) will do it better.

Think about it. Isn't that quite ridiculous? This cuck is so fucking afraid that he can't compete with his windows-only renderer that he stops talking about his plans, lol.
At the same time his well known quality renderer was surpassed in quality by open source freetards in a very small amount of time, for all major operating systems. No wonder this cuck is scared but what's even funnier: mpv is open source! He could steal all he wants, e.g. the superior color management.
Don't forgot: people can't "steal" from him (only ideas) while he could/can steal everything from mpv. He doesn't have to respect the license as nobody can see his code anyway. But still, he's constantly loosing ground.

I guess, as he once said, he really wanted to make a pro version of madVR but is scared that mpv either already has or will implement whatever pro features he can offer so there is no way for him to monetize madVR anymore. And I guess that makes him pretty salty.

I'd really like to see this cuck in real world.

File: 1458776398806.jpg (353KB, 920x900px) Image search: [Google]

353KB, 920x900px

>>56885386

>>56885386
>mpv is open source! He could steal all he wants, e.g. the superior color management.
mpv is GPL-licensed, so he'd have to open source madVR in order to do so.

Then again, madVR is technically a GPL violation already. So is SVP. Sadly too small for any company to consider suing.

>>56885386
>I guess, as he once said, he really wanted to make a pro version of madVR but is scared that mpv either already has or will implement whatever pro features he can offer so there is no way for him to monetize madVR anymore. And I guess that makes him pretty salty.
I swear on my mother's grave, if madshi ever makes a pro version of madVR I will work day and night to make sure mpv implements all “madVR pro only” features

>>56885119
>mpv's scaler would correct the shift
Yes, image is shifted when upscaling is exactly 2x.

>>56885421
>mpv is GPL-licensed, so he'd have to open source madVR in order to do so.
Yah of course but let's face in reality: nobody cares. Especially not closed source programs. Who would notice it or even be able to prove it...
He can do what he wants and he certainly would if it were to benefit him.

>>56883291
N-no one? pls

>>56885438
>I swear on my mother's grave, if madshi ever makes a pro version of madVR I will work day and night to make sure mpv implements all “madVR pro only” features
Exactly and I'm pretty sure you are not the only one.
But let's be honest, if he ever were to make a pro version then he is not going to bring new pro-only features but rather declares some existing ones as pro-only. Maybe both.

Either way, mpv and the community around it made it completely impossible for him to make any money with madVR as he planned for a long time.
Since apparently he can't fight with a good product (even though he has full access to the competitor) he tries to discredit mpv wherever he can.
A natural reaction but a cuck and dick-move nonetheless.

>>56885119
>>56885482
Oh wait, I think I know what might be happening.

You might be hitting this line of code:
https://github.com/mpv-player/mpv/blob/master/video/out/opengl/video.c#L2044

Basically, if you're using --scaler-resizes-only (enabled by default) and the NNEDI3-prescaled texture matches the your output size, the shift would not get corrected by mpv (by design).

Try with a different sized window

>>56885488
>Who would notice it or even be able to prove it...
Well we know madVR includes GPL'd OpenCL code because he publishes the source code for that part of the program already.

However what he fails to realize is that using GPL'd code makes his entire renderer a derivative work of that code, which under the GPL means he has to publish the source code of the rest of madVR as well.

He doesn't do this, so legally the author of the corresponding script could sue him.

>>56885632
>Well we know madVR includes GPL'd OpenCL code because he publishes the source code for that part of the program already.
Yeah but only because people know what NNEDI is. He can't really say "hey guys, I have this new algo that looks like NNEDI but is totally different code!".
However, he can just take mpv's color management code and claim he improved his color management, or some existing scaler algorithms, debanding, dither, .... Nothing suspicious here. With nnedi it's obvious though.

>>56885616
Yes, there is no shift with scaler-resizes-only=no

>>56885616
So, the shader assumes that after doubling the resolution, it will be shifted/scaled appropriately by mpv. But mpv doesn't do shit to it, because it just sees a correct resolution already? I'd say that's still a bug in the shader then, it's not presenting a correctly positioned texture then, right?
It seems what you're saying is true though, this only happens with 720p video fullscreen on my WQHD monitor.

>>56885963
See the comment in the source code. Basically, it's a conflict of interests.

1. Shifting by a half-pixel is a very bad idea. It will basically make everything blurry, and destroy the point of even using NNEDI3.

2. Shifting by a half-pixel requires a full, expensive scaler pass. The point of scaler-resizes-only is to avoid having to run the heavy-weight scaler when a no-op could do the same thing for much cheaper.

So there's no way to simultaneously have
>high quality
>cheap noop scale
and
>correct half-pixel adjustment

So the choice of mpv is to pick the latter over the former, because being off by a half-pixel is far less big a deal than being blurry and expensive.

>>56886007
So, there's no way to do it right. So using NNEDI3 is going to always be shit if you happen to be doubling the resolution.
Fair enough.

>>56886031
>Half-pixel offset
>Shit
Why do you care?

>>56886121
Because I don't think having the entire image be faulty is good. Also, it's shifted compared to chroma at that point, right? That means everything is misaligned. And considering I don't think NNEDI3 would give that much better scaling compared to ewa_lanczossharp, this error is just too much for me.

>>56886230
>Because I don't think having the entire image be faulty is good.
???
It's just off by a half-pixel. I seriously don't get what the big deal is

>Also, it's shifted compared to chroma at that point, right?
No

>>56886275
>???
What don't you understand? The entire image is shifted, thus the entire image is faulty. I'm not saying it's a big deal, but it's a bigger deal than having no obvious errors.

>No
But it's a luma prescaler. Which presents the image shifted compared to what the original is. Chroma would still be scaled to the correct position. Thus the end chroma would be shifted compared to end luma.

File: ab.jpg (425KB, 2000x1000px) Image search: [Google]

425KB, 2000x1000px

>>56886323
>What don't you understand? The entire image is shifted, thus the entire image is faulty.
But it doesn't impact image quality in any way

Or tell me which one of these two images do you think looks better?

>Chroma would still be scaled to the correct position.
No

>>56886503
>But it doesn't impact image quality in any way
I disagree, considering it's not correct with the original image. If it was, it wouldn't be "shifted".

>Or tell me which one of these two images do you think looks better?
False dilemma, doing one doesn't mean I can do the other.

>No
How is that possible then? The prescaler only touches luma. mpv doesn't know what the prescaler does in regards to the position. So how could it scale the chroma "incorrectly" as well then?

>>56886540
>False dilemma, doing one doesn't mean I can do the other.
So you agree that they're the same quality? Even though one is shifted by 1 pixel?

>How is that possible then?
Feel free to look at the source code?

>mpv doesn't know what the prescaler does in regards to the position.
Yes it does. That's what the OFFSET field in the header is for

>>56886564
>So you agree that they're the same quality?
See >>56886540
>False dilemma, doing one doesn't mean I can do the other.
Maybe try to understand what I said before commenting about it immediately. One is clearly worse then, just compare it to the original image and it should be obvious because one's shifted compared to it. I don't have the original image, so fuck if I know.

>>56886564
>Feel free to look at the source code?
I'm just asking you mate.

File: zero_cool.webm (2MB, 1920x800px) Image search: [Google]

2MB, 1920x800px

>>56886503
Left looks better

t. robot eyes

>>56886621
>Maybe try to understand what I said before commenting about it immediately. One is clearly worse then, just compare it to the original image and it should be obvious because one's shifted compared to it. I don't have the original image, so fuck if I know.
Those sure are a lot of excuses I'm hearing out of you. Either you can see a difference in quality or you can't

>>56886671
Seeing a difference in quality =/= being able to tell that the image is faulty.

>>56886671
>Those sure are a lot of excuses I'm hearing out of you.
Maybe that should indicate to you that what you're asking is just wrong, considering there are a lot of things wrong with it.

>>56879603
how do you get ui like that to see if mpv drops any frames?

>>56886897
>stats.lua
doesn't say anything to me, windows user here.

>>56886690
>>56886700
And yet my point continues to stand. Either you see a difference in quality or you don't.

If you don't see a difference in quality, then there's no point in distinguishing

>>56886852
https://github.com/Argon-/mpv-stats

>>56886906
Yeah I realized I should've posted the link...

>>56886925
thanks!

>>56886907
The only way to be able to see a difference in quality is by knowing the original image. The one that most closely simulates the original image is the one that has the "highest" quality. That's what differentiates scalers from each other, how accurate they are in getting the original image across without changing information, or artifacts. Considering there is no original image, it's impossible to know.
I could post an image of a black square here, and an image of a tree here, and ask you which one is the best quality, and you would not be able to answer, because it all depends on what the original image is.

I'd say there's definitely a point in distinguishing between a scaler that more accurately portrays an image and a scaler that shifts the image. It's why people choose one or the other, it's why there's even the option of using one or the other.

>>56887009
>I could post an image of a black square here, and an image of a tree here, and ask you which one is the best quality, and you would not be able to answer, because it all depends on what the original image is.
Yes but that's clearly not what I did. I posted two images that are exactly the same, except slightly different crop (off by 1 pixel, which is even more of a difference than the half-pixel shift introduced by NNEDI3).

The point is that there's no functional distinction, because you can't possibly make one. The content is effectively, essentially, practically, aesthetically and qualitatively identical.

You could apply a 1-pixel crop to any movie you watch and be unable to appreciate any difference. Claiming to experience one is completely and utterly placebo, and you are making a mountain out of a molehill for absolutely no reason other than massive autism. The pixel shift has absolutely no relevant whatsoever.

The only way you can possibly argue it affects quality is by bending the definition of quality so hard that you can apply mathematical reasoning to deduce their difference. It's a complete and utter non-argument

If you don't want to use a scaler because of a half-pixel difference, then I honestly have no words for you. You should rate a scaler based on how well it upscales the image, and not whether it's off by a single pixel (which is lossless for 99.999% of the image)

>>56886925
one more question, what are VO frames?
Seems like I get no dropped or delayed frames but one VO when I switch to fullscreen.

>>56887174
>Yes but that's clearly not what I did.
It's basically the same aside from it being less extreme. It's what people call an "analogy".

>except slightly different crop (off by 1 pixel, which is even more of a difference than the half-pixel shift introduced by NNEDI3).
So, it's not even scaled. In which case, it's not the same comparison as comparing prescaling to not prescaling.
But considering that information, the original is obviously closer to the original, so that quality it higher.

>The point is that there's no functional distinction
There is, the image is shifted compared to the original image. It's a big artifact. People don't like artifacts in video rendering.

>>56887174
>You should rate a scaler based on how well it upscales the image
That's what's happening here mate. You're the person that's dismissing arguments about how the scaler is producing errors because it's "not noticable", or just "placebo". It's factual that it's faulty, that there's artifacts.

>>56887230
>It's basically the same aside from it being less extreme. It's what people call an "analogy".
No, but here's an analogy for you: Your argument is like taking a question comparing two nearly identical cars (except for 1mm difference in seat hight) and turning that into a question comparing a tricycle and a formula 1 car.

>>56887176
>Seems like I get no dropped or delayed frames but one VO when I switch to fullscreen.
Going with the default framedrop mode (VO) the VO dropped frames are the "normal" (and bad) dropped frames. It's normal getting some when seeking or changing window size, depends on many factors. What's important is that they don't increase during regular playback.

https://mpv.io/manual/master/#command-interface-vo-drop-frame-count
https://mpv.io/manual/master/#options-framedrop

>>56887508
I am not trying to turn this into a question comparing a black image and an image of a tree though. If I was, then maybe that analogy would make sense, but right now it does not.

>>56887979
And yet you're the one trying to invalidate me saying the equivalent “there's no practical difference between these cars. The difference is only 1 mm. That's impossible for you to feel” of by bringing up nonsensical hyperbole involving tricycles and formula 1 cars.

>>56888050
That's a very generalized analogy, but yes I guess.

>>56878941

n playlist-next; write-watch-later-config     # Next file
b playlist-prev; write-watch-later-config     # Previous File

s screenshot     # take a screenshot
S screenshot video     # without subtitles
Ctrl+s screenshot window     # with subtitles and OSD, and scaled

q cycle ontop
ESC quit

#Audio
MOUSE_BTN3 add volume 2
MOUSE_BTN4 add volume -2
UP    add volume 2
DOWN  add volume -2
+ add volume 2
- add volume -2

anyone else wanna share?