/mvsm/ - mpv vs madvr

>low mpv preset:
profile=opengl-hq
deband=no
cscale=spline16
scale=spline16


>medium mpv preset:
profile=opengl-hq


>high mpv preset:
profile=opengl-hq
cscale=ewa_lanczos
scale=ewa_lanczossharp


>insane mpv preset:
gist.github.com/igv/a015fc885d5c22e6891820ad89555637
gist.github.com/igv/2364ffa6e81540f29cb7ab4c9bc05b6b
profile=opengl-hq
scale=ewa_lanczossharp
opengl-shaders="~~/shaders/CrossBilateral.glsl,~~/shaders/ SSimSuperRes.glsl"


>low madvr preset:
chroma upscaling: bicubic100 + ar
image upscaling: lanczos3 + ar
dithering: ordered


>medium madvr preset:
chroma upscaling: catmull-rom + ar
image upscaling: jinc3 + ar
refinement: superres (1)
debanding: medium/high
dithering: ordered


>high madvr preset:
chroma upscaling: super-xbr100 + ar
image upscaling: jinc3 + ar
debanding: medium/high
enhancements: sharpen edges (0.5) + ab (50%) + ar
dithering: ordered


>insane madvr preset:
chroma upscaling: nnedi3 32 neurons
image upscaling: jinc3 + ar
doubling luma: nnedi3 32 neurons
refinement: superres (3)
debanding: medium/high
deringing: on
dithering: error diffusion 2

Other urls found in this thread:

youtube.com/watch\?v\=3wC1dD-JwWc
screenshotcomparison.com/comparison/186421
github.com/mpv-player/mpv/blob/master/TOOLS/lua/autocrop.lua
bugs.madshi.net/view_all_bug_page.php
raw.githubusercontent.com/bjin/mpv-prescalers/master/nnedi3-nns16-win8x4-all.hook
github.com/mpv-player/mpv/blob/master/video/out/opengl/video.c#L2044
github.com/Argon-/mpv-stats
mpv.io/manual/master/#command-interface-vo-drop-frame-count
mpv.io/manual/master/#options-framedrop
twitter.com/NSFWRedditGif

Snif, no more commits on the vulkan branch.

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.

>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 ;_;

how do I get rid of these libva messages?

$ mpv 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 :(

ao=alsa
also-resample=no

TY user!

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?

vo=opengl
profile=opengl-hq

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

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?

stats.lua

N af toggle lavfi=dynaudnorm

next?

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

vf_opencv when?

"60fps" videos on mpv?

Works fine for me?

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

I meant software tweaking like SVP

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

Exactly

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

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

MOUSE_BTN3 add volume 2
MOUSE_BTN4 add volume -2

i prefer this:
MOUSE_BTN3 add volume -2
MOUSE_BTN4 add volume 2

makes a lot more sense if you use smartphones etc.

>if you use smartphones etc.
kill yourself

...

Probably...

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

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?

>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?

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.

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 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.

higher profile - less distortions

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?

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"?

>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.

>//!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.

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?

>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?

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.

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.
github.com/mpv-player/mpv/blob/master/TOOLS/lua/autocrop.lua

>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.

>vo=direct3d_shaders

The fuck are you smoking?

bump

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

>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.

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

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

scam sell your chink trash to some other Cred Forums autist and get something better

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

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

By being sharper than ewa_lanczos.

By how much?

About 3 sharp units.

TIL: 3 sharp units are completely unnoticeable.

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?

because it's shit and unoptomized

his dithering implementations are completely mongoloid

>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.

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

Sounds like an issue with the NNEDI3 script you're using

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

>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

how many windows only programs share their source code?

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

By exactly 1.9107694051635482%

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

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

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

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

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

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

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

Well it works fine for me

What are you even trying to do?

>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.

...

>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.

>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

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

>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.

N-no one? pls

>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.

Oh wait, I think I know what might be happening.

You might be hitting this line of code:
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

>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.

>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.

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

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.

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.

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.

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

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.

>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

>???
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.

>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

>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?

>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

>So you agree that they're the same quality?
See >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.

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

Left looks better

t. robot eyes

>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

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

>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.

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

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

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

github.com/Argon-/mpv-stats

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

thanks!

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.

>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)

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

>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.

>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.

>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.

>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.

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

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.

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.

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

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?