Install mpv: https://mpv.io/installation/ Wiki: https://git

Hobbyist player for a hobbyist OS.

>>61217504
>>>/v/

>>61217522
You know I'm right.

>>61217724
Nope. I use Linux at work.

The only thing that keeps me on Win now is games and foobar. Dont care about madVR anymore. Thanks mpv devs!

What is vulkan and how do I use/install it

>>61217805
Its something mpv devs stole from Madshi!

>>61217834
Disgusting.

Is foozoor a good guy or a bad guy? Ive seen him praising mpv once.

>>61217834
Madshi is a directX shill.

>>61218024
i think he was joking. as in "the mpv devs stole linux from madshi!".

How do you use lavfi i'm trying to use the now depreciated pan?

File: 2017-07-04_19-46-17.png (6KB, 293x121px) Image search: [Google]

6KB, 293x121px

I'm trying out chromium instead of firefox but I can't find an extension that lets me open mpv with just a right click on any link like in pic related.

What do you chromium bros use?

>>61218197
chromium based browsers cant do that
either use the terminal or drag and drop urls
or keep using firefox like a normal person

>>61218238
>chromium based browsers cant do that
What? Why not?

There's tons of extensions like "Open with Firefox" etc, why couldn't they open mpv similarly?

>>61218197
https://github.com/agiz/youtube-mpv

>>61217504
As opposed to the profesionnal anime watcher that choses the patrician player, madvr.

>>61219025
mpv is literally created by anime lovers. madshi doesnt give a fuck about anime.

>>61218900
>having to run a server for something so basic
Absolutely disgusting, I might actually vomit.

File: Capture.png (23KB, 383x82px) Image search: [Google]

23KB, 383x82px

Dead.

File: Screenshot_2017-07-04-14-05-23-1.png (106KB, 1257x605px) Image search: [Google]

106KB, 1257x605px

>>61219109
Alive.

>>61219109
No, faggot
https://sourceforge.net/projects/mpv-player-windows/files/64bit/

>>61219109
lachs0r's just making a build everytime there is an official release. I think there should be one soon.

>>61219243
why does the stats.lua graphs show on lachs0rs build but not shinchiros newest build?

>>61218480
>There's tons of extensions like "Open with Firefox" etc, why couldn't they open mpv similarly?
https://addons.mozilla.org/en-US/firefox/addon/watch-with-mpv/

is it possible to watch .gif with mpv? i saw a frame, but they are always frozen

>>61219282
Works for me.

Can someone explain how 10-bit depth achieves smaller encodes with apparently better perceived quality. I'm reading that one github explanation, but can't internalize what exactly happens.
It says that LCD displays can only display up to 6-bit depth and use dithering to represent higher bit depths. What exactly happens when passing colour data to the device? Is pixel data passed to the device and the device decides on dithering? Or does dithering have to be hardcoded? Does 8-bit depth get truncated to 6-bit depth for the display?

Why would 10-bit depth not require hardcoded dithering (I assume the bitrate loss from ditching dithering leads to the smaller size here) and 8-bit releases require it? Wouldn't a 10-bit depth also be truncated to 6-bit depth? How does this magic actually work?

>>61219364
the actual graphs that move in realtime?
I'm using hwdec=d3d11va btw

>>61219282
Update stats.lua

>>61219409
You need to download stats.lua
To activate it you press shift+i, to see more stats you press "2".

>>61219388
>LCD displays can only display up to 6-bit depth
This hasn't been true for a long time. Almost all modern panels are 8bit, though some do use different color profiles which can limit their overall number of colors.

The 10bit encode being smaller than an 8bit encode is mostly bullshit, as you'd have to be retarded to apply filters at the encode point. What I mean is that a dithered 8bit gradient IS more data than a non dithered 10bit gradient, but that should never happen, as you should be dithering the 8bit encode ONLY during playback.

>colors truncated
Colors are mapped between spaces in a lookup table.

>>61219569
I will need to do my own comparison I suppose to figure it out, but the github article says that when both encodes are encoded with same settings, both dithered
>The video track of the 8-bit encode has 275 MiB, while the 10-bit encode has no
more than 152 MiB and doesn’t look worse at all
What's the deal here? If both are dithered and 10-bit encode requires extra bit precision, where do the space savings come from?

Similarly, if 10-bit encodes do not reduce the file size, what is the point of using them? Reduced banding? Does the encoder somehow deduce/simulate the extra color information required to represent color gradients better?

>Colors are mapped between spaces in a lookup table.
Any technical term for this so that I can read up more? Does the device color buffer (or whatever you call it) store color values as 10-bit, even when only 8-bit depth can be represented? Do the values in lookup table not get truncated to make them suitable for display?

>>61219569
TN panels are limited to 6-bit.
Modern TN panels are 6+2-bit(6-bit panel with 8-bit dither).
IPS panels can be 8 and 10-bit, some 4k monitors that advertise 10-bit color depth are just 8+2-bit(8-bit ips panels with 10-bit dither).

>>61219774
>Does the encoder somehow deduce/simulate the extra color information required to represent color gradients better?
The encoder can apply filters (debanding, etc) in 16-bit before dithering to 10-bit.
But even without filters a 10-bit encode from an 8-bit source is preferable because 8-bit encodes are prone to worsening source banding or outright introducing it, especially at low bitrates.

i want high quality and smooth experience, or just smooth experience

give me the best you've got

>>61220264
For a "smooth experience"
interpolation
video-sync=display-resample
tscale=oversample
tscale-radius=3.0

>>61220293
>tscale-radius=3.0
?

>>61220306
It provides better interpolation quality, I forgot to mention that I have my monitor overclocked to 72hz.
tscale is limited to a radius of 3

how many hours a day do you dwellers spend on configuring mpv?

>>61220375
What do you mean by "better" ? Legit curious.

>>61220404
My config is always open.

>>61220411
oversample is too blurry.
tscale-radius=3.0 makes it a bit sharper.

>>61220421
mine too

>>61220441
>tscale-radius=3.0 makes it a bit sharper.
Any tradeoffs?
>>61220404
Almost 20.

opengl-shader=1.glsl
opengl-shader=2.glsl
opengl-shader=3.glsl

Who the fuck though this was a good idea?

>>61220503
https://github.com/mpv-player/mpv/commit/e4bc563fd2dcff1059624efb7b948b2886a382ab

>>61220503
I think its good but it seems that they reverted this already.

>>61220465
None.

This is my full scaling conf:
# Scaling
scale=ewa_lanczossharp
cscale=ewa_lanczossharp
dscale=mitchell
tscale=oversample
tscale-radius=3.0
correct-downscaling
sigmoid-upscaling
sigmoid-slope=10.0
linear-scaling

>>61220527
Where is video-sync=display-resample ?

>>61220527

dscale=mitchell 
correct-downscaling
sigmoid-upscaling
linear-scaling

If you use profile=opengl-hq you can delete these.

>>61217400
>Test vulkan and post logs if it gives you any kind of problems.
I did and resizing crashes on x11 and memeland is not ready for the desktop.

>no build for 2+ months
Did they at least fix stuttering on macOS?

>>61220527
In the # Other section

How do I bind seeking to start of the file?
Tried

set start 0

but it returns an error.

>>61220618
Meant for >>61220560

>>61220210
Where does this extra color information come from? In case of anime BDMV, aren't most of them 8-bit?

>>61220626
seek 0 absolute

>>61220527
>sigmoid-slope=10.0
cargo cult meme

>>61220713
All the anime BDMV are 8-bit, and color information come from the filters.

>>61220527
post everything related you've got to video quality

>>61220733
Test it you dumb fuck.

>>61220733
>cargo cult meme
What's wrong with it?

>>61220785
Test for what? You think it improves quality? Post proof then. sigmoid-slope=10.0 is a meme.

>>61220718
Doesn't work. But fuck it, it is easier to seek minus 99 hours so I just did that.

>>61220840
what about

set playback-time 0

>>61220778
# Video
vo=opengl
profile=opengl-hq
deband=yes
deband-iterations=2
deband-range=12
deband-grain=0

# Scaling
scale=ewa_lanczossharp
cscale=ewa_lanczossharp
dscale=mitchell
tscale=oversample
tscale-radius=3.0
correct-downscaling
sigmoid-upscaling
sigmoid-slope=10.0
linear-scaling

# Other
opengl-backend=angle
angle-max-frame-latency=1 #Lower values should make VSync timing more accurate.
hwdec=dxva2-copy
vd-lavc-threads=16
ad-lavc-threads=16
interpolation
video-sync=display-resample

>>61220845
That one works. Thanks.

>>61220867
*I don't use any opengl-shaders and I use "vd-lavc-threads=16" because I have a Ryzen 7 1700

>>61220840
I just tried

n seek 0 absolute

and it works for me.

Any way to compare 8-bit and 10-bit encodes using avisynth with something like Interleave? Would the depth of one encode get converted to another depth or can avisynth keep them separate?

>>61221086
>avisynth
avisynth doesn't have real 10-bit support, use avisynth+ or vapoursynth

>>61220590
Post logs and try again with --vulkan-direct=no

>>61219774
>Any technical term for this so that I can read up more? Does the device color buffer (or whatever you call it) store color values as 10-bit, even when only 8-bit depth can be represented? Do the values in lookup table not get truncated to make them suitable for display?

A color LUT / colormap; https://en.wikipedia.org/wiki/Lookup_table#Lookup_tables_in_image_processing

>What's the deal here?

>>61220133 mostly explained it, the caveat is that this is only valid when you have low bitrates and dither at the encode point. Arguably, the CORRECT way to do video encode / playback is to apply filters* (such as debanding) only at playback, thus avoiding this situation entirely and making 10bit color mostly a meme.

*some filtering and alterations are better applied at / before encode, such as color correction, degrain, deghost or other alterations to "correct" the source material. To be clear though, this doesn't include visual enhancements like debanding or antiringing.

>>61221603

2nd ref should point to >>61220210

Can someone explain ign's SSIM to me please? In madvr you could chose SSIM as a downscaler but in mpv SSIM is used after downscaler? So ign's version is not a downscaler but something like enhancer to downscaler?

Are igv's shaders working with this plexmediaplayer macosx release?
https://nightlies.plex.tv/public-test/plexmediaplayer/public-testing/latest/

OMG! It seems haasn found a way to make mpv really faster!

><hanna> and now I've gotten scale and cscale efficient enough to the point where I can stop using auto-profiles and apply my full on meme placebo scalers to all resolutions \o/

>>61222100
How? Vulkan?

>>61222116
https://github.com/mpv-player/mpv/pull/4575

>>61222100
>>61222116
Nah, but it probably means: vulkan is even more useless now.

File: 1477890417973.jpg (63KB, 484x490px) Image search: [Google]

63KB, 484x490px

>>61222125

>>61219388
>Can someone explain how 10-bit depth achieves smaller encodes with apparently better perceived quality.
More efficient encoding of gradients, because you can just encode a 10-bit gradient without noise whereas for 8-bit you need to dither

>>61219388
>Why would 10-bit depth not require hardcoded dithering
Because you can dither during playback

>>61222100
So, he'll just use haansblur for everything now?

>>61222187
>>61222201
Why do you need dither 8-bit but not 10-bit if the display itself cannot display the bit depth? Wouldn't you need to dither both at playback to simulate the extra bit depth that display cannot handle?

>>61222268
The display handles dithering from 8 to 6 bit (assuming it needs to)

>>61222134
Sure, if you don't care about cpu usage, gpu usage, you bank account, not wanting desync when you alt+tab, whether you run free software, are a slave of the gpu mafia and whether or not throwing more power at an issue doesn't solve your problems anymore. That's a lot to not care about.

>>61222275
Why do you need to dither the 8-bit encode then as >>61222187 stated if it is implicitly handled by the display? In fact, wouldn't you need to dither the 10-bit instead of the 8-bit encode to simulate the 10-bit depth?

Are there any drag and drop based playlist plugins. mpv is literally shit for watching a series because of lack of manual playlist support.

>>61222318
click the user scripts link in the OP

>>61220812
>Test for what? You think it improves quality? Post proof then. sigmoid-slope=10.0 is a meme.
It had very slight improvements or more like less artifacts as shown in various comparisons in many old threats.
Granted, I don't really see the difference without eye balling an image. However, it doesn't cost anything, so why not?

>>61222312
because 8 bits are insufficient to cover the range of human vision, we need at least 10 bits (standard gamut) or 12 bits (HDR / wide gamut).

if you try encoding a gradient 8 bits undithered, it will look like shit (banding), so you need to compensate by adding more dithering/noise (expensive to encode)

but you can just encode a gradient as 10 bits for free because the rounding artifacts are within the limits of human perception

>>61222100
>>61222125
What about Intel?

>>61222300
>not wanting desync when you alt+tab
works fine for me with opengl

honestly the most important improvement vulkan would bring with it is the ability to be cross-platform, which opengl can't really satisfy (because microsoft hates it)

>>61222493
Try it and find out?

<atomnuker> because if its slower on intel or amd who gives a fuck, these poor subhuman bastards can't even run the supreme upscalers out there
<atomnuker> but hey, memcpys are free, right, lets have more of 'em
<atomnuker> some guy wrote some amazing assembly for in-place intraprediction? bah, gotta find more places to memcpy, my cpu is getting cold and my electricity bill is down a percent
<hanna> atomnuker: shitposting is free, writing code is not
<hanna> either contribute improvements or shut the fuck up
<hanna> I'm sick of yo
<hanna> you
<atomnuker> well I'm sick of you too
<atomnuker> fyi I wrote over 2000 lines of code for vo_vulkan and I do not enjoy constant reminders of how I should just give it up

What's this beef about? Is the new code really slower for AMD/Intel?

File: 1488384160297.png (312KB, 714x574px) Image search: [Google]

312KB, 714x574px

>>61222546
EVERYONE STOP FIGHTING!!!!!!!!!

>>61222590
Nah that's always good for foss, atomnuker should create his own optimized vulkan/linux fork to beat the madvr supremacy.

>>61222546
>Ps. if anybody is reading this, feel free to send me a recent-enough AMD GPU so I can optimize scaling for AMD as well (not just nvidia) :^)

>>61222462
Does that mean most 8-bit encodes are encoded with dithering (implicitly) or do specific settings have to be enabled for dithering?
>>61219569 says that 8-bit encodes should be dithered only at playback, what's the point of dithering at encode time apart from bloating the size?

>>61222696
Yeah but he also claims 10 bits encodes don't save size, which is demonstrably and empirically false and has been known for like a decade

File: mpv-shot0005.png (102KB, 704x480px) Image search: [Google]

102KB, 704x480px

>>Does that mean most 8-bit encodes are encoded with dithering (implicitly)
Yes, otherwise all gradients would look like complete shit. This is what 8 bit looks like if undithered

All 8-bit encodes have to be aggressively dithered to prevent this from happening

>>61222546
Same speed as before.

>>61222766
So if 8-bit encodes are meant to be dithered at playback (no encoded dithering) and require less precision bits for color depth, where does the size saving come from with 10-bit encodes?

File: 1462686107858.jpg (21KB, 640x432px) Image search: [Google]

21KB, 640x432px

>>61222546
So this is the power of the mpv dev team

>>61223089
It's all his >>61222134 fault.

>>61222891
Did you misread my post? I'm stating the exact opposite

any good GTX1070 windows 10 configs?
any improvement on my current

##Video##
profile=opengl-hq
scale=ewa_lanczossharp
cscale=ewa_lanczossharp
dscale=mitchell
correct-downscaling=yes
dither-depth=auto

##Deband##
deband=yes
deband-iterations=2
deband-range=12
deband-threshold=48

##Interpolation##
interpolation=yes
video-sync=display-resample
tscale=oversample

##Other##
opengl-backend=dxinterop
hwdec=no
vd-lavc-threads=4

[upscale]
profile-desc=cond:get('width',0) <= 1366
sigmoid-upscaling=yes
sigmoid-slope=10.0
opengl-shaders="~~/shaders/SSIMSuperRes.glsl,~~/shaders/KrigBilateral.glsl,~~/shaders/enhance_detail.glsl,~~/shaders/anti-bloating.glsl"

[downscale]
profile-desc=cond:get('width',0) >= 1440
opengl-shaders="~~/shaders/SSimDownscaler.glsl"

>>61222796
Does AVC 8-bit encoding enforce that by default or are there any flags that need to be enabled (or are you meant to dither with external filters)?

>>61223489
Does that mean his assumption of 8-bit encodes being dithered at playback is wrong, if presumably 10 bit encodes save space by reducing dithering bit rate?

>>61222766
You're retarded.

>>61223637
>Does AVC 8-bit encoding enforce that by default
No, it's completely up to the encoder.

>Does that mean his assumption of 8-bit encodes being dithered at playback is wrong
Not necessarily. “Dithering” is a repeatable operation, and usually you have to do it before every quantization step. In a typical playback chain there might be three quantization steps: 1. inside the encoder (e.g. 10->8), 2. inside the video player/driver (e.g. 16->8), 3. inside the display (e.g. 8->6)

>>61223667
I think (16->8) inside the encoder as well is more likely, either during mastering (assuming the studio masters with 16-bit content) or after filtering (assuming you degrain/smooth the source).

So to put it another way, 16->10 quantization doesn't necessarily require dithering, whereas 16->8 always does.

>>61223667
>>61223714
Dithering only needs to be applied in the output filter chain, not at every step along the way.

>>61223637
The only reason to dither at encode time is if your target viewer / player isn't able / configured to apply filters or do dithering (probably most people).

>Does AVC 8-bit encoding enforce that by default
No.

>>61223750
>The only reason to dither at encode time is if your target viewer / player isn't able / configured to apply filters or do dithering (probably most people).

That's why in >>61222891 I asked how 10bit encodes can save more file size compared to 8bit encodes, if 8bit encodes that were not dithered at encode time do not come with the extra bit rate dithering overhead and require less space to store color depth information (8 compared to 10).
Unless that post was stating otherwise from what I am assuming here, in which case I am just retarded in reading comprehension.

>>61223667
>Dithering only needs to be applied in the output filter chain, not at every step along the way.
Wrong. Dithering needs to be applied before every quantization step, unless you use an intermediate bit depth high enough (at least 10 bit, ideally 16 bit) to avoid banding.

This is exactly why all filter processing is usually done in 16 bit, so you only need to dither at the very end. This is also why people use 10-bit encodes.

I usually use mpv on gnu/linux, but I have to use windows for a while. What windows builds are best, and how do I install them properly?

>>61224276
Drop the files into a folder, run the executable.

>>61223937
>if 8bit encodes that were not dithered at encode time
don't really exist

Also I think 10-bit is more efficient for gradients than 8-bit even if you don't dither, because encoding a linear function is cheaper than encoding a stairstep, but I could be wrong about that

>>61223937
>>61224318
10-bit encoding is also more efficient than 8-bit encoding even ignoring gradients and dithering, because 10-bit encoding also means 10-bit precision for motion compensation and other filtering steps, which are more accurate than 8-bit. This means that the accumulated error due to in-loop deblocking etc. is smaller, resulting in higher quality. Not to mention, 10-bit prediction vectors allow you to more accurately find similar blocks, further reducing the amount of error you have to encode.

Furthermore, 10-bit actually can never be “less” efficient than 8-bit, because all you have to do to recover the overhead is quantize more, thus resulting in the same quality level as before.

So really, dithering is a red herring. 10-bit encoding is just strictly better than 8-bit encoding, regardless of what your 8-bit content looks like. Better representation of undithered gradients are just an added bonus.

>>61224392
Put another way, 10-bit encodes trade processing time (in the encoder/decoder) for bitrate reduction. Which is exactly what we saw in practice: 10-bit decodes were smaller but more difficult to play back.

>>61221719
>So ign's version is not a downscaler but something like enhancer to downscaler?
Most likely, what's going on is that the filter named simply “SSIM” just has the upscaling filter hard-coded into it. (probably Lanczos or Jinc)

The design philosophy of the mpv developers seems to be trying to make everything as flexible and independent as possible, instead of hard-coding specific associations, and this probably extends to igv's SSIM shader design.

>>61223646
>You're retarded.
I'm sure you think the developers of x264 are retarded as well

x264.nl/x264/10bit_02-ateme-why_does_10bit_save_bandwidth.pdf

>>61221719
SSIMDownscaler is just a sharpener (in madvr too), it doesn't remove artifacts like SSimSuperRes.
You can call it SSIMMitchell.

you guys got any build of ffmpeg with libfdk_aac for windows? I don't want to compile that shit now.

File: 1473794453120.jpg (115KB, 1280x720px) Image search: [Google]

115KB, 1280x720px

Which of these two images do you prefer?
Both are 720p upscaled to 4k....

Image 1: https://my.mixtape.moe/eqqyea.png

Image 2: https://my.mixtape.moe/rnuldz.png

Original is here.

>>61226348
Image 2 is objectively better.

>>61226348
2nd is a less blurry, but not super noticeable.

>>61226348
The second one is upscaled with NGU, Can see it from a mile away.

>>61226400
>Image 2 is objectively better.
I agree.
>>61226440
>2nd is a less blurry, but not super noticeable.
Right. The whole anime is fucking soft for some reason.
>>61226502
Nope. it's mpv... with a bunch of shaders.

61197961

>>61197961

Two questions:
1. If I use adaptive-antiringing.glsl can I remove the scale-antiring from the config?
2. What are good values for tscale-radius and tscale-clamp?

Thanks.

>>61226655
1. Yes

>>61226519
Post conf nigga.

>>61226936
Probably >>61197987

>>61220609
On OSX but not building with homebrew? What kind of degenerate faggot are you?

>>61225357
That only applies if the source is 10bit, because quality is measured relative to the source.

All of my points stand.

>>61221719
>In madvr you could chose SSIM as a downscaler
In madvr 1D is Catmull-Rom and 2D is Jinc
mitchell or gaussian is better tho.

>>61227446
>All of my points stand.
None of your points stand.
Go read an encoding guide and comeback here retard.

>>61227484
Wait, are you sure? Ive never heard this info and ive been lurking doom9 and used madvr for years.

>>61227686
https://forum.doom9.org/showthread.php?p=1757360#post1757360
>"SSIM 1D" is the same as "SSIM" in v0.90.9 (just a bit faster, I hope). It's based on a Bicubic kernel. "SSIM 2D" is based on a Jinc kernel, so it's rather slow.

1D has some ringing, so it can't be mitchell. It's either Catmull-Rom or Bicubic60.

>>61227745
Thanks man. I always wondered why 2D is so much slower.

>>61226348
>.jpg
Nigger stop. Dont use fucking jpg to compare scalers.

I'm on windows and I am having trouble with youtube-dl. Any link I run through it fails on the reason of failure to recognize file format.

File: muh dick.jpg (557KB, 1920x1080px) Image search: [Google]

557KB, 1920x1080px

This new stats thingy is soo goood. Makes testing things so much easier. Thank you based haasn and Argon!

What was the setting to make shaders not affect subtitles? Adaptive-sharpen adds way too much ringing to them

>>61228148
--blend-subtitles=video / no

>>61227960
Any help on this? Searching sort of gives me a dead end.

>>61226348
I prefer the one in the middle, she has the sexiest eyes

>>61228173
Thanks

File: pajeet_cant_into_reading.png (171KB, 1055x1414px) Image search: [Google]

171KB, 1055x1414px

>>61227446
can't even make this shit up anymore

https://addons.mozilla.org/en-US/firefox/addon/watch-with-mpv/
Make sure you have youtube-dl installed

>>61227994
>unknown unknown unknown
get these meme shader authors to add descriptions for their passes

>>61217400
is this the best media player

>>61228252
I like it.

>>61227960
Bump please help, I can't goto bed unless I some how fix this

>>61227960
>I'm on windows
>>61228294
>I can't goto bed unless I some how fix this
I have a solution but I have a feeling you will not accept it as valid.

>>61228357
>yay more memes.

I'll just goto bed then. Its like 430am

>>61227994
How do I access that? Is it not part of stats.lua?

>>61228388
https://github.com/Argon-/mpv-stats/commit/96916139d27c74c542d3372240bee738bada0d2d

>>61228406
Finally i'll be able to compare angle and dxinterop

>>61228425
Don't forget to set your gpu to highest performance first.

>>61228425
post results