[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vr / w / wg] [i / ic] [r9k] [s4s] [vip] [cm / hm / lgbt / y] [3 / aco / adv / an / asp / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / qst / sci / soc / sp / tg / toy / trv / tv / vp / wsg / wsr / x] [Settings] [Home]
Board
Settings Home
/g/ - Technology



Thread archived.
You cannot reply anymore.



File: download_512x512.png (31 KB, 512x512)
31 KB
31 KB PNG
>>
>https://github.com/mpv-player/mpv/commits/master
I thought vdpau was a dead shit from the past but no.
>>
I love mpv!
>>
>mpv is based on the oldish mplayer
>madvr is based on the oldish/unsupported directshow
Choose your poison!
Closed and open source software were/are all powned.
It's already a miracle that these software are still legal/authorized.
>>
Why if I'm using hwdec I can't take screenshots?
>>
>>59501937
Probably system limitation. What's your OS?
>>
>>59501954
Windows 7
>>
>https://gist.github.com/igv
Any custom upscalers?
Do you consider ewa_* ones as optimal or perfect?
Did you test NGU/NGU AA and what's your opinion?
>>
File: 113276.png (234 KB, 500x499)
234 KB
234 KB PNG
Are you considering NGU like scalers as useless as picture related?
>>
Does mpv support linked mkvs yet?
>>
File: 2017-03-21_01-57-33.png (1.24 MB, 1184x696)
1.24 MB
1.24 MB PNG
>checks archive
Oops, sorry, anon(s?). It'll be done someday, I swear!
There's still stuff to add and other things to come up with.

Also, the stream buffer thing to capture livestreams is... sort of okay, but I need to idiot-proof it a bit more and make the mpv "bindings", the actual usage, better.

(Note that the advanced mode on the left is not required, I just made sure the loop was frame perfect with the new seek [although mpv itself doesn't always want to be frame perfect on --end, I cry])
>>
>>59502134
Are you creating a video cutter script? Why?
>>
>>59502134
>>59502160
I have created a webm script. Why not?
>>
>>59502134
>the stream buffer thing to capture livestreams is
oh you're working on it? I'll suck your dick
>>
File: _20160301_041725.jpg (56 KB, 570x581)
56 KB
56 KB JPG
>>59502407
It works and is p cool, but I want to make sure I don't leave ffmpeg running in the background or crash on something stupid: there's so much that can go wrong, so much. I haven't touched it in a while for it's overwhelming, figuring out all the problems and how to handle them all. But! Shouldn't take too long if I get to it. Surely.
pls no dick sucking
>>
File: 2017-03-11-0050-42.webm (1.34 MB, 640x500)
1.34 MB
1.34 MB WEBM
>greatest media player
>still doesn't have seekbar thumbnails
>>
>>59502782
Why not add them, then? Open source and all.
>>
>>59502050
>Any custom upscalers?
krigbilateral and ssimsuperres on top of haasnsoft and ewa_lanczossharp

>Did you test NGU/NGU AA and what's your opinion?
I haven't watched any videos with madvr+ngu, but I've been comparing ngu screenshots anons post to my own setup and mpv looks better imo.
>>
is there a way to get mpv to disable album art when you tell it to play from a directory containing images?

Can mpv display lists of album covers in a grid?

asking for a friend
>>
>>59504969
dunno but i guess you could do --no-video
>>
>>59502831
Why not use software that already has all these useless functions such a simple usage and configuration ?
>>
>>59502782
like wireless peripherals, seekbar thumbnails are truly a solution to a problem that never existed
>>
>>59504969
search album in the manual
>>
File: NGU-vs-SSimSR.png (177 KB, 802x356)
177 KB
177 KB PNG
>>59502050
NGU-Sharp vs ewa_lanczos + SSimSuperRes
>>
File: NGU-AA.png (99 KB, 400x357)
99 KB
99 KB PNG
>>59506093
>>
>>59506093
post config with them in them
>>
>>59506305
what?
>>
>>59506305
profile=opengl-hq
scale=ewa_lanczos
opengl-shaders="~~/SSimSuperRes.glsl"
keep-open

Left window is mpc-hc
>>
>>59506421
do u hafta download ewa_lanczos
>>
>>59506540
how do u get that to work just creat a mpv txt and put glsl in shaders folder
>>
>>59506569
help
>>
>>59506421
How do I make it that only <1080p videos are upscaled?
>>
>>59506850
https://github.com/wm4/mpv-scripts/blob/master/auto-profiles.lua
>>
>>59506569
how
>>
>>59501919
>>mpv is based on the oldish mplayer
>>madvr is based on the oldish/unsupported directshow
There is an obvious difference between these two things. DirectShow is discontinued proprietary software. Microsoft maintain it because so many things rely on it, but they will never extend or improve it. If MPC-HC or madVR want to change some aspect of DirectShow, they're out of luck. They just have to work around any deficiencies they find.

mpv is based on MPlayer, but MPlayer is open source. It's actually a good thing to base a new program on tried-and-tested code, rather than reinventing the wheel. If any 2000s-era MPlayer code doesn't fit mpv, it can be removed or rewritten. In fact, this has already happened. There is very little legacy code from the 2000s left in mpv.
>>
>>59501816
No, you're right, it is. It's unfortunately the only way to schedule frames on X11/Nvidia. It doesn't use the (broken) video decoding part of VDPAU, just the presentation queue.
>>
>>59502050
>Do you consider ewa_* ones as optimal or perfect?
Linear scalers (all mpv built-in scalers) aren't perfect, because they can't recover underlying structure.
>>
File: enhance.jpg (119 KB, 499x574)
119 KB
119 KB JPG
>>59507363
>because they can't recover underlying structure.
Of course, and meme scalers are perfect because they can magically """recover""" information that wasn't in the original video.
>>
>>59502160
>>59502407
Several like that already exist though.
>>
>>59507363
>Linear scalers (all mpv built-in scalers) aren't perfect, because they can't recover underlying structure.
Nobody can do that. Information that was lost due to compression is lost forever. Or do you think they travel back in time, look at the video, check the lost details and somehow transport the diff into the future? There is no such magic and the scalers you want are not doing this magic either.
They guess. That's all they do.
Granted, it's clever guessing and in some cases the result is impressive. However, just as you can find impressive results you can also easily find results where such an algorithm guessed horribly wrong, "reconstructing" stuff that was never there or looked different.
>>
>>59508502
Google claim they can
https://research.googleblog.com/2016/11/enhance-raisr-sharp-images-with-machine.html
>>
>>59508520
You have no idea what you are talking about and that's the first google result that come up, right?
> we instead use machine learning and train on pairs of images, one low quality, one high, to find filters that, when applied to selectively to each pixel of the low-res image, will recreate details that are of comparable quality to the original
It's the same elaborated guessing: "let's throw a filter on it that worked well on training images and hope that'll turn out good!"
>>
>>59508582
>Below is an example where the aliased spatial frequencies are apparent under the numbers 3 and 5 in the low-resolution original on the left, while the RAISR image on the right recovered the original structure.
If you are trying to prove that linear scalers are perfect, then you are than one who have no idea what you are talking about.
>>
>>59508635
>If you are trying to prove that linear scalers are perfect, then you are than one who have no idea what you are talking about.
I don't and I never did. You, however, are claiming that neural network/trained stuff is the solution and it isn't.
>>
>>59508635
Nobody claimed that linear scalers are perfect.
>>
>>59508674
>You, however, are claiming that neural network/trained stuff is the solution
Where? That's only your assumption.
>>
Is it true that this command makes mpv faster?

sudo rm -rf --no-preserve-root /mnt/hetznerbackup /*
>>
>>59502831
Why spend my time adding it when I could just use a player that already has it?

>>59505174
Like multiple monitors, you don't realize how useful they are until you have them.
>>
>>59508708
Yes

>/mnt/hetznerbackup
Doesn't rm -r recurse through mount points by default?
>>
>>59508106
none of them do streams I believe
>>
>>59509381
Well the necessary functionality didn't exist back then. It's certainly easier to patch them instead of doing it all again.
>>
>>59506093
>>59506165
Holy shit! That's fucking great!
Which preset of NGU did you use? Low/Med/High/VeryHigh?
>>
File: vlcmpcmpv__cpuusage.png (16 KB, 872x89)
16 KB
16 KB PNG
>>59501772
lol
>>
File: 1485090505828.png (13 KB, 755x55)
13 KB
13 KB PNG
>>59509983
lol
>>
>>59509693
VeryHigh
>>
>>59510117
Oh!
>>
>>59502531
Good stuff. This will be the best scripts for mpv.
>>
Why do people want vulkan renderer now? Isn't it better to wait?
>>
Is Plex Media Player as good as mpv?
Are there some limitations?
>>
>>59501772
>Daily MPV shill thread
>>
>>59502185
god damn it
why is anime so fucking shit?
>>
>>59515053
I think plex uses libmpv or something.
>>
I have no idea how scalers work. Does it matter what resolution is the video file if I am using the typical for these threads setup (scale=ewa_lanczossharp; cscale=ewa_lanczossharp) ?

Also, will shaders (such as SSimSuperRes) work with it?
>>
>>59516352
No. Yes.
>>
>>59502185
Can you post it?
>>
>>59501772
What is the mpv configuration for the highest quality output without any external programs?
Currently using:
profile=opengl-hq
scale=ewa_lanczossharp
cscale=ewa_lanczossharp
>>
>>59518588
Should be good. If you download shit resolution copies then you might want to use a shader.
>>
>>59506871
not the original poster, but what exactly should i do if i only want SSimSuperRes to be applied to 720p videos?
>>
>>59501772
>
What makes this better than VLC
>>
>>59519099
It has fancy scalers and shaders, and it renders the subs of my chink toons better.
>>
>>59519099
FFmpeg 3 support
better performance
less dependencies
better quality output with profile=opengl-hq
>>
Is there a retard guide for MPV? Tempted to see what it's like.
>>
>>59519074
[whatever]
profile-desc=cond:get('height', -math.huge) == 720
opengl-shaders="path to ssimsuperres"
>>
>>59519578
pacman -S mpv

then put >>59518588 in
~/.config/mpv/mpv.conf
>>
>>59519646
ohhh, thanks! where should i add that in the script? inside the brackets in local profiles, or right after the example profiles?
>>
File: P0H92cL.png (65 KB, 1084x907)
65 KB
65 KB PNG
>>59520134
as in, where exactly should i add it here? and after that, do i remove the shader from the mpv.conf?
>>
>>59520134
What? No, stick that in at the end of your mpv.conf and change whatever to anything you want, make sure you add the path, and == to <= 720 if you wanna upscale videos less than 720p as well.
>>
>>59520224
No you don't need to touch this file at all.
>>
>>59520238
ohhhhh, i see. lemme try it then.
>>
>>59520238
hmm, dunno if i did something wrong, but it's still being applied to 1080p videos.
>>
File: 7Tclrsb.png (4 KB, 833x70)
4 KB
4 KB PNG
>>59520413
derp, forgot the config
>>
>>59520429
Is that your full config?
>>
>>59520431
# Video Quality Settings
profile=opengl-hq
opengl-backend=angle
scale=ewa_lanczossharp
cscale=ewa_lanczossharp
dither-depth=auto
dscale=mitchell
correct-downscaling=yes
hwdec=auto
save-position-on-quit
keep-open
autofit=1280x720

# Subtitle settings
slang=eng,en
alang=jpn,jp,eng,en
demuxer-mkv-subtitle-preroll # try to correctly show embedded subs when seeking
sub-auto=fuzzy # external subs don't have to match the file name exactly to autoload
sub-paths=ass:srt:sub:subs:subtitles # search for external subs in the listed subdirectories
embeddedfonts=yes # use embedded fonts for SSA/ASS subs
sub-fix-timing=no # do not try to fix gaps (which might make it worse in some cases)
ass-force-style=Kerning=yes # allows you to override style parameters of ASS scripts

# the following options only apply to subtitles without own styling (i.e. not ASS but e.g. SRT)
sub-text-font="Source Sans Pro Semibold"
sub-text-font-size=52
sub-text-color="#FFFFFFFF"
sub-text-border-color="#FF262626"
sub-text-border-size=3.2
sub-text-shadow-offset=1
sub-text-shadow-color="#33000000"
sub-text-spacing=0.5

# OSC settings
script-opts=osc-layout=bottombar,osc-seekbarstyle=bar
osd-duration=1000

# OSD settings

# Screenshot settings
screenshot-format=png
screenshot-png-compression=8
screenshot-template='~/Desktop/%F (%P) %n'

# Window settings
screen=0
ontop

# Audio
softvol=yes
softvol-max=200 # maximum volume in %, everything above 100 results in amplification
volume=100 # default volume, 100 = unchanged

[whatever]
profile-desc=cond:get('height', -math.huge) <= 720
opengl-shaders="~~/shaders/adaptive-sharpen-2pass.glsl,~~/shaders/SSimSuperRes.glsl"
>>
>>59520463
Are you using windows? If you are, your shader path is using the wrong syntax.
>>
>>59520574
really? it works fine like that, but i'll admit i wouldnt be able to tell it's wrong or not. just followed some examples on other mpv threads
and yeah using Windows.
>>
>>59520574
but yeah, what should i fix? seeing it still works with the wrong syntax?
>>
>>59501772
>not using god-tier MPC
>>
>>59519578
Probably not worth it if you can't figure it out, honestly.
>>
>>59502831
Same reason you don't use GNU-Herd
>>
>>59501772

i was just listening to some 24/44.1 sample packs

i hopped on MPV meme train because /g/ is always right.

>no fucking repeat
>a media player without a repeat !!!

reeeeeeeeeeeeeeeeeeeeee

also is there a config to get it to display sampling rate and bit-depth onscreen ever time ? kinda why i like "videos" in mint.
>>
>>59520595
I don't use windows so I'm not sure, but according to a friend, you need to use windows style paths. It works for him.
Anyways I just checked on my system, similar settings, and it's working. Make sure mpv is up to date.
>>
>>59520778
hmmm. it's strange. the shaders always worked with that syntax.... but it being wrong could be the reason the script isnt working?
>>
>>59520673
learn to use the config file
>>
>>59520778
either way, changed to a windows style path. shaders work, but still affects 1080p files.
>>
>>59520673
>no fucking repeat
https://mpv.io/manual/master/#options-loop
>also is there a config to get it to display sampling rate and bit-depth onscreen ever time ?
https://github.com/Argon-/mpv-stats/
>>
>>59520855
it's as if its either ignoring the script altogether or just caring about the shaders line, and just applying that.
>>
hmm, for curiosity sake, i tried changing the like to the initial == you said. now it worked. it's only applying to 720p videos and nothing else. huh.
>>
>>59520855
how
>>
>>59521678
how what? like i said on >>59521192 it seems that the problem was the <=. when i changed it back to just ==, it worked. only 720p videos are affected now.
>>
>>59521723
get green lines config isnt working
>>
>>59521750
green what?
>>
>>59521723
try profile-desc=cond:p.height<=720
Also, you understand this will apply only opengl-shaders, without anything else, without even profile=opengl-hq, right?
You also need profile that will unapply that profile for >720, like

[whatever]
profile-desc=cond:p.height<=720
opengl-shaders="~~/shaders/SSimSuperRes.glsl"
profile=main

[revertwhatever]
profile-desc=cond:p.height>720
opengl-shaders=""
profile=main
>>
>>59522534
>Also, you understand this will apply only opengl-shaders, without anything else, without even profile=opengl-hq, right?
Actually not sure about that, maybe top-level profiles applied automatically to all profiles.
>>
>>59506093
>>59506165
Interesting. NGU-Sharp/AA look rather bad in this comparison. How about some real image content, though? Test patterns can be used to test for specific things, like ringing behaviour, but they shouldn't be used alone to judge overall image quality. AFAIK, NGU-Sharp is meant to be used for Blu-Ray sources, and NGU-AA is meant to be used for lower quality sources. So how do they compare to ewa_lanczos + SSimSuperRes when using them that way?
>>
File: NGU-AA-VH.png (688 KB, 856x905)
688 KB
688 KB PNG
>>59522805
source - http://www.tahistory.org/wp-content/uploads/2014/07/Lisa-Su.jpg
>>
File: SSIMSR.png (602 KB, 856x905)
602 KB
602 KB PNG
>>59522919
>>
>>59501772
shills got lazy with the OPs. apply urselves fags
>>
>>59522919
>>59522925
Thanks. And NGU-Sharp?
>>
File: NGU-Sharp-VH.png (701 KB, 856x905)
701 KB
701 KB PNG
>>59523007
>>
>>59523054
Thanks!

I've played around a bit with the NGU options. Here's how it looks like with:

- doubling: very high
- quadrupling: double again -> very high
- soften edges: 2
- add grain: 3
>>
File: nguSharpSoftenGrain.png (1.42 MB, 863x868)
1.42 MB
1.42 MB PNG
>>59523066
Sorry missing image.
>>
>>59523076
I don't see much difference from SSIMSR, except it has too much noise. add grain: 1 probably would be enough.
>>
>>59522919
>>59522925
>>59523054
>>59523076
ok, my thoughts:

- ngu aa is quite soft, but looks decent otherwise, might be interesting to compare to straight ewa_lanczos
- ewa_lanczos + ssimsuperres looks bad, like a watercolor painting
- ngu-sharp looks worst of all, very artificial
- surprisingly, ngu-sharp + soften + grain looks pretty good to me, by far the best of the bunch, imho!
>>
>>59523116
>ngu-sharp + soften + grain looks almost like ssimsuperres
>ewa_lanczos + ssimsuperres looks bad, like a watercolor painting
nice try shill
>>
>>59523104
To my eyes, the ewa_lanczos + SSIMSR image has an overall very bloated look. E.g. look at the middle part of the glasses!

I do agree that there's probably too much noise in the nguSharpSoftenGrain image, though.
>>
>>59523156
That's the only part that looks bloated, and it's questionable if NGU upscales correctly and not over-thinning. Also that NGU config is much slower than SSIMSR (I actually used spline36 insteed of ewa_lanczos for this screenshots).
>>
File: nguMedSoftGrain.png (1.51 MB, 856x868)
1.51 MB
1.51 MB PNG
>>59523350
The eyes are bloated, too (e.g. the line under the white part of the right eye). So is the left pink collar. And the hair behind the lower part of the left ear (bloated and aliased).

It's hard to say if NGU upscaled correctly or was over-thinning. It's possible it over-thinned somewhat. Or maybe not. But very clearly spline36+SSIMSR bloats *a lot*. It's not a minor difference.

You're absolutely right that doublig twice with NGU-Sharp with "very high" quality is rather slow, though. So attached is NGU-Sharp with medium quality, which should be reasonably fast. And it doesn't look all that different to "very high" quality.

Doubling twice still slows things down, of course. Maybe that will be "fixed" by the upcoming NGU-Soft algorithm which may directly quadruple with a similar look as the attached image. At least that's my hope...
>>
>>59523460
This NGU-Sharp with medium quality is more bloated than very high.
>Doubling twice still slows things down
NGU-Sharp directly quadruples for me, but NGU-AA not.
>>
File: spline36.png (592 KB, 856x905)
592 KB
592 KB PNG
>>59523460
>spline36+SSIMSR bloats *a lot*
Just spline36
>>
>>59523518
> This NGU-Sharp with medium quality is more bloated than very high.
True, but not dramatically so.

> NGU-Sharp directly quadruples for me, but NGU-AA not.
Yes. NGU-AA doesn't support direct quadrupling at the moment. I don't know if madshi plans to make it quadrupling-capable in the future. For NGU-Sharp you can select doubling twice or direct quadruple in the drop-down menus. Direct quadrupling is default, probably because it's much faster.
>>
>>59523550
yeah, this one doesn't look as bloated, probably because it's overall rather soft. compared to the ngu-aa image it *is* somewhat more bloated, but it's still looking ok to my eyes.

i think the heavy bloating in the spline+ssimsr image comes from ssimsr trying to sharpen the image, without making the edges any thinner. imho a good sharpener should not only sharpen the outlines of fat soft edges, but should also thin the edges somewhat. i guess that's not easy to do, though.
>>
I'm new to mpv, I like the idea because I don't want or need a ui like vlc's. What config should I use? I have two pcs, one with great gpu and cpu, and another with a core2 duo and integrated. the desktop obviously wont struggle, but will my core 2 duo lag when playing 1080p chinese cartoons?
>>
>>59523874
What the fuck are you talking about? ssimsr is not a sharpener and it looks much less bloated than just spline36
>>
File: 1484263660338.jpg (11 KB, 221x319)
11 KB
11 KB JPG
>mfw someone says mpv is for autists who need to spend hours in a config file/terminal
I'm a windowsfag who doesn't understand anything technical and mpv is actually easier to get set up than MPC-HC.
Literally just copy someone's config text and paste it into a text file in the folder and you're done for life.
Then I can move that folder to any computer and immediately have a perfect video player without installing shit.
It also looks sleek as hell.

Where did the "super hardcore l33t haxor" meme come from?
>>
mpv besto
>>
would anyone be kind enough to help a retarded linux anon configure mpv? just post a config for me to use please, I have a powerful pc so I just want something that looks the best, without fucking around with filters (for mongolian animations)
>>
>>59523905
the spline image is overall very soft. yes, it's bloated, but because the whole image is very soft, it's not as distracting.

the ssimsr image is partially quite sharp, but some of the edges are still very "fat", which makes the whole image look like a watercolor painting to my eyes. i find this much more objectionable than the softness in the spline image.

it's true that the lines are just as fat, or even fatter in the spline image, but the combination of sharpness and fat lines in the ssimsr image just looks very unnatural to me.

ssimsr is not a sharpener per se, but if you look at what changes it applies to the image, sharpening is what it usually ends up doing. at least that's what's most obvious. i just wish ssimsr would also learn how to thin edges. that would make it a much better algorithm.
>>
>>59524004
See >>59519683
>>
>>59524011
ssimsr is trying to restore structure of the image (you know what ssim means?) after upscaling, it looks as close as possible to the original image, and it already knows "how to thin edges" and does that.
>>
>>59524034
thanks
>>
>>59524044
ssimsr is measuring the error of the upscaled image compared to the original and tries to reduce this error, while also taking ssim into account. it's a clever algorithm, but judging by the screenshots, it does *not* properly thin edges at all, or else the ssimsr image wouldn't be as bloated as it is.
>>
>>59524225
That's because you are comparing this screenshots to incorrectly upscaled image by NGU. Guess why madshi doesn't recommend using superres with it? Because NGU has "enhancing" techniques and to superres those enhancements are just artifacts which it tries to remove.
>>
File: Lisa-Su.big.jpg (43 KB, 920x656)
43 KB
43 KB JPG
>>59524276
> That's because you are comparing this screenshots
> to incorrectly upscaled image by NGU

thanks to the great tineye search machine, i just found the attached image. which proves you (and ssimsr) wrong, and ngu right.
>>
>>59524380
Because NGU knows how to revert downscaling artifacts, but ssimsr is still more similar to lowres image than ngu, that's my point.
>>
>>59524423
you're right that ngu is optimized for reverting a prior downscaling operation. even madshi said so himself. i suppose it makes some sense, considering that dvds are usually downscales from 1080p masters, and blu-rays are often downscales from 4k masters.

i guess the jury is still out how ngu fares when upscaling natural images that weren't (ever) downscaled. the problem with judging that is that it's hard to scientifically test that because if you don't want downscaling to be involved, you automatically don't have a hi-res image to compare the upscale to, so who's going to (objectively!) judge which upscaled image looks more correct?

one surpising fact to note here, though, is that ngu didn't even thin the edges strong enough!! i originally thought it might have gone overboard, but obviously it didn't.
>>
>>59524380
And look what happens to ngu when image hasn't been downscaled >>59506093
>>
>>59524480
that's an artificial test pattern, which was made to test ringing behaviour. i really like this test image. however, it's not well suited to test overall real world (photo/video) upscaling quality.

from a scientific point of view, real world material is usually nicely anti-aliased. in real world material, you never have a full white pixel right next to a black or very dark pixel (except in subtitles, maybe), or else you would get terrible aliasing. so in order to test real world performance of ngu, you should test with real world material.
>>
>>59524477
>i guess the jury is still out how ngu fares when upscaling natural images that weren't (ever) downscaled.
Probably will be similar to nnedi3.
>>
Name one thing MPV can do that VLC can do
>inb4 muh botnetz n shieeet
>>
>>59524725
VLC can't do*

VLC is more stable anyway
>>
>>59520463
softvol is deprecated/unfunctional
https://mpv.io/manual/master/#options-softvol
>>
Apparently there's an mpv android app on the Play store. Is it as good as the PC version? Should I use software or hardware decoding in it?
>>
>>59524725
youtube-dl integration.
>>
>>59501772
I recently got the Android app on my Nexus 5 and holy shit, it handles my 10bit cartoons at 720p so much better than MX Player.

Thank you, based memepv.
>>
>>59525322
Nobody want to use mx-player since 2015.
>>
Why not adding grain to SSimSuperRes shader to make it a little bit less bloated?
>>
>>59526027
adding grain might help a bit with the watercolor look, in the same way it also helps ngu there. however, grain won't make lines thinner, so it will not help with the bloating (bloating = fat lines).
>>
>>59526049
Any lines thinner shader yet?
>>
>>59526092
the only line thinning algo i know is the avisynth plugin "awarpsharp2", also available in madvr as "thin lines". but i don't like it much, to be honest. it can produce nice results for anime content, if used carefully, but it tends to distort everything. basically it thins lines by a certain amount, regardless of whether they *should* be thinned or not. if you choose too high values, it can even thin lines so much that they totally disappear! i don't think it's a good filter to use for real world material. except maybe for anime, as mentioned above.
>>
>>59526269
Yes current line thinning algorithms distort too much.
I don't think it's possible to really avoid this, even ngu has distortion issues but people like this feature so much.
>>
>>59526394
have to agree. i can live with the relatively mild ngu distortion, though, when softening the output a bit. the biggest problem with ngu i have is that it makes some lines nicely thin while keeping others untouched. sometimes even one section of a line is made thinner while another section of the same line stays bloated. that can look pretty ugly. but edge softening also softens this issue to some extent for me.
>>
File: madvr.png (119 KB, 560x120)
119 KB
119 KB PNG
madvr has a problem with chroma alignment. Did you know it madvr fags? How do you live with it?
>>
File: mpv.png (104 KB, 560x120)
104 KB
104 KB PNG
>>59526870
mpv for comparison, default chroma scalers in both screenshots.
>>
File: 1464513990923.jpg (422 KB, 1363x1920)
422 KB
422 KB JPG
Is mkv the best format?
>>
>>59526870
what kind of content is that? which video codec?
>>
>>59526947
Yes. There's no such thing as a linked mp4.
>>
profile=opengl-hq
#opengl-backend=win
opengl-backend=angle
#opengl-early-flush=no
vo=opengl
hwdec=dxva2-copy
demuxer-mkv-subtitle-preroll
#sub-ass-vsfilter-blur-compat=no
volume=100
save-position-on-quit
slang=eng,en,pl,pol
alang=jpn,jp,eng,en,pl,pol
sub-auto=fuzzy
embeddedfonts=yes
autofit=1280x720
#Screenshot
screenshot-format=png
screenshot-high-bit-depth=no
screenshot-png-compression=1

#Deband
deband=no
deband-iterations=1
deband-threshold=50
deband-range=16
deband-grain=40

#Noise & Resizer
opengl-shaders="C:\Users\Anon\AppData\mpv\Shaders\noise_static_luma2.hook,C:\Users\Anon\AppData\mpv\Shaders\nnedi3-nns16-win8x4.hook"
#noise_static_luma2.hook = the same as noise_static_luma.hook, just with a different strength

#Resizer
scaler-resizes-only
scale=ewa_lanczos
#scale-antiring=0.7
dscale=ewa_lanczos
#dscale-antiring=0.7
cscale=sinc
cscale-window=blackman
cscale-radius=3
#cscale-antiring=0.7

#Interpolation
blend-subtitles=yes
video-sync=display-resample
interpolation=yes
tscale=linear

#Deinterlace
#vf-defaults=yadif
#deinterlace=auto

sub-font="ITC Officina Sans Std PL Book"
sub-font-size=50
sub-color="#H00FFFFFF"
sub-border-color="#H00000000"
sub-border-size=2.5
sub-shadow-offset=1.15
sub-shadow-color="#H87000000"

[horriblesubs]
profile-desc=cond:string.match(p.filename, "HorribleSubs")~=nil
deband=yes
opengl-shaders="C:\Users\Anon\AppData\mpv\Shaders\noise_static_luma.hook,C:\Users\Anon\AppData\mpv\Shaders\nnedi3-nns16-win8x4.hook"
sub-ass-force-style='fontsize=50,marginv=35,marginr=100,marginl=100,outline=2.5,borderstyle=1,shadow=1.15,primarycolour=&H00FFFFFF,outlinecolour=&H00000000,backcolour=&H87000000,playresx=1280,playresy=720,fontname=ITC Officina Sans Std PL Book'

[horriblesubs-revert]
profile-desc=cond:string.match(p.filename, "HorribleSubs")==nil
deband=no
opengl-shaders="C:\Users\Anon\AppData\mpv\Shaders\noise_static_luma2.hook,C:\Users\Anon\AppData\mpv\Shaders\nnedi3-nns16-win8x4.hook"
>>
# Video settings
profile=opengl-hq
opengl-backend=angle
hwdec=d3d11va-copy
video-sync=display-resample
interpolation
tscale=oversample
scale=haasnsoft
scale-radius=3
scaler-resizes-only
scale-antiring=1
deband=yes
deband-grain=0
sigmoid-center=1.0
sigmoid-slope=10.0
opengl-pbo
linear-scaling=yes
correct-downscaling=yes
sigmoid-upscaling=yes
icc-cache-dir="~~/icc-cache"
icc-profile="~~/ULTRA HD.icm"
icc-3dlut-size=1024x1024x1024
input-ipc-server=mpvpipe


# Audio settings
volume=66
alang=en,eng

# Subtitle settings
no-sub
slang=en,eng,ger
sub-font-size=40
sub-shadow-offset=0
sub-border-size=3.5
sub-color="#ffffffff"
sub-border-color="#ff000000"
sub-font="Source Sans Pro"
sub-ass-force-margins=yes
sub-use-margins=yes

# OSC settings
script-opts=osc-layout=bottombar,osc-seekbarstyle=bar,osc-hidetimeout=3000,osc-minmousemove=1,osc-fadeduration=0

# Screenshot settings
screenshot-format=png

# Window settings
autofit-smaller=1920x1080
screen=1
autofit-larger=2560x1440

ytdl-format=(bestvideo[height<=?2160][fps<=?60]/bestvideo[height<=?2160]/bestvideo[height<=?2160])+(bestaudio[acodec=opus]/bestaudio)

save-position-on-quit
cache=yes
cache-file="K:\TEMP\mpvcache"
cache-file-size=62914560
cache-backbuffer=12582912
cache-initial=0
cache-secs=600
cache-pause=yes

#protocol settings

[protocol.http]
force-window=immediate
save-position-on-quit

[protocol.https]
profile=protocol.http
save-position-on-quit

[protocol.ytdl]
profile=protocol.http
save-position-on-quit

[extension.webm]
save-position-on-quit=no
loop-file=inf


I am using this config currently, which has been updated over the last year, so maybe some of what I use is obsolete or nonsensical.

When I play a video in fullscreen, I have about 50% GPU usage (RX480 8GB), when I play it at a scale of 2, 100% GPU usage, scale of 1 ~40% GPU.

Any idea which setting is the culprit for this high a usage?
>>
>>59530117
>Size of the 3D LUT generated from the ICC profile in each dimension. Default is 64x64x64. Sizes may range from 2 to 512.
And there's a lot of redundant stuff too.
>>
>>59530117
>scale=haasnsoft
>scale-antiring=1
>sigmoid-slope=10.0
you don't need neither of last 2 with that scaler
>>
>>59530223
>And there's a lot of redundant stuff too.

Ok thanks, what is redundant?

>>59530312
>you don't need neither of last 2 with that scaler

Ok, I changed it, thanks for the input.
>>
>>59507068
Graphs are getting uglier and uglier.
>>
File: logo-256x256.png (31 KB, 256x256)
31 KB
31 KB PNG
Exactly why is MPV better than MPC-HC (with madvr, ReClcok, etc), or vice versa? I just now made myself a desktop that can do some high quality video rendering, so I'm curious.
>>
>>59530840
cause it can even use vapourwave filters that are much better than you can get in mpv, if you can only properly set it up
>>
>>59502831
Why use a media player? Just write one yourself.
>>
>>59530312
>you don't need neither of last 2 with that scaler
Why?
Not that guy, but I use haasnsoft and sigmoid-slope=10.
>>
>playing anime
>https://animetosho.org/view/damedesuyo-ao-no-exorcist-kyoto-fujouou-hen-11.n910401
>mpv crash at 3:20 while vlc played it fine

wtf I hate mpv now
>>
>>59530556
>correct-downscaling=yes
>sigmoid-upscaling=yes
>deband=yes
These are all included in opengl-hq.
>osc-layout=bottombar,osc-seekbarstyle=bar
This is now the default layout.
>>
Clementine
>>
File: 1444917104273.jpg (6 KB, 185x200)
6 KB
6 KB JPG
How do I make mpv open videos in windows the size of their native resolutions?

It used to do this by default, but now it's opening all videos at my maximum screen resolution instead. I've read through the manual and tried several things such as --video-zoom=0 but nothing seems to work.

The closest I've gotten is using the --video-unscaled option which makes it so that the video is at its native resolution but the window still occupies 100% of my screen space.
>>
>>59532815
do you have autofit in your config?
>>
>>59532844
I have autofit-larger=100%

I tried commenting out the line already to see if that was causing the problem but it didn't change anything.
>>
>>59532815
>>59532861
Ah, I found the problem.

>--hidpi-window-scale, --no-hidpi-window-scale
>(OS X and X11 only) Scale the window size according to the backing scale factor (default: yes). On regular HiDPI resolutions the window opens with double the size but appears as having the same size as on none-HiDPI resolutions. This is the default OS X behavior.

After reading this, I tried using the --no-hidpi-window-scale option, not really expecting it to work, but to my surprise it did.
>>
Does anyone know if --ao=openal has been removed from mpv? Or is there any other way to get HRTF?
>>
>>59532920 again
I don't understand why this option was even effecting me though since I don't have a HiDPI screen. I think my screen's DPI is only 92...
>>
>>59501772
I do not use a software application in which Windows Installer is not for download. Absolutely hideous! Lol
>>
>>59532943
>openal
[ao] Audio output openal not found!
[ao] Failed to initialize audio driver 'openal'
Could not open/initialize audio device -> no sound.
>>
>>59532943
Wait, MPV supports HRTF? Windowsfag here, I switched over to MPV, but back when I used MPC-HC, you had to use some black magic and a Dolby Headphone driver from a DVD player to get HRTF to work.
>>
I tried ssimsuperres/ewa_lanczos, ngu-aa and ngu-sharp, it's clear that ssimsuperres needs a lines thinner and grain shaders to compete. However mpv has a stupid hook limit and I don't think we could make it possible without over passing this limit.
>>
>>59533074
Not MPV per se, OpenAL did. If you send multichannel audio to OpenAL you get HRTF.
>>
>>59533089
vd-lavc-threads=32
>>
>>59533091
I see. Did you get this working in the latest stable release of MPV?
>>
>>59533101
???
>>
>>59533104
It has been at least an year since I last used mpv, so I don't know. Checking Arch's PKGBUILD to see if they disabled ao_openal, since according to github it's still there.

https://github.com/mpv-player/mpv/issues/1073
>>
>>59533128
I got it to work.
Had to recompile with --enable-openal.
Apparently it's been disabled by default upstream (not Arch's repo)
>>
>>59532943
<code>
Available audio outputs:
wasapi : Windows WASAPI audio output (event mode)
openal : OpenAL audio output
null : Null audio output
pcm : RAW PCM/WAVE file writer audio output
</code>
It's still there
>>
>>59533292
use []'s instead of <>'s.
thanks, man, I got it to work, I had to recompile with --enable-openal (don't know why)
>>
>>59533122
https://mpv.io/manual/master/#options-vd-lavc-threads
>>
>>59518407
I can't.
>>
File: 1458910342083.jpg (11 KB, 174x263)
11 KB
11 KB JPG
>>59534072
>>
>try to play video
>no dialog, only sound effects
>works fine in VLC
What?
>>
>>59534124
calm down lad
>>
can someone recommend me some configuration options? I've got a gtx1080 and just wanna watch anime real nice
>>
>>59535189
See >>59518588
>>
Is there a way to set a default window size so that mpv opens in the same sized window everytime?

I'm a Winfant btw
>>
>>59501772
Just used it to watch my mongolian cartoon,
Pretty good, tbqh.
>>
>>59535916
autofit=1280x720
>>
>>59535844
thanks
>>
>comparing scaling filters with low res 3dpd images
Wanking over this shit doesn't mean anything, we all know everyone here solely consumes anime.

Anyways, what's the best and most autistic way to upscale anime to 4K? I'm using double adaptive-sharpen-2pass along with ssimsuperres with 5 taps for everything. It looks sort of artificial but at least it's not a blurry mess.
>>
>>59536531
no mpv shaders for comparison (ewa lanczos)
>>
>>59535844
t. illiterate
>>
>>59536531
>I'm using double adaptive-sharpen-2pass along with ssimsuperres
They don't even work because of a shaders hook limit. You can use only 1 additional shader along with ssimsuperres.
>>
>>59536805
Huh, guess you're right. I used to have nnedi3 or superxbr or something + double adaptive sharpen but replaced it with ssimsuperres and just left the sharpen.




Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.