/mpv/ - please show me it can do better

>madshi's fags on suicide watch
https://github.com/mpv-player/mpv/commit/d8a3b10f45eb10fb34ce9da3a9a76e3bd8644e3d

Yesterday i tried MPV after i saw MadVr takes over my second monitor too and i was unable to browse web while watching stuff. 5 Minutes of video play and i removed mpv shit instant and accepted my fate with madvr even though i can't browse web. Videos in mpv are muddy, why does it praised so much in /g/? its literally shit

I don't care I watch my 4/3 anime on my CRT.

>>60830104
With yellow prosubs?

>>60830053
You probably used a too GPU intensive config, it's a common problem because people think its minimal look means it's a player with minimal features. Mpv is an advanced media player with tons of features, it's not just a windows/directshow renderer like madVR.

>>60830144
this is what i have
[CODE]
vo=opengl
profile=opengl-hq
opengl-backend=angle

scale=ewa_lanczossharp
cscale=ewa_lanczossoft

interpolation
tscale=catmull_rom # $ mpv tscale=help for possible options
temporal-dither
video-sync=display-resample

target-prim=bt.709
target-trc=bt.1886
gamma-auto
icc-profile-auto

blend-subtitles

osd-font-size=14

osd-level=3

slang=enUS,en,eng,ptBR,pt
alang=jp,jpn,enUS,en,eng,ptBR,pt

quiet
[/CODE]

>>60830170
Post your madVR config or you GPU model a least.
MadVR has absolutely no tscale=catmull-rom equivalent feature!

>>60830200
this is my mpv config, i have r9 280x

>>60830209
Any dropped frames?
Did you try tscale=oversample?

>>60829986
>work has resumed on vo_vulkan
>haasn is back
Nice.

File: Untitled.jpg (286KB, 1659x662px) Image search: [Google]

286KB, 1659x662px

>>60830227
no dropped frames its just muddy

left one is madvr and rightone is mpv with the config i posted above. madvr videos look sharper.. i didn't try anything else because i found that config on some mpv forum.

>>60830259
vo_vulkan? where?

>>60830295
https://github.com/atomnuker/mpv

>>60830293
You can use adaptive-sharpen.glsl and/or SSimSuperRes.glsl shaders to increase sharpen, there are much better algorithms than what the current madVR release provides.

Adaptive-sharpen:
https://gist.github.com/igv/8a77e4eb8276753b54bb94c1c50c317e

SSimSuperRes:
https://gist.github.com/igv/2364ffa6e81540f29cb7ab4c9bc05b6b

Please read the shaders description and don't forget SSimSuperRes works only when upscaling.

>>60829910
I just use Kawaii Codec Pack.
Other than the name being web as fuck It's a pretty cool guy that doesn't (seem) to afraid of anything.

>>60830320
Oh thanks! Github search is really a terrible shit!

Why does no one show comparisons in motion? Filters always look great in still frame, but looks weird as fuck once in motion.

>>60830398
Sometimes yes, I literally hate extra sharpen/oilpainting algo in motion.

File: madvr.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

CHALLENGE: produce an equally sharp mpv upscale to this MadVR upscale with as little noise and artifacting as this upscale, and post repeatable mpv settings for the upscale. MadVR settings for this are NGU AA medium image upscaling, and adaptive sharpen 1.0 + crispen edges 1.0 + anti-ringing filter under upscaling refinement. This image utilizes AMD denoising but you can use mpv's dn3dHQ for an equivalent effect.
Note: your subjective opinion of how nice this looks is irrelevant to this challenge. This is a technological experiment to see if mpv can produce an upscale this sharp without noise or artifacting. Refusing to attempt the challenge because of ambiguous buzzwords will be interpreted as an admission of defeat.

>>60830200
>tscale=catmull-rom
What is tscale?

File: [Doki]_Yuru_Yuri_-_07v2_(1280x720_h264_BD_AAC)_[DF080EE5].mkv_snapshot_01.33_[2017.06.09_15.31.28].png (2MB, 1280x720px) Image search: [Google]

2MB, 1280x720px

>>60830445
Source image attached.

>>60830137
If possible.
I usually watch from DVD or cassette, so I don't have much choice.

>>60830293
This is proof that madvr/windows poo in loos like oversharpening and artifacting. Jesus christ my fucking eyes are bleeding. I think I need to wash them with a bicubic filter just to fix things.

>>60830351
>there are much better algorithms than what the current madVR release provides
Such as?

>>60830053
>why does it praised so much in /g/?
It's FOSS. That's the only reason. /g/ would openly recognize MadVR's superiority if MadVR were FOSS.

>>60830500
>This is proof that madvr/windows poo in loos like oversharpening and artifacting. Jesus christ my fucking eyes are bleeding. I think I need to wash them with a bicubic filter just to fix things.
Left is unscaled, retard. Chroma alone wont make such a difference.

>>60830505
*they -> adaptive-sharpen.glsl and SSimSuperRes.glsl

Igv improved them a lot!

>>60829986
>https://github.com/mpv-player/mpv/pull/4498#issuecomment-307331514
Good.

>>60830544
How do you know that madVR has outdated versions? Also no point in superres really anymore thanks to NGU. Adaptive sharpen sometimes useful but i doubt there will be much difference with the never version.

File: 00074what.jpg (92KB, 482x569px) Image search: [Google]

92KB, 482x569px

Why does /a/ care so much about the sharpness of still shots? You're either watching on a CRT (Inherently blurry) or something like LCD/OLED (slow pixel response), so you're not going to see the sharpness benefits anyway.

Even for your 0-12fps cartoons it really doesn't matter.

>>60830575
>Adaptive sharpen sometimes useful but i doubt there will be much difference with the never version.
If there was a big improvement mpvfags would be able to complete the >>60830445 challenge and prove it.

>>60830518
it's FOSS and multiplatform don't forget about multiplatform
also it's more streamlined and easier to configure even with the text config

File: toorushrug2.jpg (197KB, 1167x921px) Image search: [Google]

197KB, 1167x921px

>>60830600
I see the difference in motion. Maybe your eyes suck.

>>60829910
>wasting time in getting the best filtering
>not getting quality releases

>>60830518
No madVR is just too limited by design.
Windows only is THE limitation.

>>60830575
Igv's versions are different than ones provided by Shiandom or bacondither!

>>60830608
For over processed rendering, you should use vapoursynth.
I think mpv is the only one player that currently works well with it at realtime.

>>60830633
Why not both?

>>60830615
>easier to configure even with the text config
That's really a matter of opinion with mpv configuration you either have a steeper learning curve re: memorization of config lines or you need to use a reference for configuration. Also, cycling back and forth between MadVR profiles isn't exactly difficult or slow.

>>60830629
Maybe you should actually be paying attention to the content instead of the lines.

>>60830650
why would you (over)filter a high quality release?

>>60830644
Linux is poor for multimedia anyways, especially due to its audio limitations. Windows has a lot more audio processing options and audio is half the sensory experience of watching videos/anime.

what am i looking at? is this some shader for compressed vids

>>60830663
What do you mean by "filter"?

File: gabuface2.jpg (36KB, 347x394px) Image search: [Google]

36KB, 347x394px

>>60830662
I see the difference in motion when not even looking for it. I think your eyes suck.

>>60830671
Let me guess,you only use pulseaudio,right?

>>60830671
Nah that's the opposite, linux audio is future proof.

File: 00096what.png (309KB, 703x703px) Image search: [Google]

309KB, 703x703px

>>60830687
If you knew anything about how signal processing works, you'd know it's not your eyes which cause things to look "better".

If you really cared about quality why are you downloading low quality releases in the first place?

Do you even think about things like this? My life is pretty disordered, but come on anon.

>>60830671
That's why 99,99999% of multimedia devices are running a linux based system. Dude, are you still living in the 90s?

>>60830692
No; never used it. I use Equalizer APO system-wide in Windows and I use a VST headphone crossfeed in foobar and MPC-HC (via ffdshow's winamp plugin support and a VST loader for winamp) which I got off the old Changstar forums and which is the only crossfeed I know of which does not affect frequency response.
Neither a high quality system-wide parametric equalizer with a sufficient number of bands nor a high quality crossfeed (bs2b rolls off treble) exist for Linux.
These are only the examples which affect me, but Linux in general is much more limited regarding audio options.

>>60830735
Signal processing is bullshit.

>>60830735
Dude high quality releases also need to be upscaled and can be bettered. Also not everything have a high quality release. Stop with this filtering bullshit.

>>60829910
Is ngu sharp destructive though? I refuse to use any options that are destructive.

>>60830658
you just have to ask a faggot from /g/ and they will come with the best option they found with their weaponized autism

>>60830700
>>60830739
You cannot possibly be arguing that linux has more audio software processing options available to it than Windows. That's patently false. The way Linux's audio is structured in general is irrelevant to this if that's what you're getting at; I'm talking about how many different examples of audio processing software have been developed for the OSes.

>>60830755
>equalizer
>ffdshow
You are literally living in the 90s! That's just AWESOME! :D

>>60830762
Stop using shitty low PPI monitors and you wouldn't have this problem. 4k monitors aren't every expensive these days.

>>60830053
>MadVr takes over my second monitor too and i was unable to browse web while watching stuff

That's kind of weird. Definitely haven't had that problem.

Hope you find a media player that works out for you, though.

>watching low quality media that requires any sort of upscaling or filtering
>not having a 24p capable display

It's funny how superior you all think you are, when you're actually the biggest plebs on /g/ for needing to use media players like this in the first place. I bet you're all tpb-using pajeets that download everything in 720x300 xvid.

>>60830769
I've done a lot of testing with this and I've come to the conclusion that NGU AA with higher adaptive sharpening settings is better than NGU Sharp with lower adaptive sharpening settings with great sources and the difference only grows with poor sources.
>>60830735
I seek to optimize both what sources I download and the signal processing of said sources. As >>60830762 said upscaling high quality sources can still be improved with signal processing like sharpening and denoising.
>>60830757
>subjective improvement is bullshit because muh signal purity
Absolute illogical, superstitious hodgepodge. When you believe in things that you don't understand, then you suffer. Superstition ain't the way.

>>60830778
You are describing what we call business. Software were developed for Windows because Microsoft was and is in some countries forcing manufacturers to sell Windows with their products.

>>60830780
What do you even mean?

>>60830779
Literally every pair of headphones ever made except maybe the ER4S and literally every pair of speakers ever made can benefit from equalization. LAV audio filters offer no discernible benefit to me over ffdshow audio processing, which contains far more features and can even provide virtual headphone surround sound with surround-encoded videos.

>>60830829
Upscaling isn't as noticeable on pixel dense displays.

>>60830853
ffdshow, lavfilters, mpv, ... this is just THE ffmpeg

File: satanialaugh.webm (3MB, 1440x810px) Image search: [Google]

3MB, 1440x810px

>>60830780
>buying meme resolution displays before the 4k standard is finalized

>>60830878
Madshi's fags use $30k 8K projectors.

>>60830780
I'll say this since this is 4chan home of the weeb: patrician anime viewers will possess a 1080p display for anime viewing since there is still zero anime content above 1080p right now, and since the majority of new seasonal anime shows contain lineart scanned at resolutions above 720p according to anibin.

>>60830445
>mpvfags will ignore this post because it proves their software is deprecated until it can use NGU AA

>>60830878
4K has two finalized standards. What the fuck are you even talking about?

>>60830932
Upscaling on high PPI still looks better simply because there are more pixels to work with.

>>60830932
People like scaling their shits and and hear their GPU roaring.
Please be tolerant and kind.

>>60830949
NGU-AA is really similar to NNEDI3 and a lot of people don't like them.
Madshi wasted 8 months of his life for that shit.
Please don't compare active people with doom9 retarded winfags.

>>60830996
>lot of people don't like them
These people are known as technophobes. They are more concerned with "muh signal purity" than which fucking image looks less goddamn blurry.

>>60830955
>Upscaling on high PPI still looks better simply because there are more pixels to work with.
Upscaling to a higher resolution does not magically create image detail out of nowhere. Upscaling, period, does not magically create image detail out of nowhere. Enjoy your placebo.

>>60829910
Nagoo+linux

>>60831026
Dumbass, it better represents the source image because it has more pixels to represent it. When you have a "normal" PPI monitor there aren't enough pixels, so non-integer scaling has more obvious distortion (blurring).

>>60830869
Are you retarded?

>>60830949
I use MPV because it seems bretty decent and I'm actually watching that post, eagerly awaiting replies to make my anime look slightly better
>tfw always download 1280x720 but monitor is 1440x900

File: smugsawako.jpg (34KB, 359x377px) Image search: [Google]

34KB, 359x377px

>>60831061
http://www.rtings.com/tv/learn/4k-ultra-hd-uhd-vs-1080p-full-hd-tvs-and-upscaling-compared
>If you're only watching 1080p or even smaller resolution content, it won't give you a boost in quality.

>>60830955
Um.
what the fuck mang

>>60831114
Are you really so retarded that you don't know the difference between "boost in quality" and a "better represent the source"? Why are you even on /g/?
Jesus Christ.

>>60831130
see>>60831061

>>60831111
I've already posted it twice and a bunch of people tried and failed. I wouldn't hold your breath. NGU AA's primary advantage to me as an anime viewer who often upscales 720p to 1080p is that it handles high sharpening settings very well without producing much noise (the little noise that sharpening produces is eliminated by video card denoising or dn3dhq, which ffdshow supports) or any objectively definable artifacts like ringing or aliasing. I can't say the same about any other image upscaling algorithm that MPC or mpv use, at least in my testing. I'm actually hoping someone can complete the challenge successfully, though, because mpv's subtitle renderer is really great and fast.

>>60831149
>Are you really so retarded that you don't know the difference between "boost in quality" and a "better represent the source"? Why are you even on /g/?
So, tl;dr: you're sperging out over a technicality which has no impact on actual perceived quality. Got it.

>>60831178
Just stop.

>>60831192
OK. I won the argument, so sure, I'll stop.

>>60831208
Sure. Typical /g/.

>>60831149
My "what the fuck" was referring to displaying 1080p content on a 4K monitor rather than a 1080p monitor

>Technically Top Tier: 4K anime on 4K monitor
>Ascended Tier: 4K anime on 1080p monitor
>Very Suitable Tier: 1080p anime on 1080p monitor
>Placebo 2.0: Fucking Kill Yourself Tier: 1080p anime on 4K monitor

>>60830170
>pt-BR

Yeesh

>>60831227
That's what nearest neighbor scaling is for anon. Not all anime is 1080p either. LoGH being a great example.

>>60831227
This is hypothetically accurate except >1080p anime doesn't exist. There's usually one series per season at best with full 1080p lineart (KyoAni series have always had full 1080p lineart as far back as Hyouka) and that's the limit.

File: yall-mind-if-i-think-on-19238723.png (178KB, 500x689px) Image search: [Google]

178KB, 500x689px

>>60831266
(That's why I put the qualifier "technically" in there)

Most anime is 1080p at best, which looks better on a 1080p monitor than a 4K monitor,

>>60831257 I *believe* that Native Physical Pixels is a better upscaler than nearest-neighbor-2x, though that would need some testing to confirm

Jinc still the best.

>>60831227
You don't even need 1080p for anime in the first place, 720p is more than enough. Especially given how anime doesn't even have decent artwork anymore. If it was hand drawn anime, I can see an argument being made for 1080p, but the modern shit just had no need for anything above 720p.

>>60831402
...at being blurry.

File: 1479005876572.jpg (5KB, 250x250px) Image search: [Google]

5KB, 250x250px

>>60831227
I've had enough

>>60831450
:^)

>>60831431
http://anibin.blogspot.ca
The majority of new series contain >720p lineart. You are incorrect. KyoAni and JC Staff in particular consistently release anime that's well worth the 1080p downloads.

>>60831484
Yeah cause I totally want to admire the lineart of the latest Kyoanus moeshit.

>>60831484
Yeah but anime in general looks pretty horrible these days.

File: staythere.jpg (109KB, 924x664px) Image search: [Google]

109KB, 924x664px

>>60831505

File: 1481777459424.jpg (152KB, 419x419px) Image search: [Google]

152KB, 419x419px

MPV pros... I need good settings for anime watching. Usually 720p source enlarged to 4K screen.

Which settings are the best?

>>60831551
yes, kyoanyfags are usually >>>/v/ fags

>>60831266
Hyouka is 720. I saw one scene with a few lines that were >720.

>>60831576
https://github.com/haasn/gentoo-conf/blob/xor/home/nand/.mpv/config

Post configs.

>>60831591
https://anibin.blogspot.ca/2012/04/1_26.html
You are incorrect. When anibin says 955.5 vertical resolution for KyoAni, that actually means full 1080p. KyoAni's After Effects use fucks with anibin's estimate methodology, and behind the scenes footage on the Chuu2 BDs show the animators working in full 1080p.

>>60831588
>>60831622
>I actually have to explain the meme
The (ancient) meme is that /v/ only likes battleshonen and that people who say "moeshit" need to go back to /v/.

>>60831629
I this a pasta?

>>60831576
NGU AA... oh wait, you're using a deprecated media player.

>>60831627
https://hastebin.com/xocukeyawo

File: 1475938758522.jpg (753KB, 1280x1442px) Image search: [Google]

753KB, 1280x1442px

>>60831604
I don't want the whole thing anon. I'd like to make a new preset called [anime] and just use it for when I watch anime (instead of YT/TV shows... when I use HW decoding).

Could you improve on this please?

[hw-high-quality]
profile=opengl-hq
deband=yes
deband-iterations=4
target-prim=adobe #monitor dependent
target-trc=gamma2.2 #monitor dependent
icc-profile-auto
icc-cache-dir=~/icc-cache
icc-3dlut-size=256x256x256
scale=ewa_lanczossharp
cscale=ewa_lanczossoft
dscale=mitchell
tscale=robidouxsharp
dither-depth=auto
scaler-resizes-only=yes
sigmoid-upscaling=yes
correct-downscaling=yes
temporal-dither=yes

 

File: Untitled.jpg (566KB, 1920x2160px) Image search: [Google]

566KB, 1920x2160px

Above is MPV
Below is MPC-HC+MadVR

HOW CAN FOSSFAGS EVER RECOVER?!?

>>60831402
Jinc is the best linear scaler. It is especially good with anime.

>>60831759
with better scalers once these obnoxious shitposters calm down

I mean, I don't know what the situation is (I use MPV currently because it seems suitable enough, though I could always go for a better scaler); but holy fuck you MadVR people make Apple users seem non-autistic

>>60831805
>holy fuck, you MadVR people make Apple users seem non-autistic
I say this, my first time hearing about this program(? scaler?) being this thread

>>60831805
You realize the top is a 360p release and the bottom is a 1080p release and he's just baiting for replies because he knows madvr is shit

>>60831759
Please use horizontal comparisons in the future

>>60831759
Post source image.

File: deebly_goncerned.png (89KB, 600x500px) Image search: [Google]

89KB, 600x500px

>>60831829
interesting
I should have guessed that from his tone
and on second look it's really obvious :|

There was another thread where someone was shitting on Gigabyte for no apparent reason, and screeching "shill" or "not an argument" at anyone who asked for clarification…I wonder if it's the same troll?

or is that image actually MadVR upscaling a 360p video? :O

>>60831604
What is this?

>>60831629
The number of 1080 details are minimal.

File: laughing_girls.jpg (39KB, 540x360px) Image search: [Google]

39KB, 540x360px

>>60831759
>trying this hard

>>60830445
>still no attempts
ITT: mpvfags admitting defeat

File: 1477542767585.jpg (98KB, 1147x646px) Image search: [Google]

98KB, 1147x646px

>>60831759
>images not the same size
do you work for CNN or NYTimes? kek
fuck off troll.

File: smoothiesmooth.webm (3MB, 1280x724px) Image search: [Google]

3MB, 1280x724px

>>60831805
Relax, it's a super low-quality bait I made because this topic is just so darn picky.

I use DXVA2 all around, and I'm more of a smooth playback person than a soft/sharpen person.

>>60831829
No both are the same sourece, I just printscreened before the scaling begun.

>>60831845
Meh.

>>60831859
Eh...

>>60831924
I was trying as little as possible.
You knew it was a super obvious bait image, or you should've known.

>>60831948
No, I just couldn't bother doing it properly.

>>60831957
Smooth motion or SVP? Smooth motion is patrician-tier; SVP is absolutely disgusting.

>>60831978
his webm doesn't look like shit so presumably it isn't SVP

>>60831978
>One makes things smooth, the other is named Smooth but doesn't make it smooth.

https://www.youtube.com/watch?v=Y5XdghuwLOM

>>60832001
I use SVP, but I have hardware good enough to render it.

>>60832011
>I use SVP, but I have hardware good enough to render it.
not having the hardware for it isn't the problem, it's inherently bad.

File: SVP.png (24KB, 502x582px) Image search: [Google]

24KB, 502x582px

>>60832011
Forgot image.

>>60832015
>works_for_me.webm
As shown, considering you thought I didn't use it because it didn't look like shit.

File: 1493333525795.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

>>60832011
Troll video. Smooth motion produces flawless, judder-free panning shots for me. What smooth motion does NOT do is this bullshit.

>>60832028
Why not just use smooth motion? Or you like soap opera effect?

File: GkzStsx.jpg (42KB, 600x338px) Image search: [Google]

42KB, 600x338px

>svp

Can anyone suggest any improvements here? Aiming for maximum quality here.

profile=opengl-hq
scale=ewa_lanczossharp
cscale=ewa_lanczossoft
dscale=mitchell
scale-antiring=0.7
cscale-antiring=0.7
dither-depth=auto
correct-downscaling=yes
sigmoid-upscaling=yes
deband=no
volume-max=100
hwdec=no
video-sync=display-resample
fs=yes

>>60831957
Why is your source fucked up? One Piece openings are 30fps.

>>60829910
I feel like I'm missing out heavily using mpv, but I refuse to use Windows.

>>60832099
You need some debanding.
>>60832118
You dont miss much if your mpv settings are correct and you use Jinc as luma upscaler.

>>60832062
Because mpv interpolation is more flexible!

>>60832118
this is the vague suspicion I have

but there's too much shitposting to tell

I don't want to leave my /comfy/ to watch animu
not to mention the Torrent drive is EXT4

We need more posters like >>60832157 who actually meaningfully contribute to discussion

how come i cant find any info on ngu-sharp's algorithm

>>60832180
>I don't want to leave my /comfy/ to watch animu
Especially with animu you wont notice almost any difference between doubling/quadrupling with NGU and Jinc in mpv. Just configure mpv properly, read manuals and forget about madVR.

t. avid madVR user.

>>60832118
You're missing out heavily with regards to upscaling. If you weren't somebody would be able to complete the >>60830445 challenge which mpvshills are ignoring at all costs right now.

>>60832202
What kinda info you want?

File: Smooth Line.webm (2MB, 932x526px) Image search: [Google]

2MB, 932x526px

>>60832054
You have problems with your computer, your fault.

>>60832062
Because smooth motion doesn't do much at all.
Did you cry when watching 48Hz movies in cinema as well?

>>60832073
>SVP

>>60832110
No they are 23.976 fps

>>60832118
Not really, keep using what works for you rather than messing with settings more than you watch the videos.

>>60832203
This.

>>60832203
The difference between NGU AA and Jinc alone is small, but the difference between them once sharpening is applied is massive.

>>60832224
>Because smooth motion doesn't do much at all.
Well, it removes judder, good enough for me.

>>60832240
Jinc+superres is good enough for most people. You probably wont notice difference with NGU AA+sharpeners without zooming 300%.

File: REDLINE_(2009)_[1080p,BluRay,flac,x264]_-_THORA.mkv_snapshot_00.40.08_[2016.11.04_10.55.46].png (3MB, 1860x1044px) Image search: [Google]

3MB, 1860x1044px

>>60832257
Good for you.
Not enough for me, I use SVP instead to force 144FPS.
Both happy, with our respective solutions.

>>60832222
like it's license and how i can replicate it for my software

>>60832265
I will notice it in motion without zooming without even looking for it. I've tested it. Then again, I have 20/10 vision and am not viewing anime on an itty bitty display; I'm viewing a 32" display 3 feet away.

>>60832272
Id like to try SVP or something similar with hentai. I really really like 60FPS hentai.

>>60832265
If I don't want to deal with GSGL, is

vo=opengl-hq:scale=ewa_lanczossharp

all I need in my mpv.conf?

>>60832275
>replicate it for my software
You cant. Its madshi's baby. He is gonna sell it for avisynth soon so you can guess the license.

>>60830053
>MadVr takes over my second monitor too and i was unable to browse web while watching stuff
It doesn't do anything of the sort, you probably have it set to full-screen exclusive mode in which case it behaves the same way a full-screen game does, i.e. you cannot use another monitor without going windowed or alt-tabbing.

>>60832275
You mean copy paste, edit a few lines and removing anything that is related to the word NGU and credit to the maker, then claiming it your own?

>>60832290
There's a non-jew version with limited features which I use as seen in the webms I've posted.

Just make sure to switch to Animation instead of Film.
Play around with masking and quality until your PC manage to render with without having interpolation artifacts or max everything if you have good enough hardware.

I could've shown you a webm of 60FPS hentai if I actually stored hentai.

File: 1464079349-chrome-2016-05-24-10-42-28.png (181KB, 452x382px) Image search: [Google]

181KB, 452x382px

>svp

File: 500px-Anime-1.jpg (25KB, 500x315px) Image search: [Google]

25KB, 500x315px

>>60832333
>svp

File: REDLINE_(2009)_[1080p,BluRay,flac,x264]_-_THORA.mkv_snapshot_00.35.47_[2016.11.04_10.55.20].png (3MB, 1860x1044px) Image search: [Google]

3MB, 1860x1044px

>>60832333
>>60832347
>SVP

Use something better than a Pentium 2 with ATi 3D Rage II+DVD and you might be able to achieve better image quality.

>>60832330
>You mean copy paste, edit a few lines and removing anything that is related to the word NGU and credit to the maker, then claiming it your own?
no?
>>60832306
isnt mpv open source?

hey anon with the alternate webm script, are you ever going to post it? should I give up on waiting?

>>60829910
How about doing something productive instead of watching chinese cartoons?

>>60832330
what is a software that has NGU implemented so I can just reverse engineer it out and piss him off

>>60832559
The one you make yourself from scratch that hopefully is as good as NGU.

Are you Sgt. Mark IV perhaps?

>>60832589
Just dled madVR and ill document his algorithm

personally, it's because this really pissed me off
http://madshi.net/madCodeHookDescription.htm
>pay for it after he sees you make "good code"

>>60832605
I see nothing wrong with it.
It's his program and if people use a feature he made for ill purposes he can and have already pulled the plug.

Don't like it? Make your own.

>>60832622
>"buy" it once, release it for free
how mad are you madshi

File: rawr.png (171KB, 548x464px) Image search: [Google]

171KB, 548x464px

>>60832664
Awesome, I just turned into a person that made a program.

>>60830445
>mpv's dn3dHQ
Can't find it.
Which tab?

>>60830445
Useful to use it under image refinement instead?
To battle shitty upscales of 900p anime studios provide?

File: Screenshot_2017-06-09_21-39-18.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

>>60832224
>Not really
Nigga madVR upscaling is still the best out there. No one has posted something that looks as clean as that Yuru upscale. I'm an mpvfag, my config can't produce an image with that quality. If anyone has better configs, please do help out.

I haven't tinkered them, they're just almost open-gl defaults:

###Video###
vo=opengl
scale=ewa_lanczossharp
sigmoid-upscaling
cscale=ewa_lanczossoft
dscale=mitchell
correct-downscaling
deband
tone-mapping-param=0.5
hwdec=no

###Audio###
ao=alsa
audio-channels=5.1
volume=100

###Track Selection###
alang=ja,jpn,Japanese,en,eng,English

###Subtitles###
slang=jpn,ja,Japanese,eng,en,English
demuxer-mkv-subtitle-preroll=yes
sub-gauss=.4
sub-gray

###Screenshot###
screenshot-format=jpg
screenshot-template="/home/anon/screenshots/%F_%#04n_%p_[%tY_%tm_%td_%tH]"

###Playback Control###
[extension.gif]
loop-file=inf

[extension.webm]
loop-file=inf

[extension.jpg]
loop-file=inf

[extension.png]
loop-file=inf

###Program Behavior###
save-position-on-quit
ytdl

###Window###
keep-open=yes

>>60832790
Nigga, I'm not defending MPV, I'm using MPC-HC+MadVR(+SVP) myself.

>>60832790
>not even using profile=opengl-hq
>not even using ssimsuperres and adaptive sharpen
yea you wont get close if you dont even know how to use mpv

>>60832829
Are you misunderstanding something?
That post was that I'm missing something by using mpv.
Your psot said not really.
My very last post is admitting madVR to be superior on retards like me.
>>60832859
Look at the output of profile=opengl-hq, most of the stuff is already there on the config. On the others, I have no idea.

>>60832877
Well then, I did misunderstood you.
Anyway, just put Windows on a liveCD(USB) and...
>EXT4
Oh...
As long as the anime works, eh?

>>60832877
On the others, I have no idea.
No shit

>>60832895
Fuck Windows.

>>60832912
Post screenshots and configs then, you faggot.

File: 1494637491354.jpg (47KB, 300x300px) Image search: [Google]

47KB, 300x300px

>>60832918
Your choice.
That's the price you have to pay for using FOSS.

Yes I think my horrible pun was fun

>>60832681
Anyone knows?

>>60832928
why? I dont care what you use. If you like madvr more than fucking use it.

I need people to test the vo_vulkan renderer.

https://github.com/atomnuker/mpv

Pretty much guaranteed to work perfectly on intel with wayland now. Probably the same with amd. x11 works too but just don't resize by dragging the window. No idea if it crashes on nvidia (or worse) since apparently everyone else is too busy to test, probably with playing games and wanking.
Run in the main mpv repo directory (with ./build/mpv <file> since it looks for shaders in ./wip_shaders).
If it crashes, try --vulkan-direct=no. If it still crashes provide a log with -v.
Only supports yuv420/yuv422/yuv444 at 8/10/12 bits currently.

Its so fast it'll happily do 8k at 30fps on a Skylake, the bottleneck is now during decoding.

>>60832958
Might as well not point anything out then you shitter.

>>60829910
what is ngu sharp?

and which is better for down-scaling? MadVR or mpv?

>>60830445
I use MadVR but that looks like shit. Reminds me of using HQx4, SuperEagle, or SuperSAI filters for video games.

>>60832975
atleast you're less ignorant now
you're welcome

>>60832895
I have windows running on an Ext4 HDD (in a VM), fucking witchcraft.

>>60832991
Much appreciated. Now go fix your terrible grammar.

>>60832895
>not using GPU passthrough on a KVM

File: 1487612269652.png (1MB, 1184x696px) Image search: [Google]

1MB, 1184x696px

>>60832417
Anon will sure deliver. We just have to wait.

>>60832968
Also make sure --hwdec is explicitly set to no if its enabled in a config file, I know some freaks use it.

>>60832982
NGU is an image upscaling (doubling) algorithm, NGU Sharp is a variant which is tuned to produce a sharp image. No idea about downscaling, though I imagine there wouldn't be any real-world visible difference if both are configured adequately.

>>60832988
I can imagine the autism emulation threads cause whenever someone mentions higher res or filters over accuracy.
On the image, mpv looks a little soft in comparison and it looks even worse.

>>60832968
https://pastebin.com/raw/V8MZR4T1

>>60832221
>>60831939
>>60830949
>>60830608
stop replying to yourself

>>60833253
Higher res is fine, CRT filters are fine if that's what you want, but shit like SuperEagle is just awful.

File: mpv yuru.png (3MB, 1920x1080px) Image search: [Google]

3MB, 1920x1080px

>>60830445
Am I close?

>>60833280
yep

>>60833259
Autodetection was broken, fixed, do a git pull. Or use --vulkan-backend=x11 to force it.

>>60833280
You're close, but you're blurrier.

>>60830293
look at her eyes. massive oversharpening. Winbabbies must be blind, or they've adapted maybe

>>60832365
CPU/GPU speed does not affect single frame output of SVP. It affects ability to use SVP without dropped frames.

File: ss.jpg (342KB, 1920x1080px) Image search: [Google]

342KB, 1920x1080px

>>60833366
It's working now. Tested it on a few different videos. https://pastebin.com/raw/Jhx4riZW

It always segfaults on this file though https://pastebin.com/raw/y6tXhvyH

I have to make sure to turn off tiling on the workspace because any fucking with the window will completely freezes my PC and I have to reboot. I can switch to fullscreen fine but it looks like this.

>>60830200
GTX 1060 6gb. What should my mpc+madvr settings be?

>>60832968
I just tested it on nvidia (driver 381.22 + X11). Manual resizing locks up driver (I assume this is expected, but I tested it anyways), subtitles don't appear, and when showing 720p scaled to 1080p stuff full screen on a 1920x1200 screen, the last line in the video is repeated to the bottom of the screen.

File: lsfmod via vapoursynth.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

>>60830445

>>60833826
Almost. Extremely fucking close. The Japanese characters just off Akari's left shoulder are ever so slightly sharper in the MadVR upscale. Post your settings. That's very impressive.

>>60833826
Aw shit nigga, that's pretty neat.

File: Untitled-1.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

>>60830445
Posting NGU AA Very High to see if it makes a difference.

>>60830633
>thinking that his FHD BD rips don't need chroma upscaling
What a retard.

>>60833826
Sauce on lsfmod?
I'd have to do something like

vapoursynth=<lsfmod.filename>

on the conf?

>>60834206
Can't tell a difference between NGU AA medium and NGU AA very high. Comparing >>60833826 with either NGU upscale, they are incredibly close but the NGU AA upscale is very very very very very slightly sharper. Both the lineart and the stuff on the bulletin board in the background (especially) is very slightly sharper on the MadVR upscale. I will readily concede that it's so close that it makes little practical difference.
I'm very curious to learn of the repeatable settings >>60833826 needs to provide. I'm very curious as to exactly how demanding those settings are, as NGU AA is extremely fast. If >>60833826 is using NNEDI3 it's almost certainly much slower than the NGU AA settings.

>>60830955
Upscaling to higher resolution with non destructive algorithms is only going to give you more fabricated pixels and potentially make it look blurry. Some algorithms can still look sharp, and neural network based upscaling can do some extremely detailed magic sometimes, but if you consider precision and faithfulness as your requirements for "quality" you're never exactly going to get a better image by upscaling something, the best you can do it give more or less the same experience as the original.

File: circled.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

>>60830445
>>60833826
Areas circled in red are where it's clear that MadVR's upscale is still very slightly sharper. Super fucking close, though.

File: 2017-06-09-211428_3840x2160_scrot_dscaled.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

>>60830445
Hi. Can we do 4K instead? I thought we'd moved beyond poorfag tier resolutions like 1080p.

>>60834239
It's a port of an avisynth filter: https://forum.doom9.org/showthread.php?t=166582

You need to have mpv with vapoursynth support enabled, then load lsfmod.vpy:

import vapoursynth as vs
import havsfunc as haf
core = vs.get_core()
haf.LSFmod(video_in,strength=300,ss_x=2,ss_y=2,dest_x=1920,dest_y=1080,soft=50).set_output()

Probably not the best settings but I put exactly 0 thought into them. Seems to run fine on my hardware with 720p stuff, at least.

>>60834386
see: >>60830932
Also, your image exhibits significant aliasing.

>>60834441
Your dumb opinion makes me unable to take any other part of your post seriously, sorry. Please correct that and we can have a discussion.

>>60834386
Damn, that almost looks like Q-TEC.
An impressively hideous achievement.

>>60834447
You're welcome to your opinion but your opinion is objectively wrong. A 1080p display is objectively the correct tool for anime viewing right now. 1080p releases are the proper releases to download for the majority of new anime and videos are always best viewed at their native resolution.

>>60834386
oil painting

>>60834469
Sorry, I don't take my advice from poor people.

>>60834479
>>60834468
Downscaling it from 4K in GIMP seems to have introduced aliasing. Oh well. I'd like to see a MadVR image as sharp though, because I haven't seen one yet.

>>60834404
Are you going to post your complete mpv settings as the challenge required? If you do not people can assume you simply used waifu2x and Photoshop or something.

>>60834491
>rich people can't own a 4k display for live action content and a separate 1080p display for anime
Logical fallacy.

>>60834511
Prove you do. Take a pic.

>>60834501
Not that anon, but I'll try it out and post settings if successful.

>>60830445
Why do people like adulterating their waifus this inelegant way so much?

File: 02.mkv_snapshot_03.45_[2017.06.09_21.47.32].png (2MB, 1440x1080px) Image search: [Google]

2MB, 1440x1080px

>>60834656
>tfw mai waifu received a fucking Q-TEC upscale

>>60834521
>>60834511
>he can't prove it
Not surprised. It's one thing to make reasoned statements like "the benefit of 4K over 1080p is small for anime, but seeing its overwhelming benefits for literally anything else, I think the resolution is valid for comparison". It's another to rail against 4K as "non-patrician" or inferior. It shows a deeper bias against the resolution. This is why I don't take advice from poor people - or rather, people who can't afford 4K displays. It's not out of any sort of class superiority or anything like that (I'm not exactly wealthy myself, and I freely admit that), it's because there is an inherent bias against something that exists out of your price range - for every person, whether they're aware of it or not. Asking him for proof was never a move to "win" the argument - it was merely an attempt to confirm bias or lack thereof. His lack of proof has thus confirmed his bias, and that his arguments stem from emotion rather than logic.
Now, I could evaluate his arguments through logic, but really, I'm not going to change minds that way. Given that 4K as a resolution does help day-to-day productivity, I'd encourage anyone who doubts that anime on a 4K screen looks as good as on a 1080p screen to buy a 4K TV or monitor, and give it a shot. If you still disagree, sure. But my guess is that you won't.

>>60830445
>Posting a sharp mess

>>60834683
Anime on a 4k screen will, at very best, look equivalent to 1080p while dramatically increasing CPU/GPU load for upscaling. A 4k display is thus objectively inferior to a 1080p display for anime. This is not an opinion.

>>60834501
Other settings shouldn't make much of a difference. And really it could be tweaked to look even closer, but I don't actually care for this sort of filtering so I didn't bother. I just thought it was interesting how similar it looked to madshi's stuff.

File: question.jpg (66KB, 615x924px) Image search: [Google]

66KB, 615x924px

name=$( basename "$1" | cut -d . -f 1 )
    mpv --no-terminal \
        --script-opts=osc-visibility=always \
        --player-operation-mode=pseudo-gui \
        --external-file="$( youtube-dl --get-thumbnail --default-search=ytsearch "$name" )"

can this into lua? to fetch thumbnail for local audio_only files

>>60830445
Here >>60833826

>>60834731
>Other settings shouldn't make much of a difference.
Why would you post this and damage the credibility of your challenge attempt by doing so? Just post the fucking settings.

File: mpv-shot0001.png (410KB, 840x552px) Image search: [Google]

410KB, 840x552px

>>60830445
CHALLENGE: produce an equally sharp madvr upscale to this mpv upscale with as little noise and artifacting as this upscale

File: 1477103092689.jpg (87KB, 649x815px) Image search: [Google]

87KB, 649x815px

Any news on vo=vulkan?

>>60834809
see >>60832968

>>60834706
Believe what you want. I'm not going to try to convince you.

>>60834805
src: https://github.com/haasn/mpvhq-upscalers/raw/master/rose/original.png

File: 1467482096053.png (56KB, 188x188px) Image search: [Google]

56KB, 188x188px

>>60832968
>Its so fast it'll happily do 8k at 30fps on a Skylake

>>60834751

profile=opengl-hq
scale=ewa_lanczossharp
deband=no

thassit babe

File: 1353960760003.png (513KB, 975x730px) Image search: [Google]

513KB, 975x730px

>>60832559
>>60832605
>>60832622
>>60832330
>>60832202
Just extracted the shader code for NGU

https://pastebin.com/QMc1E7fP

>>60833081
this is just like when my dad went to the store for some cigarettes

>>60834945
what anime is that picture from?

>button 5 and button 6
>Decrease/increase volume
this is out of date

>>60834945
cross compiling to glsl now

File: 1472033424928.jpg (197KB, 707x831px) Image search: [Google]

197KB, 707x831px

>>60831957
How do you get smooth playback on mpv? its fucking choppy

>>60834981
What is input.conf?

>>60835002
what about it?

# Audio
axis_up add volume 5
axis_down add volume -5

Is it possible to use vapoursynth with mpv on Windows? A quick search of the internet's preliminary results seem to be telling me no.

>>60834992
vo=opengl
tscale=oversample
interpolation
video-sync=display-resample
This'll enable smooth motion, which fixes judder without introducing soap opera effect.

There it is, decompiled NGU source for GLSL

https://pastebin.com/raw/F5kT5GwT

>>60835146
is this the best quality with smooth motion?

>>60835158
NGU AA, soft, standard and sharp? Or the old NGU before Madshi split it into four separate settings?

>>60835198
The latest NGU. It's packed in as a big .exe file resource that is compressed/encrypted. The files themselves are compiled directx shader object files.

NGU being called "NGUPass"
https://transfer.sh/FXs2f/NGUPass

File: you_dont_understand.jpg (67KB, 445x618px) Image search: [Google]

67KB, 445x618px

>>60835158
Based

>>60835158
>PrcSampler;
>RandomSampler;
what are they?

>WeightSampler;
trained model?

>>60835158
Translate this weebs.

Here's the "ShaderBank" file, decrypted and decompressed. Full of every hlsl shader-object that madVR uses
https://transfer.sh/YSWOg/ShaderBank.bin

File: Despair.png (52KB, 318x470px) Image search: [Google]

52KB, 318x470px

>>60834404
Figured I might as well try it out after watching animu, but I get a

[vapoursynth] Error filtering frame.

with mpv. I had to build vapoursynth since Debian Sid is shit like that and the multiple python versions gave me a headache. I'm not even sure if I placed the scripts where they needed to be, I just know I can import vapoursynth from the terminal.

>>60835230
>PrcSampler;
>RandomSampler;
>WeightSampler;
Cant find much that implies what PrcSampler is but Random and Weight sampler are probably related to a static noise map and what is probably an edge-detection weight map

Some shader names
https://pastebin.com/yWjiBeQS

>>60835319
Prc is the pixel sampler that actually samples pixels of the video frame
TexCoord possibly encodes both the xy of the current pixel and the width and height of the video frame so that it can accurately look up neighboring pixels and such

>>60835231
"ImmConst" seems to have a lot of constants that are the gaussian blur kernel
http://dev.theomader.com/gaussian-kernel-calculator/

>>60835230
Only PrcSampler seems to get used
I decompiled using
https://github.com/James-Jones/HLSLCrossCompiler/
so there's gonna be a lot of auto-generated garbage

File: madvr-ngu-std.png (879KB, 840x552px) Image search: [Google]

879KB, 840x552px

>>60834805
Best I could get in madvr

>>60835132
yeah but it's a pain in the ass so it's not enabled in the available builds
build it yourself with https://github.com/jb-alvarado/media-autobuild_suite

>>60835271
does it say more than that? could be a few different things

>>60835604
Too late for that, I went ahead and purged vapoursynth altogether. I will definitely give it another go tomorrow, I'm just too tired for this right now and would rather be watching animu as I am right now.

>>60830575
>Also no point in superres really anymore thanks to NGU.
NGU is superres, idiot.

>>60830170
>ptBR,pt
Urgh.

why doesn't osc-visibility work?
also can the extensions be combined into one [ ]

[extension.mp3]
script-opts=osc-visibility=always
stop-screensaver=no
[extension.mp4a]
script-opts=osc-visibility=always
stop-screensaver=no
[extension.ogg]
script-opts=osc-visibility=always
stop-screensaver=no
[extension.opus]
script-opts=osc-visibility=always
stop-screensaver=no
[extension.flac]
script-opts=osc-visibility=always
stop-screensaver=no

>>60835595
>>60834386
>>60830445
so, NGU ia better at upscaling at lower resolutions but above 1080p it starts to look like an oil painting?

>>60834992
try another --hwdec and/or --vo
or turn off your compositor

>>60836103
>artifacts and blur is better

What's the catch with hardware decoding? I am testing vaapi, and cannot tell a difference in screenshots. However, the image hashes (lossless PNG of course) are different between software and identical hardware screenshots, so obviously there is a difference in decoding.

>>60836409
N times less cpu usage

>>60836151
both playback speed and the picture

i want the smoothest playback with the best quality on mpv

>>60834945
Which NGU? There are a lot of variants.
Personally I would like to see what he changed to make NNEDI3 (aka NGU-AA) faster.

>>60833738
Fixed the crash. Still working on x11 resizing.

File: Nishikino-Maki.png (202KB, 500x281px) Image search: [Google]

202KB, 500x281px

>install madvr
>use nnedi3 because every website memes its good
>all that dropped frames

fuck this shit anon.
what are the settings for NGU to have memedi3 quality?

>>60836017
At a glance it looks like options inside of those auto profiles get set after the osc is already loaded. osc-visibility works in mpv.conf when it isn't in an auto profile. maybe submit a bug report.

>>60830662
I hate this argument. Nobody fucking cares, everyone watches movies differently. Some people will notice.

>>60836192
Compared to this >>60834386 shit, yes it is far better.

>All these people knowing all this stuff by heart
>I just have mpc-hc by itself

I don't know should I feel bad about it since I'm pretty much content as long as its not streaming quality.

>>60837716
NGU does better than nnedi3 even at low lmao

>>60836438
Please this

>>60835908
Educate yourself, dumbass.

>>60838204
kill yourself retard

>>60837716
No point in using nnedi3 currently. NGU is much faster. The best NGU for pretty much all types of content is NGU Anti Alias. Set chroma upscaling to NGU AA medium. Set image upscaling to NGU AA (luma doubling: medium, chroma: normal), leave everything else as "let madvr decide". You might also want to enable debanding (medium/high) and anti ringing. Also i suggest you to tick smooth motion + untick everything under "trade quality for performance".

I watch my Berserk from twist.moe.

File: 1398604795595.webm (3MB, 640x360px) Image search: [Google]

3MB, 640x360px

>>60838237
thanks my brother of african descent you really helped me out

>retarded shitposter posting his "challenge" again
>it's just a waifu2x upscale
>people still fall for it
I know most of /g/ is tech illiterate but come on

>>60838306
Its not waifu2x actually but still looks like shit.

>>60832968
Vulkan for windows when?

>>60838304
source

File: Screenshot.png (224KB, 800x800px) Image search: [Google]

224KB, 800x800px

>>60838237
why ngu is such a piece of shit?

File: mpv-shot0002.png (82KB, 800x800px) Image search: [Google]

82KB, 800x800px

>>60838339
This is how it supposed to look

>>60832988
madVR fags are retarded, if it shipped with warpsharp they would love that

>>60838352
>muh test pattern images
Post mpv result btw.

>>60836103
I have a 4K monitor, the "oil painting effect" only happens in a very visible manner when upscaling very low-res sources such as 480p. The truth about NGU is that none of its variants (AA, soft, standard, sharp) are really a one-size fits all upscaler and which settings work best can depend very, very strongly on the quality and characteristics of the source and how much you need to upscale.

For instance for 1080p > 4K upscaling with good quality sources Standard and Sharp are almost universally better than any other scaler out there and produce sharper results with pretty much no visible oil painting effect. Going from decent quality 720p > 4K it's largely the same story but the oil painting can become slightly more visible due to having to quadruple first. When you get to 480p > 4K though the source is really, really important and there can be significant differences in end result based on which variant you use and I even have a source where old NNEDI3 looks better than any NGU variant, even NGU AA which was designed to be a replacement.

>>60838384
>Post mpv result btw.
Here >>60838352
Check the filename fag

File: videogames.webm (3MB, 638x496px) Image search: [Google]

3MB, 638x496px

>>60838327
imouto paradise

WTF i have to download?

Windows builds by lachs0r https://mpv.srsfckn.biz
Windows builds by shinchiro https://sourceforge.net/projects/mpv-player-windows/files
Compilation instructions https://github.com/mpv-player/mpv/blob/master/DOCS/compile-windows.md
MSYS2 source package https://github.com/Alexpux/MINGW-packages/tree/master/mingw-w64-mpv

Can't they just make things simpler with ONE download option?

>>60838442
Just compile it you windows faggot.

>>60830755
But it does, but you gotta nigger rig the thing with Pulseaudio hackery that never works
Thankfully mpv can use lavfi, and thus we get a decent parametric equalizer and real HRTF (from SOFALIZER) thanks to ffmpeg
>>60830853
ER4S still needs equalization due to the inherent acoustic mismatch between the drivers and ears

>>60838453
Nah, I will just keep using VLC.

>>60838442
It's an idiot filter. If you can't at least figure out how to install it, you shouldn't be using it.

>>60829910
How do I get into creating video shaders?
I've got some hlsl experience.

>>60832968
Holy fucking shit
And fuck Intel for not supporting Vulkan in their Windows drivers for older iGPs

File: NGU AA hash - Copy.png (1MB, 1920x1080px) Image search: [Google]

1MB, 1920x1080px

>>60838339
But this is not NGU.
>>60838415
Who are you trying to trick, retard? It looks like original image haasn got test image from.

>>60834751
Everything he's doing it's done on vapoursynth, fucking madvr retard

File: Jinc hash - Copy.png (967KB, 1920x1080px) Image search: [Google]

967KB, 1920x1080px

>>60838415
>>60838352
>>60838339
>>60838515

>>60838515
>>60838542
Source is https://github.com/haasn/mpvhq-upscalers/blob/master/hash/original.png

File: mpv-shot0001.jpg (611KB, 1080x1080px) Image search: [Google]

611KB, 1080x1080px

>>60838515
>>60838572

File: NGU Sharp - Copy.png (1MB, 1920x1080px) Image search: [Google]

1MB, 1920x1080px

>>60838582
What are you trying to achieve?

>>60838321
When someone else writes it after it has been merged.

I'm fucking pissed off with how every fucking open source project does vulkan on windows support first and doesn't even FUCKING bother to write linux support, despite the fact its easier. And even if they write support shit is fucking hardcoded to use x11 functions and break wayland completely. I can think of rpcs3, ppsspp, and dolphin that do all this. No wonder they all hardly work at all, they're all written with c++ by windows people or in the case of ppsspp its worse, android people.

>>60838515
>But this is not NGU.
That is NGU standard

File: Lanczos SuperRes - Copy.png (97KB, 1920x1080px) Image search: [Google]

97KB, 1920x1080px

>>60838582
>>60838352
Looks like you applied super res on top of something. Here is Lanczos 4 taps. SuperRes 3. I wonder why Jinc produces such shit result in >>60838542

>>60838515
You can't really test such a small image upscale with NGU though. NGU is used for quadrupling at most, so you're actually seeing the output of NGU + whatever other scaler you have selected, could be Jinc AR by default. The input is 100x100, NGU can scale that to 200x200 and 400x400, anything other than those 2 exact resolutions aren't showing the output of NGU alone but of NGU combined with another scaler. It's a pretty retarded test case, nobody watches 100x100 video.

File: NGU Standard hash - Copy.png (1MB, 1920x1080px) Image search: [Google]

1MB, 1920x1080px

>>60838711
Pic.
>>60838752
I agree. Why Jinc produces such a weird result compared to lanczos? Look >>60838717 and >>60838542

Newfag here. I've just been using Bicubic like a pleb. Is NGU good and can I use it with a GTX 960 or is that too shit?

>>60838792
Use >>60838237 You also can set dithering to error diffusion 2.

>>60838442
I always download the one from sourceforge

>>60838824
Thanks for the quick pointers.

>>60838781
>Why Jinc produces such a weird result compared to lanczos?
No idea really. If you want to see something really funky try NGU Soft with direct quadrupling. Now that is some real weird-ass shit.

File: 1495389627406.jpg (1MB, 3840x2160px) Image search: [Google]

1MB, 3840x2160px

>>60834816
Not sure if bait or a poorfag still paying 4k off.
Either way made me laugh, have a (You).

>>60838306
Try his settings yourself before spouting bullshit, idiot.

Is it possible to implement enhanced ewa_lanczos?
http://kaba.hilvi.org/homepage/cg/ewa/Enhanced_Ewa.pdf

>>60839110
Yes.

>>60839110
Yes but who knows when are going to do it.

>>60839163
*when the devs are

>>60832157
I know that /g/ doesn't like to spoon-feed people, but can you please explain?
I don't want to switch to windows.

>>60832099
ewa_lanczos is better at dscale than mitchell
deband=yes
Turn down the antiring to 0.5

>>60838237
drops 20 frames per second, kek

>>60839110
Ask on IRC, but probably never.

>>60838499
Read the documentation of GLSL and you could play with http://glslsandbox.com/

>>60839573
Post specs and I'll help you with your settings

>>60839726
i7 4720HQ @ 2.60GHz
nvidia GTX 960m
16 GB RAM
>laptop
D-don't laugh at me senpai

>>60839769
You are using madVR right?

>>60839813
MPC-BR + mad VR + LAV filters. x64.
I just want to know what's the best I can aim for, because every time I try some configuration I have the impression I fucked up the ONE thing that makes frames drop

>>60839858
Try these settings:

Chroma upscaling:
Jinc

Image downscaling:
Jinc

Image upscaling:
NGU Anti-Alias

Algorithm Quality
medium
medium
let madVR decide

Active doubling/quadrupling
Let madVR decide
Let madVR decide

If any(more) scaling needs to be done
Jinc AR
Use image downscaling settings