/mpv/

File: Screen Shot 2017-02-24 at 5.11.34 PM.png (2MB, 1846x1087px) Image search: [Google]

2MB, 1846x1087px

>linux nerds will never have a frontend this nice

File: trashman6.png (591KB, 778x1018px) Image search: [Google]

591KB, 778x1018px

>>59102701
>io
>manual

File: 2017-02-24_231946.png (2MB, 1291x750px) Image search: [Google]

2MB, 1291x750px

This is pretty damn good!

>>59102985
font?

File: shot0009.png (2MB, 1920x1080px) Image search: [Google]

2MB, 1920x1080px

This is a shot of the same frame except from the raw untouched bluray video.

>>59102884
>not using baka-mplayer, the best mpv frontend
kys

>>59102701
Isn't this that shitty player that can't even handle seeking for ogm files?

>>59103029
looks like profont

File: ogm seek.webm (1MB, 576x464px) Image search: [Google]

1MB, 576x464px

>>59105231
works for me. I think some had issues when seeking via dragging. anyways, ogm is depricated and buggy to start with.

>>59102884

iina?

>>59105231
You should remux your OGM shit to MKV anyway.

>>59105443
Yeah okay... I'm going to waste my time remuxing video files so I can use some inferior autist media player. Or I could just use a media player that doesn't have that problem.

>>59105376
>576x464 monitor
kys poorfag

Haven't kept up with these threads in a while, my config still viable?

# Video settings
profile=opengl-hq
scale=ewa_lanczossharp
csale=ewa_lanczossharp
tscale=oversample
video-sync=display-resample
framedrop=vo
hwdec=no

# Audio settings
ao=alsa
audio-channels=stereo
softvol-max=300

# Language settings
alang=eng, en
slang=eng, en

# OSC settings
no-osc

# Screenshot settings
screenshot-format=jpg
screenshot-jpeg-quality=100
screenshot-jpeg-source-chroma=yes
screenshot-tag-colorspace=yes
screenshot-high-bit-depth=yes
screenshot-directory=~/Pictures/mpv screens/
screenshot-template=~/Pictures/mpv screens/%F - %p

# Window settings
keep-open=yes
force-window=yes
autofit-larger=90%x90%
autofit-smaller=500x500
geometry=50%:50%

# Miscellaneous settings
ytdl=yes
ytdl-format=bestvideo[vcodec=vp9][tbr<2000]+(bestaudio[acodec=opus]/bestaudio)/best
save-position-on-quit

# Output some video stats
term-playing-msg='Resolution: ${width}x${height}, Framerate: ${fps}'

# Extensions

[extension.jpg]
pause

[extension.png]
pause

[extension.gif]
loop-file=inf

[extension.webm]
loop-file=inf

File: mpv.webm (984KB, 1280x720px) Image search: [Google]

984KB, 1280x720px

>>59102701
reminder that mpv is vlc tier software

>>59105754
dunno
>hwdec=no
why though?

>>59105794
Seemed like what everybody used before

auto-copy

and when that became a thing I was on an older version. Is that still what's recommended?

>>59105794
Isnt software decoding generally better?

What did I do wrong?

screenshot-template="%F_[%p]_[%tH:%tM:%tS][%td.%tm.%ty]"
Error opening 'E:\mpv_ss\01 Pilot (HD)_[00_16_52]_[22:06:31][24.02.17].jpg' for writing!
Error writing screenshot!

>>59106327
Windows doesn't allow file names to contain colons

>>59106509
thanks bae

>>59105769
>downloading retarded files
"it's the players fault"

File: 1456828022761.jpg (117KB, 1440x1364px) Image search: [Google]

117KB, 1440x1364px

>>59105769
>ogm

I wish I was one of the extras at NERV that day.

What is the least performance intensive tscale?

File: pbs_twimg_com_media_C5OkRK1XAAAY.jpg (46KB, 451x558px) Image search: [Google]

46KB, 451x558px

>>59102701
Is this the new anime thread? What anime are you watching this season, /g/?

File: Capture.png (825B, 134x40px) Image search: [Google]

825B, 134x40px

I'm getting very occasional framedrops at a seemingly somewhat regular interval, like 1 per every 30 seconds. What's the go?

Also pic related, is this something to worry about?

>>59109116
Yep, the latter are late frames, so while only a few frames are being dropped, a lot of frames are mistimed. You might have chosen VO settings that are too intensive for your GPU.

>>59109102
Give me Little Witch Academia or give me death

File: Capture.png (24KB, 1267x780px) Image search: [Google]

24KB, 1267x780px

>>59109136
I'm still getting 1 late frame every 5 seconds even with this config, jfc am I going to have to disable interpolation even?

>>59109194
Maybe. My laptop isn't powerful enough for interpolation (though my desktop is.)

Does opengl-backend=dxinterop work for you?

>>59109204
Oh it turns out I had foobar encoding some audio files in the background, now I'm only getting one late frame every 15 seconds or so, and one dropped frame every couple minutes. I guess it's fine, but it still kinda triggers my autism. I'd like absolutely no late or dropped frames ideally.

It seems that madVR is still the best renderer but it's a Windows only software. :(

>>59110144
foozoor is still not allowed in these threads.

>>59110158
Who is this guy?

>>59102884
How do I achieve these levels of stylishness on a Mac

File: main~01.png (751KB, 1175x710px) Image search: [Google]

751KB, 1175x710px

>>59110211
Use the iina frontend:
https://lhc70000.github.io/iina/features/

why using v4l2 input uses high cpu on mpv? only started a few git commits ago (like 5 days)

>>59111337
Probably hwec issue produced by a gpu driver update.

File: poor.png (399KB, 1058x513px) Image search: [Google]

399KB, 1058x513px

Why is this still open?
https://github.com/mpv-player/mpv/issues/3863

Are they interested to implement a similar algorithm or not?

File: anime.jpg (775KB, 1280x960px) Image search: [Google]

775KB, 1280x960px

>>>/a/153893860

What do you think, lads? Is mpv the best anime machine?

>>59111470
>Are they interested to implement a similar algorithm or not?
I wouldn't get your hopes up. None of mpv's core developers are interested in algorithms like NGU that increase sharpness at the cost of adding ringing and ugly oil-painting artefacts. It will only happen if an external contributor or shader writer implements it for mpv. I don't know why that issue is still open, but it probably shouldn't be.

>>59111495
What is that? Some delicious VHS source?

>>59111669
It's just glitch art as far as I can tell. It looks like they've done some pixel sorting on top of the VHS filter, so it's not really authentic.

>>59105899
Yes.

Why can't I take a screenshot while using dxva2?

>>59112046
mpv doesn't support it. With hwdec=dxva2, the video image never touches main memory. Taking a screenshot would require hwdec-specific code to copy the image from GPU memory to main memory. This code exists for d3d11va (Windows 8.1 and 10 only,) but no one has written the equivalent code for dxva2. It's a valid feature request, if you feel like asking on GitHub.

>>59111394
i'm not sure what you mean by hwec

the thing is, i play in mplayer:
mplayer tv:// -tv driver=v4l2:width=640:height=480:device=/dev/video0:alsa=1:adevice=hw.2:amode=1:input=1:immediatemode=0:forceaudio=yes:norm=ntsc -fps 30 -ao alsa

and it's fine (30% or so cpu)

I play in mpv:
env __GL_THREADED_OPTIMIZATIONS=0 mpv --tv-driver=v4l2 --tv-device=/dev/video0 --tv-input=1 --tv-norm=ntsc --tv-width=640 --tv-height=480 --tv-forceaudio=yes --tv-alsa --tv-adevice=hw.2 --tv-immediatemode=no --tv-amode=1 tv:///

certain options (such as --tv-fps=30) lower mpv's cpu by about 10%
cpu at around 110-140%
(env __GL_THREADED_OPTIMIZATIONS=0 is needed due to nvidia driver, without it, cpu goes to 225% on mpv)

>>59112046
>>59112156
hwdec=dxva2-copy should support screenshots I think.

>>59112169
-ao alsa or pulse makes no difference in mpv either, but without sound, it does lower cpu to 30-40%

>>59112156
Thanks for the explanation. I wanted to know I wasn't doing something wrong. Copyback does work, and ctrl+s always works. Alt+i works normally with dxva2 on MPC-HC.

>>59111552
Or maybe they wait for a new madVR shaders leak. ;)

I am getting old and I like to re-watch oldies and this kind of algorithms are really spectacular. I know the main variant of NGU is for good 720p/1080p sources only because there is no anti-ringing/artefact pass and is too sharp to produce a good result.

My preference for scaling:
1) madvr ngu pixart wip (3)
2) madvr jinc + ar
3) mpv ewa_lanczossharp + adaptive-antiringing.glsl
4) nnedi3 (I can't run it without dropped frames)

For me madshi made a better work but in play time it's almost unnoticeable.

>>59105769
>American Horror Story

>>59102701
>"""""""official"""""" builds by leet anonymous hackers such as lachs0r, shinchiro, and stolendata
>no madvr
>no xysubfilter
>no proper frontend
>counter intuitive settings defined in text file like its 1989

hows your guys player handling h.265 video playback??

kek!!!

>>59113341
>madvr jinc + ar
it's not better than ewa_lanczos + sigmoid-slope=10.0

trying to build mpv with cuda support in arch, configure shows no CUDA support, but I have all the cuda packages:
cuda
pycuda-headers
python2-pycuda
python-pycuda

what else is needed? i also have ffmpeg-nvenc which is supposed to have full cuda support

>>59113735
Never notice the sigmoid-slope option, is it that useful to reduce ringing?

>>59114881
Yes.

>>59114982
Is it the same for other scalers like lanczos or spline36?

>>59115025
Probably, I only tested with spline36 and ewa_lanczos though.

>>59115025
>Is it the same for other scalers
Yes, as long as you use opengl-hq

>>59115196
Please create an issue on github.

I'm on Win7, today I switched out to Classic Shell from Aero and started getting mad screen tearing. What gives? Went back to Aero and it's gone

>>59110167
Doom9 flammer that spams the madVR thread with false flags for both sides

>>59112937
The leaked shaders were implemented by a third party

tfw just want something that just works and got bored of learning whatever stupid language this config file uses

i just update it with brew from time to time and let it be. too lazy to customize it now that I actually have a job and shit to do

inb4 brew doesnt update it but brew cask does, i dont care anymore

Is Adaptive Sharpen preferable to FineSharp?

>>59118322
Yes! Use the single pass one from igv.

>>59118322
and use it with profile=opengl-hq

>>59118336
>>59118378
Thanks.

>>59117747
Does it mean converted shaders are available somewhere?

File: 128860.jpg (2MB, 2806x4208px) Image search: [Google]

2MB, 2806x4208px

Except ten guys on doom9, nobody uses or want to use madvr.
This closed-source, windows only and shitty directshow filter should die.
I am sure his ngu is based on protected university papers.
Yes it's better than nnedi3 but damn it is not that hard to do.

>>59118818
spotted the false-flagger

>>59118242
Honestly only reason why I am still on mpv is because of youtube-dl

The devs are rather arrogant about every request they get, and as of late madvr+potplayer just have better features and an easier way to edit settings, on here you don't need a fucking script just to make the player be on-top while playing, only thing that It might lack is the way it renders subs.

Is there any reason as to why downscaling with mpv looks so blurry?

http://i.imgur.com/5iwWrKn.png - mpv
http://i.imgur.com/bUxy8PG.png - madvr

Yes, I posted something like this before

>>59118818

madVR is the best, used by wayy more people than the 10 freetards that use mpv

File: lel.png (152KB, 1713x810px) Image search: [Google]

152KB, 1713x810px

>madVR

File: Screenshot_2017-02-26_00-20-40.png (1MB, 1920x1080px) Image search: [Google]

1MB, 1920x1080px

how am i supposed to watch anime like this

>>59119284
Change your dscale setting to something sharper...

File: screen3.webm (2MB, 1000x562px) Image search: [Google]

2MB, 1000x562px

So anon, how advanced is your webm script?

>>59120196
I tried everything.. and it was still not as detailed/sharped

My guess is that mpv uses a different method to downscale

>>59111331
If it's open source and on GitHub why won't Linux users ever have a frontend that stylish? We could rip it off by tomorrow, nigga.

>>59113720
Works just fine, no idea what you're talking about.

>>59120545
Mind sharing?

sup nigs

getting a laptop with Kaby Lake soon.

what decoder settings should I use to play H265 content that'll give me almost no cpu usage like in this video:

https://www.youtube.com/watch?v=cI-a4WZWwZc

MPV sucks fucking dog dick compared to MPC-HC with KCP.

>>59111331
This is actually pretty nice looking. The one time I'm a little jealous of macfags.

>>59123144
y

>>59120875
correct-downscaling?

What video codecs do you guys use for youtube-dl?

>hwdec on
>can't write screenshot
>can't deinterlace
why does this happen

>open video url in mpv
>pause video
>resume video
>mpv can't start redownloading the video
>video stops playing after the buffer runs out

>>59121634
I may github this when I get it absolutely perfect.
Any features I haven't realized yet?

>>59111495
actually, ngu doesn't add ringing at all, instead it removes it from the source. so with ngu you have less ringing than with ewa_lanczos.

you can get rid of the oil-painting look by softening the image and adding grain. softened with grain it doesn't look like an oil-painting, anymore, and it's still sharper (better focused) and less aliased than ewa_lanczos.

however, with bad sources, ngu sucks, because it tries to interpret compression artifacts as detail. that's where the new ngu pixart variant might come in handy, but it's still in development.

>>59113735
sigmoidal scaling helps with some ringing types, but makes others worse. same with aliasing. also, sigmoidal scaling changes the thickness of the interpolated lines. some lines get thicker, others thinner. it's not a free lunch.

File: mpv-shot0006.jpg (88KB, 1267x678px) Image search: [Google]

88KB, 1267x678px

Anybody using the converter scripts?
How do I make it so it uses the cached data in RAM instead of reading from HDD?

>>59118322
not necessarily. adaptive sharpen bloats, finesharp does not. if you do proper deconvolution, gray blurry lines become sharper, darker and thinner. adaptive sharpen makes the borders of the line outlines sharper, it may also make the line darker, but it doesn't think the line. finesharp does. so in this aspect finesharp is nearer to true deconvolution. at least for the usual line types...

File: Sleep-doesn-t-help-if-its-your-soul-that-s-tired.jpe.jpg (79KB, 800x800px) Image search: [Google]

79KB, 800x800px

I'm trying to play this file

https://hastebin.com/qanebifagi.pl

With mpv, but I'm getting this

[ffmpeg] https: HTTP error 403 Forbidden
Failed to open <THE URL>

But I can still download and even open the file in browser.
I tried spoofing a browser user agent but it didn't work as well.
Any ideas?

>>59118474
avisynth scripts

>>59118818
>ngu is based on protected university papers.
ngu is composed of a bunch of post processing shaders


>>59127866
> softening the image and adding grain. softened with grain it doesn't look like an oil-painting
>literally adding grain
It looks like an oil-painting dipped in sand instead

>>59127910
>sigmoidal scaling helps with some ringing types, but makes others worse
now that's some real bullshit

>>59127966
>adaptive sharpen bloats
bullshit again, seems like you judge by hlsl version

>>59127938
This would be amazing for streams. Anyone know how to do it or if its possible?

>>59128132
> ngu is composed of a bunch of post processing shaders
how would you know? seems unlikely to me.

> It looks like an oil-painting dipped in sand instead
depends on the amount of grain, obviously. grain isn't evil. you know that mpv's deband also adds grain to the image?

>>59128146
> now that's some real bullshit
no, it's not. some time ago i played with the sigmoidal options and while they made some ringing and aliasing artifacts less obvious, in other images they made things worse. one image i remember was an image which already had heavy (white) ringing in the source. with sigmoidal turned on, the white ringing got more aliased and stronger, compared to sigmoidal turned off.

> seems like you judge by hlsl version
true.

>>59128449
now try with sigmoid-slope=10.0 in mpv and see how there will be almost no ringing

>>59128445
I'm using both of the converter scripts the ffmpeg one works really great with youtube/other sites, but downloading the files would take long since it seems to convert the file as it downloads it.

The other converter doesn't work, it create the log file but stalls.

>>59105216
Didn't the dev ditch it long ago and replaced it with a rewrite in QML called mochi-player...?
Also, it's far, faaar away from IINA regarding functionality/features. You can configure mpv completely through GUI with IINA which is a huge normie boost. My father is using it now instead of VLC.

>>59128449
>>59128470
and also at least white dots/lines are thinner and sharper

>>59108956
oversample. Next best thing is linear (more efficient version of triangle). The others are all way harder.

>>59128449
>how would you know? seems unlikely to me.
madshi has been working on shaders for over a year, he has been become pretty good at tweaking them. it doesn't take a genius to guess what is doing now. given that he has different variant for pixel art and normal, it's evident using different combinations.

>grain isn't evil
oil painting still is. whore dressed in expensive dress is still a whore.

>>59121158
>If it's open source and on GitHub why won't Linux users ever have a frontend that stylish? We could rip it off by tomorrow, nigga.
Well, first of all, I'd bet all I have that _you_ are not able to do this, not at all.
The people who are able to probably don't care as they don't need it.

Other than that, IINA is written in Swift and Swift is supposed to be open source so it should run on Linux too: https://swift.org/blog/swift-linux-port/
However, not sure if the GUI calls/API is portable.

>>59128585
i think you probably mean comic look and not oil painting look? both exist, but i think ngu produces comic look, not oil painting look.

oil painting look means you get very weird directional distortion artifacts, like you're on drugs or something. i've not seen ngu do that.

comic look means you get sharp edges and no texture detail, which makes the image look like a comic image instead of a real image. ngu does that.

File: clown4xNguSharp.png (2MB, 768x896px) Image search: [Google]

2MB, 768x896px

>>59128713
>oil painting look means you get very weird directional distortion artifacts, like you're on drugs or something. i've not seen ngu do that.
Recheck the clown image.

>>59128764
where do you see "very weird directional distortion artifacts" in that image?

File: plz.png (919KB, 768x896px) Image search: [Google]

919KB, 768x896px

>>59129534
Are you joking? Oil-painting effects are obvious everywhere

>>59129835
in my book this is comic look (sharp edges, flat textures), not oil painting. oil painting is more like these:

http://www.general-cathexis.com/interpolation/clown4XAqua.jpg
http://www.general-cathexis.com/interpolation/ASBlowupClown1.jpg

but i suppose there are better things to do for both of us than to discuss the definition of terms. so let's agree to disagree on the terms.

fact is, whenever you upscaling images by a large factor, there can't be any real texture detail. it's simply not possible to reconstruct texture detail in a good way. so all any upscaling algo can do is to try reconstruct the edges as well as possible. doing that will automatically approach the comic look (or maybe you call it oil painting look). i don't really see how that can be avoided at these scaling factors. unless you make everything soft as hell. obviously you can blur the ngu image to make it softer.

imho ewa_lanczos looks a whole lot worse at such scaling factors, with or without sigmoidal options.

Is blurred mess better than oil painting?

I begin to prefer oil painting but this is only because it is now (ngu low/medium) less gpu intensive than to produce blurred mess.

blurred mess = ewa_*
oil painting = neural scalers like nnedi3 or ngu

>>59130158
Shape of thin objects, branches and shadows are distorted. NGU sharpens edges and flattens the textures but can't fix the shape of objects. Jinc + sharpen looks much natural in the most variety of content.

>>59130347
same here. i think normal ngu untouched is nice for max 2x upscaling. for higher upscaling factors post-blurring and some grain is needed, though.

>>59130401
can you post the clown with jnc + sharpen with parameters you like? to my eyes it usually looks worse, due to bloated lines, aliasing and ringing artifacts.

maybe you are less sensitive to bloated lines, aliasing and ringing artifacts because we're so used to these problems. so you might find them less objectionable than whatever problems you see with the ngu image. but bloated lines, aliasing and ringing *are* pretty ugly artifacts, too, imho.

would be interesting to do a blind comparison with the best ngu options we can find and the best jinc options we can find and then let a couple hundred/thousand users vote which they prefer.

>>59124597
same thing

okay, how do I PERMANENTLY DISABLE fullscreen in mpv?
I mean I don't want it ever to go fullscreen, and as for now it sometimes decides to go fullscreen when I move mouse to another monitor.
fs=no and fullscreen=no in mpv.conf doesn't seem to work

Topic slightly unrelated but...
Who I solve video flickering in videos encoded with h264 on Chromium based browsers?
There's a solution?

>>59133227
>Topic slightly unrelated but...
>Who I solve video flickering in videos encoded with h264 on Chromium based browsers?
>There's a solution?

>Slightly unrelated

It is a video. You use a video player for that. Not an internet browser.

>>59118818
kill yourself foozoor
'>>59120545
honestly this is too much work per webm. Or does it use sensible defaults?

>>59134099
What's wrong with this foozoor clone?
He is fucking right. No?

File: 2017-02-26_23-53-27 00.00.06.900-00.00.38.833.webm (2MB, 1248x768px) Image search: [Google]

2MB, 1248x768px

>>59134099
That's the advanced mode, showcasing all the major features such as multiple segments, cropping, CRF/bitrate/target size, scaling and so forth. Naturally, you can pre-define your settings in a .conf file as is tradition.

The "basic mode" is just using Ctrl+W twice, seen here. The Windows side is a bit uglier, I have to create a temporary .bat to chain commands (such as renaming the _temp away, mkvpropediting in the title, removing 2-pass logfiles...)
I can't hide the console, but it's not that bad - you get to see the progress! Windows and Linux are otherwise equivalent for the script.

Just got a video card

what are good MPV settings? Windows btw

>>59120545
>>59136274
noice

scale=ewa_hanning
scale-radius=4
scale-window=gaussian
scale-wparam=1.41

hmm

>>59136851
What's the purpose of this??
Is it better than scale=haasnsoft?

>>59119284
Still looking for an answer

>>59137037
perhaps posting your config would help get the ball rolling...

>>59137037
What's your mpv config?
Did you try with profile=opengl-hq?

>>59137037
Post your config first.

>>59136990
It's sharper than ewa_hanning but retains the property of not aliasing correctly-antialiased (i.e. antialiased with a filter that retains high frequency phase information, specifically not a box averaging filter) 45 degree angle edges.

>>59137254
OK! I created another profile in my mpv.conf.

[sharper-than-ewa_hanning-but-retains-the-property-of-not-aliasing-correctly-antialiased]
scale=ewa_hanning
scale-radius=4
scale-window=gaussian
scale-wparam=1.41

Vulkan when?

>>59137371
Probably never or maybe haasn is preparing a huge surprise.

>>59113825
anyone?

>>59137217
>>59137102
>>59137101

profile=opengl-hq
opengl-backend=dxinterop

correct-downscaling=yes
linear-scaling=yes

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

I've used every scaler.

>>59137327
good filter name

>>59137530
Try
dscale=sinc
dscale-radius=0.5

>>59137530
>linear-scaling=yes
Don't do that. It makes downscaling luma perceptually more aliased.

>>59126467
use auto-copy or disable hwdec

>>59137694
I also didn't had that before-hand neither correct-downscaling, is just things I've been told to try.
>>59137675
https://mpv.io/manual/master/#options-lanczos
http://i.imgur.com/VryprMB.png - mpv
http://i.imgur.com/8f4O7R6.png - madvr

>>59127659
did you fork the other guy's script?
lemme webm streams

>>59137762
Second image has mild ringing, which is why it looks sharper than the first.

>>59137762
Sharper doesn't mean better, maybe you should educate your eyes.

>>59137762
If you downscale with dscale=sinc then dscale-radius=# will be available, which you can use to fine-tune the sharpness. Beacuse correct-downscaling works weird, dscale-radius will also work weird. Try things like 0.5, 1, and 2. If 2 isn't sharp enough, then you want too sharp an image.

mpv support linked mkvs yet? My next release uses ordered chapters, but the one after is back to linked mkvs.

>>59137897
linked mkv was a mistake

>>59137813
I don't know ma dude it still looks better
>>59137786
I think it's due to the different debands
http://i.imgur.com/OUI68Ab.png - mad
http://i.imgur.com/o4gHb2k.png - mpv

in every downscaled shot I've taken mpv seems to be way too blurry to the point that it feels like there are some details missing.

But I guess I am blind to "ringing".

>>59137910
linked mkvs are comfy

>>59137919
Gee, did you try disabling debanding?

>>59137975
I did, on both on that recent screenshot

>>59137975
Also to add I am using default madvr settings
http://i.imgur.com/dstCz5U.png - mad
http://i.imgur.com/blLdjFV.png - mpv

Whats funny to me is how this doesn't happen when you fullscreen.
http://i.imgur.com/vVB931Y.png - mad
http://i.imgur.com/KNltQHy.png -mpv
Can't even tell the difference

File: file.png (36KB, 570x493px) Image search: [Google]

36KB, 570x493px

>>59138062
The madvr image is oversharpened.

File: [HorribleSubs] Hibike! Euphonium S2 - 07 [1080p] 00.19.36.885-00.19.44.685.webm (712KB, 554x844px) Image search: [Google]

712KB, 554x844px

>>59137770
No. Yes, I'll admit I took inspiration from it (and it made me realize that holy shit you can Lua mpv), but I rewrote the entire thing as I didn't like neither the dependency for YAD or the look of the code.

Youtube et al (YTDL) work (anything that can be seeked with --start), but clipping livestreams don't and most likely won't. I don't see any reasonable way to do that with any sort of accuracy; Linux side coouuld spawn an instance to just start encoding and the script could stop it with another keybind, but there'd be a few seconds of delay for the start and I'm not too sure about the stop. At best it'd be another script entirely; you'd be better off saving the stream by other means and clipping it afterwards.
Thanks for the suggestion, though!

>>59138100
Are we playing the stupid game of zooming in on a picture game?
http://i.imgur.com/ym6RwKc.png
http://i.imgur.com/qFCK522.png

>>59138183
I don't have to zoom in to see it. I have to zoom in to show you.

Oversharpening is the reason madvr is sharper. That's the answer to the question you asked.

>not watching vhs in current year

you fucks arent even trying

>>59138226
bitch did you just assume I don't have a hardware setup to stream VHS with a video device

>>59138195
I guess madvr sharpens out details and mpv blurs them

Time to move on

>>59138252
Yes, downscaling always blurs things. madvr is sharpening the image in addition to blurring it. mpv is only doing what is necessary to downscale.

>>59138273
Even with downscalers that are supposed to sharpen the picture?

>>59138297
Yes. They have to lowpass the image so that the downscaling process doesn't cause aliasing. If they don't lowpass the image, it's going to look like very bad pixel art. lowpass == blur.

Lowpasses with a sufficiently sharp frequency response will cause spatial ringing. For audio this is correct, for images only a small amount of ringing is ever correct, and madvr is adding too much ringing.

>>59138334
audio has scalers now?

>>59138383
Do you know what resampling is?

>>59138389
You are comparing resampling audio to a picture/video scaler?

I mean it's bad enough that you compared lanczos vs cubic and zoom in on pictures to "make a point", but I think I am done here.

>>59138408
Why are you posturing?

>You are comparing resampling audio to a picture/video scaler?
They are mathematically identical.

File: 1486608335259.jpg (422KB, 3300x1600px) Image search: [Google]

422KB, 3300x1600px

>>59119523
>tfw he made all that gibberish up to ruse mpv babbys

>>59138475
>madvr fag in denial

>Dump of madVR shader
https://0x0.st/eqp.txz
>Enhance detail ported for mpv
http://pastebin.com/4WBPQ7Cq

>>59138634
Is it better than LumaSharpenHook?
Are there other ported shaders?

Is there a general config that is agreed upon for best quality video or is there too much autism to agree?

>>59138831
profile=opengl-hq
Thats honestly all you need. for better performance you might want to add other stuff depending on your OS.

>>59138897
Cool that sounds simple. Performance isn't too much of a concern, ~5% cpu usage and trivial ram and gpu usage with opengl-hq and the largest of files I have. Performance is already a lot better than the previous setup I had (kcp.)

I don't have a CPU with SSE4x instructions. Does hwa-copy really need them for best performance?

>>59139162
Best performance doesn't matter, just good performance.

>>59138123
is your scirpt in the mpv wiki?

haven't updated in a year, mid range computer, anything i could look into?

>vo=opengl-hq:interpolation:tscale=linear:scale=ewa_lanczossharp:cscale=ewa_lanczossoft:user-shaders="~~/FineSharp.glsl"

>>59102701
Hi. While you faggots are compiling and figuring this shit out, I'm watching loli on VLC.

That will be all.

HOLY SHIT I LOVE THISPLAYER SO MUCH
I HAD NO IDEA /G/ LOVED IT TOO

>>59102701
>720p
Bitch, The End of Evangelion DEMANDS to be seen in 1080p minimum, while we wait for a 4K version to be released that will blow our minds. How dare you.

>>59130347
>Is blurred mess better than oil painting?
Yeah, you can create the illusion of detail by sharpening it and adding some monochromatic noise to the blurred mess
Oil paintings look as unnatural as it gets
>>59141253
EoE was done on 16mm film, you ain't going to get enough resolution for 4K, 1080 is already a bit overkill

i use mpv with everything default

Anyone have a better autoload.lua that works if the filenames are like
1
2
3
4
instead of
01
02
03
04

Because I'm watching The Rose of Versailles and the encoder has the episodes like that and its annoying.

>>59105769
Why is that movie playing in your presence. The Movie Gallery closed a long time ago.

>>59142383
Why not just rename the files then?

File: akko.png (296KB, 278x259px) Image search: [Google]

296KB, 278x259px

>>59120545
>1000x562 50+second
>2.1mb

How even possible

>>59142726
His picture is mostly static. I have a 1080p clip that's 50kbps.

>download mpv
>it just works

how much am I missing by not having a custom config?

is it the best player for windows as well? I added an IPS monitor to my gaymen desktop so I might watch movies on it rather than my loonix laptop

>>59142784
Just install K-lite standard.

>>59142651
Because it would take a while and I'd just like to know if a better version is out there.

>>59142901
It's a one liner in bash.

File: ngu.png (3MB, 1280x960px) Image search: [Google]

3MB, 1280x960px

>>59130716
>would be interesting to do a blind comparison with the best ngu options we can find and the best jinc options we can find and then let a couple hundred/thousand users vote which they prefer.
NGU

File: mpv-shot0002.png (2MB, 1280x960px) Image search: [Google]

2MB, 1280x960px

>>59130716
>>59144054
haasnsoft+SSIMSuperRes(removed some ringing)+adaptive-sharpen(1-pass at strength 0.5)
The only obvious difference is that mpv looks more bloated, but from the distance it seems more sharp because of that

>>59144054
>>59144075
thank you, that's a good comparison!

SSIMSuperRes + adaptive-sharpen are good filters, i think. but i have to give the edge to ngu here. e.g. the hand looks remotely natural in the ngu image. while in the haasnsoft image it looks a bit "waxy", similar to some movies (e.g. some star trek blu-rays) look like when the movie studio applied too much dnr. i'm definitely not a fan of bloating, so i prefer the ngu look - with this image at least.

you don't happen to have the original high-res version of this image for comparison, do you?

what do you other guys say?

File: mpv-shot0001.png (2MB, 1280x960px) Image search: [Google]

2MB, 1280x960px

>>59144075
>>59144280
fixed mpv screenshot

File: 1.jpg (247KB, 640x480px) Image search: [Google]

247KB, 640x480px

>>59144280
>you don't happen to have the original high-res version of this image for comparison
no, only this low res

File: Shinryaku! Ika Musume - 02 00.02.33.862-00.02.34.821.webm (770KB, 1920x1080px) Image search: [Google]

770KB, 1920x1080px

>>59140014
No, see >>59121634

Eventually, though; see, because I'm a dickwad I want something out of this. Having shown this to people and getting "...so?" in return has jaded me to not just release hours of my work for people to shrug their shoulders at. So, forgoing the "audience", I want to make this absolutely perfect, add everything I could ever imagine! and then github it with a nice README.md and stuff.

File: nguPixArtSuperRes.png (3MB, 1280x960px) Image search: [Google]

3MB, 1280x960px

>>59144333
thx. for comparison, here's the result with the ngu pixart test build + superres 1.

>>59144316
What about just using haasnsoft or ewa_lanczos without all these filters?
You should also create a page on the mpv wiki because 4chan is not the right tool to demonstrate things.

I think the main advantage of ngu is the power it needs to run, the low and medium versions are faster when doubling/quadrupling than ewa_*/jinc scalers and are much faster than using ewa_* scaler with ssimsuperres and adaptive-sharpen.

>>59146783
>are faster when doubling/quadrupling
yes, but with other scaling factors ngu probably won't be faster than ewa_* scaler with ssimsuperres and adaptive-sharpen

>>59146783
>What about just using haasnsoft or ewa_lanczos without all these filters?
http://screenshotcomparison.com/comparison/201787

>>59146844
Correct.

>>59146873
I prefer the mpv look, what scale option did you use exactly?

>>59146844
I only use NGU with doubling/quadrupling with the help of madVR profiles.

>>59146985
spline36 (same as ewa_lanczos but with more aliasing)
That's not mine comparison.

>>59147025
OK, it's the raw opengl-hq.
That's a weird result, ngu is not that noticeable in this example.

File: ewa_lanczos.png (2MB, 1280x960px) Image search: [Google]

2MB, 1280x960px

>>59146783
>What about just using haasnsoft or ewa_lanczos without all these filters?

What is a good adaptive sharpen strength for chinese cartoons?

>>59105769
>.ogm
>l..l..look guys its shit player!

>>59142791
>K-Lite
HOW ABOUT NO.

>>59147172
0.5-0.8

>>59147153
Clearly the best one!

>>59147251
yeah if you like ringing and aliasing

>>59147251
>>59147270
oh, and blur of course

>>59147270
Ringing is part of the original picture, aliasing is more problematic but it's better than the oil painting effect.

>>59147172
I use 0.4

>>59105769
Reminder that your video file is probably corrupt, and you should've checked to see if MPV was giving any errors in the terminal/console.

>>59147251
if you like soft, try the "nguPixArtSuperRes.png" one from a couple posts above. that one has a look which is somewhat similar to plain ewa_lanczos, but with much less aliasing problems, and with a little bit more sharpness. it's made with the work-in-progress "ngu pixart" algo (followed with a bit of superres), not with the regular ngu.

>>59147374
It happens with ogm files when you drag the seek bar. using the keybind or clicking is fine though >>59105376

>>59147296
>Ringing is part of the original picture
Why this >>59144316
pic doesn't have it then?

>>59147571
So what's your favourite one?

>>59147436
My point is though - if you have an issue, report it. It's super easy, and you can help fix the software - because it is open source.

>>59147905
https://github.com/mpv-player/mpv/issues/2074

GT710, Windows 10
ANGLE or dxinterop?

>>59148444
dxinterop

>>59132926
simple

disablefullscreen.lua

mp.observe_property("fullscreen", "bool", function(name, value) mp.set_property_native("fullscreen", false) end)

>>59148560
problem is with dxva2 (dxinterop) colors look washed out and d3d11va (ANGLE) look way better

>>59142383
Playlistmanager has a similar feature that on linux uses the natural order of files. On windows you can create a playlist with your files and use the scripts alphanumeric sort.

>>59148599
Use dxva2-copy.

>>59148567
>simple
yeah, no

i disabled the --ontop switch and for whatever reason it stopped going into fullscreen automatically so it works for me

thanks for your input though, i didn't know that you could write scripts like that

>>59149075
not possible with a CPU without SSE4.1 instructions.

File: Capture.png (73KB, 1135x775px) Image search: [Google]

73KB, 1135x775px

>>59149195
If you want to fix the colors you have to change this to nvidia settings and full range.

>>59149251
awesome! thanks. Now they look exactly the same

how the fuck mpv nows that some spaces are parts of the filenames and some are borders between filenames
when you execute mpv * and filenames have spaces it will work
when you pass same list as an argument it won't work
I just wanted to make simple shuffle wrap

mpv $(find "$1" -type f | shuf)

but it doesn't work
pic rel proof for spaces in *

File: Screenshot_2017-02-27_19-44-06.png (58KB, 1293x161px) Image search: [Google]

58KB, 1293x161px

>>59150361
forgot pic

>>59150361
it turns out that in bash if value goes through variable all enters are changed to $IFS value which default is space, changing it to newline made my script working

>>59147172
0.3

- mpv has advanced custom shader system
- igv ported ssimsuperres and it works better than superres from madvr
- igv ported adaptive-sharpen
- igv made adaptive-sharpen faster and better than the madvr one
- igv made adaptive-antiringing that works better than mpv one
- bjin and haasn ported nnedi3 and superxbr
- mpv has better jinc based scalers
- mpv has better color management
- mpv has advanced extension/script system for hackers and power users
- mpv is cross-platform and fully open source

So why should we keep using madvr?
Is ngu the only one reason?

>>59152496
This is too much try with 0.0015, cartoon shits don't need sharpen.

>>59153470
>cartoon shits don't need sharpen
The just don't use it,

Any similar addon for mpv without third party software?
https://chrome.google.com/webstore/detail/potplayer-youtube-shortcu/cfdpeaefecdlkdlgdpjjllmhlnckcodp

File: 1485088676574.jpg (85KB, 731x657px) Image search: [Google]

85KB, 731x657px

How do I play/pause MPV while it doesn't have focus?

I want to be able to pause during cutscenes without tabbing out

>Download mpv for Windows
>Want to configure keybinds and OSD
>No configuration files anywhere
>No configuration UI
>Delete mpv

>>59155800
https://mpv.io/manual/master/#json-ipc
could use this

>>59155938
Nice blog.

>>59155947

How do I use this exactly?

>>59155995
i linked to the manual...

>>59156015

I'm gonna level with you
I have no idea how I'm supposed to use any of this, I just want to press a button and have it pause and I don't see any anywhere there a hotkey is involved

File: a.webm (190KB, 614x626px) Image search: [Google]

190KB, 614x626px

>>59156091
it's a socket based IPC
aka, a way of sending commands to mpv from other programs
you could make a hotkey run a command that sends commands to mpv's IPC socket

>>59141253
I bet you're the kinda dude who also thinks megapixels dictate if a camera is good or not.

>>59155938
Look at the one link in the OP.
It tells you what you need on the first page.

>>59142901
>Because it would take awhile
>Not knowing how to write a simple script to automate this
>Not even at least being able to Google "auto renaming program for Win/OS X/Linux"

Cmon, Anon

post configs

>>59156192
what did you use to record this?

>>59157212
simplescreenrecorder to grab a quick x264-encoded intermediary, then ffmpeg to compress that further to a webm

>>59157237
thanks famalamadingdong

What's the equivalent of the Mpv for Android?

>>59157272
mpv

>>59144619
I'll suck your dick if I have to.

>>59138123
Would it be possible to make a script that saves the last minute of a livestream from the cache onto your hdd? Then you could just make the webm from that file instead of having to write what could be hours of 1080p 60fps footage to disc on the off chance something cool happens you want to clip.

>>59153341
- where's proof that igv's ported ssimsuperres is better than madvr's superres? i doubt it, at least mpdn's ssimsuperres is worse than madvr's superres, imho
- where's proof that igv's adaptive-sharpen is better quality than the madvr one? e.g. madvr has an anti-bloating and anti-ringing filter for adaptive-sharpen that mpv doesn't have
- adaptive-antiringing might be better than mpv, but still worse than madvr
- nnedi3 in mpv is slower then in madvr, and now gets beaten in both speed and quality by ngu pixart
- super-xbr in madvr has better anti-ringing compared to mpv and mpdn
- where's proof that mpv jinc is better in mpv? thanks to better anti-ringing, madvr's jinc is probably better, imho
- madvr has a dedicated test pattern generator for calibration, and is directly supported by calman, argyllcms, displaycal, hcfr and lightspace, not sure how you can claim that mpv would be better, which calibration software directly supports mpv?

i agree with:
- advanced extension/script/custom shaders
- cross-platform and open source
- probably faster adaptive-sharpen

there are some mpv advantages you didn't mention:
- mpv has adjustable frame interpolation filters, while madvr is limited to "oversample"
- with mpv you can select most parameters of all the algos while madvr has most of them hard coded
- mpv has the latest krig chroma upscaler from shiandow, madvr not

there are some madvr advantages you didn't mention:
- deringing
- film-mode
- error diffusion
- more reliable black bar detection features
- crispen edges, thin edges
- 3d blu-ray support
- built-in support for front projection screen masking, anamorphic lens and automated selection of lens memories via ip control
- continued development, while mpv seems to have slowed down recently
- a dev who actually invents new algos (e.g. gpu error diffusion, smooth motion, deringing, ngu)

File: mpc-hc_eva.jpg (603KB, 1920x1102px) Image search: [Google]

603KB, 1920x1102px

Has mpv fixed the floating gui shit yet or not?

Still staying on MPC-HC until it does.

>>59160374
>where's proof that igv's adaptive-sharpen is better quality than the madvr one? e.g. madvr has an anti-bloating and anti-ringing filter for adaptive-sharpen that mpv doesn't have
it's been known for a long time, and glsl version doesn't ring (can only amplify source ringing) and doesn't bloat

>where's proof that mpv jinc is better in mpv? thanks to better anti-ringing, madvr's jinc is probably better, imho
there is no ringing in mpv if configured properly, that also means ewa_* will be faster than jinc+ar

File: ringingTest.png (1KB, 100x80px) Image search: [Google]

1KB, 100x80px

>>59160614
interesting. can we put that to a test?

attached is an image which is often used on doom9 for ringing tests. the image obviously doesn't have any ringing by itself, so any ringing in the output must be caused by the sharpening or scaling algo.

how does this one look if you upscale it a lot (say 800%) with ewa_lanczos with "proper configuration"? and how does it look if you sharpen it a lot with ign's adaptive-sharpen? really no ringing at all? that would surprise me (positively).

File: [Coalgirls]_Sora_no_Woto_13_(1920x1080_Blu-Ray_FLAC)_[AF4F59ED].webm (568KB, 1920x1080px) Image search: [Google]

568KB, 1920x1080px

>>59158961
Noted, good to know.

>>59159507
No, sorry, I can't figure out any reasonable approach to that, within mpv. As far as I know, there's no way to use the cache like that.

You could fire off a stream-copying ffmpeg with segmented output and delete those segments periodically, then use those. More work to grab one and it's possible the action may fall on the segment gap, but at least it's not saving everything (well, keeping everything).
---
Also hey disregard the above, I went testing with ffmpeg after writing that. It is possible! I can split Twitch and RTSP into segments and (seemingly) piece them together with no missing frames / audio hijinks.
Question is now execution, how to bind this to mpv. At worst I'll have to throw some Python between, to interact with ffmpeg cleanly. FFmpeg would be started early and it'd spit out segments, maybe 20, 30 second long. Old ones would be removed, and on keybind the latest would be concatenated and stored for later inspection. Yes yes, this'll work.

No ETA! But I'll certainly see how nice I can make this.

File: ewa_lanczos.jpg (54KB, 900x720px) Image search: [Google]

54KB, 900x720px

>>59160657
small amount, with 480p+ content I don't see any ringing with scale=ewa_hanning + sigmoid-slope=10

File: adaptive-sharpen.jpg (54KB, 900x720px) Image search: [Google]

54KB, 900x720px

>>59160657

File: sssr.jpg (54KB, 900x720px) Image search: [Google]

54KB, 900x720px

>>59160657
ssimsuperres

File: sssr+adaptive-sharpen.jpg (55KB, 900x720px) Image search: [Google]

55KB, 900x720px

>>59160657
ssimsuperres + adaptive-sharpen (str 1.0)

File: mpv-shot0002.jpg (622KB, 1920x1080px) Image search: [Google]

622KB, 1920x1080px

>>59160657
>>59160782
>with 480p+ content I don't see any ringing with scale=ewa_hanning + sigmoid-slope=10
Here for example opengl-hq + scale=ewa_lanczos + adaptive-sharpen

File: mpv-shot0003.jpg (614KB, 1920x1080px) Image search: [Google]

614KB, 1920x1080px

>>59160922
compare to opengl-hq + scale=ewa_hanning + adaptive-sharpen + sigmoid-slope=10

>>59160782
thanks!

interesting. i can see ringing in the ewa_lanczos and adaptive_sharpen test images, but none in the ssimsuperres and sssr+adaptive-sharpen images.

so i suppose we can conclude that ewa_lanczos and adaptive-sharpen do ring (at least a little), but ssimsuperres does not.

the ssimsuperres+adaptive-sharpen image is confusing to me. if adaptive-sharpen by itself rings (see above), so should ssimsuperres+adaptive-sharpen. or did you apply them in the opposite order (first adaptive-sharpen, afterwards ssimsuperres)?

obviously that test image is an extreme case. having ringing in that image doesn't mean it will be obvious in typical real world situations. still, i think it's a good test.

>>59161055
>so i suppose we can conclude that ewa_lanczos and adaptive-sharpen
If there is no ringing in sssr+adaptive-sharpen, it means adaptive-sharpen doesn't ring

>>59160713
Thats great! Excited to see how it turns out.

>>59161070
then why is there ringing in the "adaptive-sharpen.jpg" image? see here:

>>59160786

>>59161055
>first adaptive-sharpen, afterwards ssimsuperres
no, ssimsuperres and then adaptive-sharpen

>>59161086
because sssr removed all ringing after ewa_lanczos, and adaptive-sharpen didn't add any new ringing

>>59161086
it's a same image as >>59160782
+ adaptive-sharpen

>>59161102
ah, so the "adpative-sharpen.jpg" image contains ewa_lanczos + adaptive_sharpen?

in that case it seems ssimsuperres and adaptive-sharpen both don't ring, that's pretty good!

but ewa_lanczos does ring a bit.

Would you guys recommend either...
(1) ewa_lancossharp
or
(2) ewa_hanning (or something smooth) + adaptive-sharpening

For 720p anime (1) works pretty well, mostly because I upscale it only to 1080p. I like my edges crisp and sharp here.
Real world content, however, looks horrible to me with (1), regarding ringing and other artifacts. I usually prefer my results more on the sharp side (and in case of anime it's very easy to "spot" sharpness) but can't stand ringing.
Any recommendations? I'm really not happy with the current situation and before I try every combination of scalers + shaders someone might have some insights already.

>>59161132
Correct

>>59161158
This >>59160931 is my daily config (fast and very good quality), but I upscale a lot.

>>59161171
The reason this looked pretty good was why I asked. It kinda looks like the sharpness of ewa_lanczos without the ringing. However, the high slope of 10 irritates me because it's so far from the default (which I assume was chosen as a good default)?

>>59161196
FWIW 10 is better in everything than default 6.5. Why 6.5 is default remains a mystery.

>>59161221
I think haasn replied to this on irc.
There are some issues with white dots.
I am not an expert so don't ask me to explain more.

>>59161298
White dots look better to me with 10.

>>59161298
Do you have a date or log? My BNC is usually idling in the mpv channel but I only found one message from 2017 matching "sigmoidal".

>>59161356
What about white or slope?

>>59161298
>>59161356
>>59161406
I found the question in a log from 2017-02-25:

<loyldr> from 4chan: >madvr jinc + ar >it's not better than ewa_lanczos + sigmoid-slope=10.0
<loyldr> Is changing sigmoid-slope to 10.0 as good as JincAR from madVR?

However, there's no answer.

>>59161351
I think he said white dots are artefacts produced by too high sigmoid-slope value.
You should add an issue on github, haasn is still active there. ;)

>>59161434
no answer?

>>59161438
I know, for values more then 10 there are artifacts with white dots/lines, but 10 is safe.

>>59161440
Nobody highlighted this guy or said anything about sigmoid ever again (one exception, but it wasn't related).

>>59161450
Ok, it seems you made a lot of tests so please share it on the mpv github repo.

>>59161496
One of them:
>>59160922
>>59160931

Why does no one seem to use anti-ringing with mpv?

>>59161567
>Why does no one seem to use anti-ringing with mpv?
You can't (shouldn't) with ewa salers and ewa scalers are best scalers.

>>59161587
Why is that? I'm ignorant if it causes some sort of artifacts.

>>59161646
He is just shitposting

ok, how do i launch mpv so it doesn't take focus away from my current application?
it takes a few seconds for yt-dl to open player (at least with my shitty connection) and during these seconds i usually start doing something else, then *mpv window appears* and takes focus for itself

if i recall correctly, mpchc had /nofocus switch; is there something similiar in mpv? (crtl+f in manual gave me no results)

File: Screen Shot 2017-02-28 at 13.11.12.png (2MB, 1800x1200px) Image search: [Google]

2MB, 1800x1200px

>>59161646
It replaces ringing with blocking.
See below the black lines, especially visible close to the blue circle.
Command line was:

--no-config --pause --keep-open --no-border --profile=opengl-hq --cscale-antiring=1 --scale-antiring=1 --scale=ewa_lanczossharp --window-scale=6

Try it again with spline36 instead of ewa_lanczossharp and notice that there is zero blocking.

>>59161680
Spotted the shitposter. Please don't tell me you were one of the guys talking about the sigmoid stuff above...

File: Screen Shot 2017-02-28 at 13.13.24.png (2MB, 1800x1200px) Image search: [Google]

2MB, 1800x1200px

>>59161646
>>59161787
For comparison: here's the same just with scale=spline36.

File: test-AR.png (3KB, 150x100px) Image search: [Google]

3KB, 150x100px

Forgot the original image.

>>59161787
I see what you mean. Is the difference between spline36 and ewa_lanczos worth forgoing the antiringing filter?

>>59161998
Probably depends on the source. I use spline on my laptop and ewa_lanczos on my desktop with dedicated GPU. Doesn't mean that's good though, especially with bad sources and high upscale ratio is dislike the result (spline36 isn't doing any better here though).
Therefore I'm inclined to try the combination of ewa_hanning + adaptive sharpen from above.

So.. I discovered I've been saving shit in .png

Will I lose any quality if I save the .png's as .jpg compared to taking screenshots as .jpg?

>>59162194
Use a decent image viewer that lets you select the quality level.

>>59162194
No. Ultimately, it all depends on the quality options you choose.

i have to say this thread is more informative than the last 20 mpv threads combined :)

Should I change sigmoid-center to 1.0 or it's not needed?

>>59162533
not needed, looks worse in some cases to me.

>>59162479
Unfortunately, the discussions here got completely unscientific. Nearly nobody with decent knowledge is left, nobody is willing to actually check what think or heard is correct (like >>59161680) and some might believe such people.
In this state, all "knowledge" here, all "shared wisdom", should be considered highly questionable until shown otherwise. But nobody makes an effort to verify anything, so today these threads are nearly useless.

>>59162533
>>59162634
Some guy above claims 10 is best.

>>59162533

n cycle-values sigmoid-center 0.75 1.0

Test it yourself. I really don't see any difference.

>>59162479
Yes, the last time was when haasn developed the shader hook ecosystem.

HAASN, you miss us!

>>59162783
>Some guy above claims 10 is best.
That's sigmoid-slope, not center.

>>59162783
>Some guy above claims 10 is best.
>sigmoid-center
>sigmoid-slope
Learn to read

>>59161438
>haasn
>active
he's too busy playing diablo
https://haasn.xyz/posts/2017-02-18-the-diablo-iii-paragon-system-visualized.html

>>59162533
>>59162783
Oh sorry, "center". Misread the post. Of course nobody claims 10 to be better for center.

>>59162783
>makes a statement without any proves >>59161587
>got called "shitposting"
>nobody is willing to actually check what think or heard is correct
how retarded are you, omg?

>>59162838
>how retarded are you, omg?
This is common knowledge for ages, the man pages mentions the possible introduction of new artifacts. It is so known that people often suggest deactivating AR for ewa.
And all the images showing exactly this are here.

>>59162903
>It is so known that people often suggest deactivating AR for ewa.
I mean in code, like making it impossible to use AR for ewa.

>>59162808
>be haasn
>play diablo
>make weird graphs
Reminds me when I was 7 and played pokemon with EV/IV to increase specific attributes of my pokemons. For each pokemon, I had a spreadsheet with autistic data. Other kids called me the "gameboyer".
I waste so many hours like that...

>>59162783
most recent mpv threads were like that, i agree, but this one seems above average, with some hard evidence that ssimsuperres and adaptive-sharpen don't ring, while ewa-lanczos does. and with some (at least half way) useful mpv vs ngu comparison screenshots etc...

>>59163037
Definitely. Unfortunately, this is going to end now.

>>59163056
No more reply limit for 4chan thread?

>>59162913
>I mean in code, like making it impossible to use AR for ewa.
I would remove it all-together from mpv

>>59163163
>I would remove it all-together from mpv
It's still useful for default opengl-hq with spline.