/mpv/ - the Sup Forumsreatest media player

Wiki:
github.com/mpv-player/mpv/wiki

Manual:
Git: mpv.io/manual/master/
Stable: mpv.io/manual/stable/

User Scripts (including opengl shaders):
github.com/mpv-player/mpv/wiki/User-Scripts

High quality video output profile (goes into mpv.conf):
Git: profile=gpu-hq
Stable: profile=opengl-hq

Configuration Files:
mpv.io/manual/master/#configuration-files

Input.conf:
github.com/mpv-player/mpv/blob/master/etc/input.conf

Post your system specs and config if you're asking performance related questions.

Other urls found in this thread:

mpv.srsfckn.biz/
bitbucket.org/rorgoroth/mingw-w64-cmake/downloads/
sourceforge.net/projects/mpv-player-windows/files/
sourceforge.net/projects/mpv-player-windows/files/Test/
jenkins.maeyanie.com/job/mpv/
my.mixtape.moe/vhsmgz.webm
my.mixtape.moe/ksclqu.zip
github.com/mpv-player/mpv/commit/719a435d3636c7b848b49f06e6a2ea55e7ee656a
twitter.com/SFWRedditImages

Windows Builds:

Stable Portable: mpv.srsfckn.biz/
Stable Installer: bitbucket.org/rorgoroth/mingw-w64-cmake/downloads/
Dev Snapshots: sourceforge.net/projects/mpv-player-windows/files/

WIP d3d11/VulkanR2: sourceforge.net/projects/mpv-player-windows/files/Test/
WIP d3d11/VulkanR2 Installer: bitbucket.org/rorgoroth/mingw-w64-cmake/downloads/

Vapoursynth: jenkins.maeyanie.com/job/mpv/

Config for D3D11/Vulkan:
gpu-api=d3d11 |..or..| gpu-api=vulkan
hwdec=auto-copy
profile=gpu-hq
vd-lavc-dr=yes

Why aren't those in OP anymore?

still wondering why mpv locks up with high cpu usage when opening images.

Does direct rendering even work on d3d11?

Nope. I think it does on Vulkan though.

I noticed that it also happens to gif but only after they play. It locks up once an attempt to loop happens.
side question, can I merge multiple extension profiles to one? The doc doesn't say if this is possible. I haven't tried [extension.png extension.jpg] nor [extension.png, extension.jpg]
config that I'm using here:
autofit-larger=60%x60%
autofit-smaller=40%x40%
correct-downscaling=yes
cscale=ewa_lanczos
deband=yes
dither-depth=auto
force-window=immediate
geometry=50%:50%
gpu-api=vulkan
hwdec=auto-copy
keep-open
load-stats-overlay=no
no-border
no-osd-bar
no-taskbar-progress
ontop
scale=ewa_lanczossharp
screenshot-directory="~/pictures/"
screenshot-template="%F - %02n"
sigmoid-upscaling=yes
snap-window
tscale=oversample
vd-lavc-dr=yes
volume-max=100
volume=60
[extension.gif]
profile=loop
[extension.jpg]
profile=image
[extension.jpeg]
profile=image
[extension.mkv]
profile=video
[extension.mp3]
profile=music
[extension.png]
profile=image
[extension.webm]
profile=loop
[image]
profile=default
image-display-duration=inf
no-resume-playback
[loop]
profile=default
force-seekable=yes
loop=inf
[music]
profile=default
audio-display=attachment
[protocol.ytdl]
profile=default
[video]
blend-subtitles
deband-iterations=2
deband-range=12
fbo-format=rgba16f
fullscreen
glsl-shader="~~/shaders/FSRCNN_x2_r2_32-0-2.glsl"
glsl-shader="~~/shaders/LumaSharpenHook.glsl"
interpolation
save-position-on-quit
temporal-dither
video-sync=display-resample

how to fix vertical tearing?

using smplayer as overlay

Why...

Just tested, yeah it does work. ~1500 FPS under Vulkan without direct rendering against ~1700 with direct rendering, lol.

Haasn changed something after the first few weeks of rev2 that makes it instantaneously crash half the time now when I try to open mpv with --gpu-api=vulkan under Windows. Always opens just fine with rev1, used to always open just fine in the first few rev2 builds.

Did you try update driver first?

Bisect it? I think they (cough wm4 cough) broke it sometime between Oct 15 and Oct 21.

Wasn't this only really necessary for Nvidia?

here is what happens with nothing added to the player: my.mixtape.moe/vhsmgz.webm
A 4k 60fps video has it at 5% cpu usage.

Redpill me on wm4 and haasn

Wm4 is a code quality autist.
Haasn is a video quality autist.
Wm4 is about making it simple and maintainable.
Haasn is about choice and modularity.
Wm4 is about reliability and stability.
Haasn is about fun and adventure.

Basically haasn will give you cool features and cool shit for you to play around with, while wm4 makes sure the player still works and remains sane. While also insulting haasn's code.

I'm a Windows user with nothing but an Intel i5 2500k.

I barely have to change any of the config other than screenshot format, sub font, etc.

This is the best player I have ever used. Faster and cleaner than mpc-hc. Glad I listened to you guys.

Anyone interested in FSRCNN vs RAVU comparison?

Go for it

>mfw FSRCNN is actually better than NGU Sharp now
igv didn't waste his time. Now i understand why madshi rushes madVR 1.0.

RAVU r4.

FSRCNN HQ r1.

WHAT THE ACTUAL FUCK

If it works fine with the earlier rev2, can you bisect it?

Those images seem doctored to me. RAVU should have a pixel shift to the top left, FSRCNN should have none. In your image it seems to be the other way around. What gives?

These shaders are developing too fucking fast, I'm getting tired of changing them.

I'd say haasn cares about code quality just as much, he just usually has a radically different idea about what the correct design should look like, which wm4 usually ends up overriding; resulting in a weird mismash between wm4's and haasn's styles that both of them end up hating.

Maybe it's time to stop jumping on every bandwagon you see

No!

God. Look at this sharpness.

There may be even better version of FSRCNN soon (hybrid of FSRCNN + LapSRN)

>Those images seem doctored to me. RAVU should have a pixel shift to the top left, FSRCNN should have none. In your image it seems to be the other way around. What gives?
Its 2X upscale of bjin's test image. Same config. Nothing is altered. AFAIK mpv fixes shift only when scale is involved.
I actually was surprised myself.
Holy shit!

This looks great.
How do you add these shaders though?
Just add post shaders="path" ?

glsl-shaders="~~/FSRCNN_x2_r2_16-0-2.glsl"
FSRCNN-MQ is enough IMO

Me again. I just compared to madVR's NGU Sharp Very High. Not posting a result for now because the mpv one was affected by ICC profile. Here is my conclusion:
>FSRCNN has better, thicker more natural looking lines than NGU Sharp
>FSRCNN has less errors in its result
>FSRCNN slightly has more ringing (wont notice until like 200% zoom)
>sharpness level is mostly the same but FSRCNN has sharper looking textures
>FSRCNN can also remove minor compression artifacts
RIP NGU.

I did some comparisons of my own. Which looks best, guys?
my.mixtape.moe/ksclqu.zip

Thanks.

doesn't look like it is activated

convert_script.lua is broken. Any alternatives?

aah never mind, I fixed it

G-guys i think FSRCNN just killed RAVU...

Just wait for the next MadVR. It's going to be huge.

RAVU r4.

FSRCNN HQ r1.

igv, i summon you. Pls, i have a question.

FSRCNN has MUCH less ringing than RAVU and has much less bloated lines without removing details (for anime). It is also sharper but just enough to not look oversharpened. I'm very impressed. When the fuck it became so good?

go away poozoor

Former MPC user here.
MPV is great!
(Still tailoring the customizable layout to my liking.)

Performance wise it's smoother and snappier.
The interface/controls was a little less intuitive but once I found out I can remap it I went all in.

Thanks for the help guys!

I am not poozoor. poozoor only talks. I actually make comparisons.

RAVU r4.

FSRCNN HQ r1.

You don't use r1 for aliased shit.

>all this blur
I see why haasn like it now.

I use RAVU
it's pretty nice
I use spline36
Plus I actually trust bjin's code unlike igv's

Thanks for the tip! I have a question. Look at the middle part of the grills's skirt in second pic (fsrcnn). It looks quiet a bit brighter than the one in RAVU.

wtf i dont trust igv's code now

What happened to mpv being a simple frontend for ffmpeg? Now it's fucking full of bloated meme filters.

Just use ffplay.

> Requires --fbo-format=rgba16f (rgba16hf for vulkan)

igv, is FSRCNN works normal in this scenario?
>FSRCNN -> Mitchell.

why wouldn't it?

Last few times i tried it ended up much less sharp than NGU Sharp -> SSIM1D. Same with Mitchell+SSIM.

Btw hats off to igv for FSRCNN-HQ, I think it's done in a pretty clever way. His use of the four image components for the 2x2 deconvolution step is nice.

Would be interesting to see if merging the final step into a single pass is faster though

In madvr it's catmul-rom + ssim.
Also depends on ssim strength.

>Would be interesting to see if merging the final step into a single pass is faster though
No, it's not faster.

mpv was NEVER a simple frontend to ffmpeg. May I remind you that mpv started out as MPlayer/mplayer2?

What would be faster is adding support for multiple render targets to mpv.

I'm getting segfaults with [Coalgirls]_Serial_Experiments_Lain_XX_(1008x720_Blu-Ray_FLAC)_[F0EF8AF8].mkv and [Coalgirls]_Haibane_Renmei_XX_(1280x720_Blu-Ray_FLAC)_[8AA17B0E].mkv I think it might be something to do with ordered chapters because I have other coalgirls releases that don't have OC and they work. They used to play fine last time I tried them a while back. I'm on git master. Anyone else experiencing this? Would it help to post a log?

Is wm4 the only sane mpv user?

Have you tried using compute shaders + larger textures to output multiple texels at the same time?

Basically say you have a 1920x1080 image; instead of rendering four feature maps into 1920x1080 framebuffers, you could render a single feature map into a 3840x2160 framebuffer, at index (2 * gl_GlobalInvocationID) + offset where offset ranges from (0,0) to (1,1).

Multiple render targets in mpv would be great; perhaps it could be implemented as some sort of equivalent of texture arrays. I don't think it's that close on the horizon though, and would have lots of difficulties. Maybe something for libplacebo?

>In madvr it's catmul-rom + ssim.
>Also depends on ssim strength.
Well, i understand. Still even downscaled with or without SSIM it looked less sharp. Same doesn't happen with RAVU. I mean, no way RAVU should be sharper in ANY scenario. Ill try again later and post pictures.

>not watching everything at native resolution so the only scaling operation is on chroma

I thought about it, but haven't tried.

I did a bisect and the commit that broke it was 719a435d3636c7b848b49f06e6a2ea55e7ee656a

github.com/mpv-player/mpv/commit/719a435d3636c7b848b49f06e6a2ea55e7ee656a
>fuck up everything and vomit ponies and rainbows
yep sounds about right

Thought: Instead of rendering to four adjacent texels it might actually be faster to make the textures “planar”.

Basically, instead of a 3840x2160 texture, use a 1920x4320 texture and render to e.g. (50,50), (50,1130), (50,2210) and (50,3290) in one pass.

Yes. He should remove all this placebo garbage from mpv.

looks like fsrcnn has improved a lot. but after comparing, i still find ngu sharp better. see attachments.

deband-grain=80
deband-iterations=4
deband-range=6
deband-threshold=156
Best debanding options for anime.

does mpv have streamlink built in

what's with the color discrepancy

I like FSRCNN more with cartoons. NGU Sharp thins lines too much. FSRCNN also works better with low res videos.

One pic is with ICC profile. The other one made by another user without it.

Tell us what is the shape of NGU already.

> Here is my conclusion:
> FSRCNN has better, thicker more natural looking lines than NGU Sharp

thicker yes. but that isn't necessarily a good thing. hey, ravu has thicker lines than fssrcnn, so ravu must be better? and mitchell has even thicker lines than ravu, so mitchell is best?

i don't see, at least not in these 2 images, where fsrcnn would have better or more natural lines. they're just thicker. otherwise they look pretty similar to me.

> FSRCNN has less errors in its result

where do you see errors in these 2 ngu images? to my eyes fsrcnn and ngu look pretty similar, except for line thickness and ringing.

> FSRCNN slightly has more ringing (wont notice until like 200% zoom)

make that a *lot* more ringing.

how does speed compare, btw? haven't actually benchmarked latest fsrcnn hq shaders myself yet.

my overall impression is that igv has done a pretty good job on optimizing fsrcnn and making it look better (much less ringing than it used to have), but imho it hasn't reached ngu sharp quite yet. maybe soon?

no
you can play youtube/twitch streams with youtube-dl. although i have found a few sites streamlink supports but yt-dl doesn't so i use both.

line thickness probably has to do with gamma vs linear light. at least ngu line thickness seems to have changed when the changelog said:

* NGU Sharp now targets middle ground between gamma and linear light

what do you mean with "shape"? for technical details you'll have to ask madshi, maybe if you catch him in a good mood?

I've got a pretty specific question, i have Youtube-dl and draging my link on mpv work.

But i want to stream it through a Torsocks in 640x360 webm format on windows. Is it possible?
I know that Hooktube is able to deliver 640x360 webm but i'm am not able to configure mine. If anybody could help me that would be great.

gpu-api=vulkan
profile=gpu-hq
scale=haasnsoft
cscale=haasnsoft
tscale=linear
interpolation=yes
blend-subtitles=yes
deband-grain=0
deband-iterations=2
deband-range=12
deband-threshold=48
temporal-dither=yes
video-sync=display-resample

>thicker yes. but that isn't necessarily a good thing. hey, ravu has thicker lines than fssrcnn, so ravu must be better? and mitchell has even thicker lines than ravu, so mitchell is best?
NGU Sharp thins lines too hard compared to RAVU, FSRCNN and ewa_lanczos. Its especially bad in anime and other cartoons. Detail gets removed.
>i don't see, at least not in these 2 images, where fsrcnn would have better or more natural lines. they're just thicker. otherwise they look pretty similar to me.
That was my conclusion after testing a bunch of different anime images. Still valid though even with those posted images.
>where do you see errors in these 2 ngu images?
Again. Maybe not much in those images but thats what i noticed with lots of other anime.
>make that a *lot* more ringing.
I don't get it.
>how does speed compare, btw?
About the same as NGU Sharp Very High. Probably a bit slower.
>my overall impression is that igv has done a pretty good job on optimizing fsrcnn and making it look better (much less ringing than it used to have), but imho it hasn't reached ngu sharp quite yet. maybe soon?
FSRCNN progress at incredible speed. igv started working on it only 3 months ago... I think in 3 more months it will completely BTFO NGU. Even at current state they look very similar while FSRCNN still has room for improvement.

madshi, love you anyway

C:\Users\fag> mpv E:\Downloads\Planet.Earth.II.S01.2160p.BluRay.HEVC.DTS-HD.MA.5.1-SUPERSIZE\DISC1\BDMV
Playing: E:\Downloads\Planet.Earth.II.S01.2160p.BluRay.HEVC.DTS-HD.MA.5.1-SUPERSIZE\DISC1\BDMV
[file] Cannot open file 'E:\Downloads\Planet.Earth.II.S01.2160p.BluRay.HEVC.DTS-HD.MA.5.1-SUPERSIZE\DISC1\BDMV': Permission denied
Failed to open E:\Downloads\Planet.Earth.II.S01.2160p.BluRay.HEVC.DTS-HD.MA.5.1-SUPERSIZE\DISC1\BDMV.

I'm probably making a retarded mistake here because I'm pretty hungover. pls help.

Put the directory in quotes

Same error

Also you're trying to access your E drive from your C drive

Shouldn't be a problem. I always play my video from my E drive like this.

Is there a way to tell mpv to prefer .ass subtitles over for example .srt subs?

mpv bd:// --bluray-device="E:\Downloads\Planet.Earth.II.S01.2160p.BluRay.HEVC.DTS-HD.MA.5.1-SUPERSIZE\DISC1\BDMV"
or, you know, just open the actual video files.

Mpv has stopped playing video.Using the latest git as of 5 minutes ago.
When I run it from cli mpv file.mpq it just says "playing file.mpq" but dosent actually run it. It does this for various file types, not just mpq. What could cause this?

Yeah, sorry I just tried that myself and it worked fine. Perhaps it's because there are backslashes in the filename? Windows gives me an error with that.

>What could cause this?
probably

C:\Users\fag> mpv bd:// --bluray-device="E:\Downloads\Planet.Earth.II.S01.2160p.BluRay.HEVC.DTS-HD.MA.5.1-SUPERSIZE\DISC1\BDMV"
Playing: bd://
src/libbluray/disc/disc.c:424: error opening file BDMV\index.bdmv
src/libbluray/disc/disc.c:424: error opening file BDMV\BACKUP\index.bdmv
[bd] Couldn't open Blu-ray device: E:\Downloads\Planet.Earth.II.S01.2160p.BluRay.HEVC.DTS-HD.MA.5.1-SUPERSIZE\DISC1\BDMV
No protocol handler found to open URL bd://
The protocol is either unsupported, or was disabled at compile-time.

I guess I could just play the m2ts files, but I'm still curious what's fucking it up. I've played BDMV folders normally by just dragging the folder into the mpv window before.

There aren't any backslashes though.