I wanted to ask if mpv always created a window and closed it immediately when called without arguments if `force-window` is configured to `yes` or `immediate`. But as it turns out, I built mpv just before the issue was fixed, presumably by this commit:
https://github.com/mpv-player/mpv/commit/544240c829cd9c9a180b7a689868904024c96fe4
although the commit message doesn't cover that case.

[mpc]
force-window=immediate
idle=yes
keepaspect-window=no
auto-window-resize=no

profile=high-quality

Rate my config.

>>102473848
kill yourself

>>102473839
>[mpc]

Make them stop fighting /g/
https://github.com/mpv-player/mpv/pull/14876

>>102474316
didn't read til the end for my IQ's sake. but someone should tell kasper that audio frames are called samples.

File: snu.webm (938 KB, 920x1632)

938 KB WEBM

Any tips on how to upscale and interpolate low quality videos?
Webm related for example.

File: ranma_ep53.webm (4 MB, 480x360)

4 MB WEBM

here are my ffmpeg commands to create the highest quality webms under 4mb (bitrate is adjusted per video). anyone see any issues?

ffmpeg -i "53.mkv" -ss 10:14.405 -to 11:02.620 -c:v libvpx-vp9 -an -sn -vf subtitles="53.mkv:stream_index=4",scale=-1:360 -b:v 0.7M -pass 1 -f null NUL
ffmpeg -i "53.mkv" -ss 10:14.405 -to 11:02.620 -c:v libvpx-vp9 -an -sn -vf subtitles="53.mkv:stream_index=4",scale=-1:360 -b:v 0.7M -pass 2 output.webm

>>102474754
That webm is broken
>>102473413
I'm a windows user and tech illiterate. Yt-dlp plus ffmpeg is amazing and makes me feel like one of you.

>>102474754
>Any tips on how to upscale and interpolate low quality videos?
yes. contact the whore and ask for a higher quality source.

>>102474768
great stuff
should i pre-modify videos to match the -ss and -to?

>>102475145
no, that's adjusted per clip.

>>102475218
should i pre-edit videos to place the to-be-rendered subtitle track at index 4?

>>102473413
>download mpv
>it's shit
/g/ is retarded

asking again because last thread was a shitshow

I've been using kdenlive to cut some clips out of a longer movie and make them into short webms. The full video is apparently 15 fps, but playing it in VLC makes it look much smoother than the display inside kdenlive (and davinci resolve for that matter). The final webm also plays very choppily in VLC compared to the original.

I'd like to make the webms look just as smooth as the original video.
Is there some setting I need to enable in kdenlive on the project? Some sort of interpolation I need to explicitly flag?

>>102475388
do you think we have crystal balls or something?
there are too many potential variables that can be involved.
and there is no reason to use anything other than ffmpeg to cut videos.
so use ffmpeg to see if the problem persists.
if it does, upload a sample or post mediainfo output so we can at least know a little bit more about the source video.

Is there truly no frame-perfect way of trimming videos on ffmpeg?

Which settings do you use with AV1 and Opus? I just know about preset 0 and crf 0.

>>102475354
Explain, how it is shit?

File: 1723838506922400.jpg (58 KB, 976x850)

58 KB JPG

>>102473413
what's the point of librempeg

>>102475284
no, use the subtitle track you want to, or remove the subtitle filter for no subs.

>>102475609
pure development.
no drama or bureaucracy or any other *nontechnical nonsense*.

>>102475672
Does it not have precompiled binaries?

>>102475465
>and there is no reason to use anything other than ffmpeg to cut videos
can I do that down to the frame?

>>102475568
>>102475807
I've used -to and -ss to trim clips in cases where frame accuracy is important. Just use timestamps with milliseconds. Is that not precise enough for you?

video-sync=display-resample
video-sync-max-video-change=0
video-sync-max-audio-change=0
interpolation=yes
tscale=oversample

rate my config

>>102476542
This one >>102476505 was better.

>ffmpeg -i video.mp4 -c copy video2.mp4
>video2.mp4 is almost twice the size of video.mp4
why

>>102475688
>***pure development.***
you're not a part of the target audience.

>>102476668
another crystal balling question

>>102475871
>Just use timestamps with milliseconds. Is that not precise enough for you?
That's what I do, but there's a bunch of trial and error there to find the right timings for the right frames. I wish it were easier.

I noticed the option "-g 9999" makes the seekbar slower to seek. Does it decrease file size and does it affect quality?

>>102476684
>trial and error
that's just the nature of the beast. video decoding is a state machine meaning calculating the current frame is performed by starting at frame 0 and counting. Sure it can be done but it's a wasteful process to keep repeating. Timestamps are more efficient just using some trial and error.

>>102476803
any value larger than frame_rate*10 is retarded.

>>102477310
Well I see 9999 thrown around in "make da most optimized webmz ever" guides

>>102477384
it's thrown around by the likes of libvpx/libav1 devs to win 0.1% of PSNR/SSIM/.... vs other encoders.
if you're not playing them stupid games, stop looking for stupid prices.

>>102477435
forgot to mention that this is used with samples that specifically would compress well with exactly one I frame.

>not using I frames
retarded

What does this entails for the future of mpv and shaders:
https://devblogs.microsoft.com/directx/directx-adopting-spir-v/

>>102477490
nothing, yet. We need d3d12 backend

>>102477536
What exactly is the difference between SPIR-V and GLSL and why mpv shaders are still on GLSL?

>>102477557
GLSL (OpenGL Shading Language) is a human-readable and writable shading language, similar in syntax to the C language. SPIR-V is an intermediate representation (IR) format. GLSL is compiled into SPIR-V, which is then used by the GPU driver to generate low-level machine code that the hardware can understand.

In the case of OpenGL, the driver typically compiles GLSL directly.

Now, let's consider DirectX, which is currently a mess in mpv. Shaders written in GLSL are first compiled to SPIR-V, which is then translated to DXBC (DirectX Byte Code) and sent to DirectX. DXBC is used by D3D11. In D3D12, the format used is DXIL (DirectX Intermediate Language), which is SPIR-V counterpart.

Vulkan natively supports only SPIR-V and does not have a direct GLSL replacement. GLSL shaders in Vulkan are first compiled to SPIR-V before use.

GLSL (OpenGL) is equivalent to HLSL (DirectX) in terms of shader languages.
SPIR-V (Vulkan) is equivalent to DXIL (D3D12) and DXBC (D3D11) as an intermediate representation format.

Microsoft, simply retire its own IR for SPIR-V which will simplify and unify a lot of things. Also HLSL to SPIR-V compilers exist and are widely used.

HLSL is the industry standard for DirectX. GLSL is traditionally tied to OpenGL, though both can be compiled to SPIR-V for Vulkan use. (and back to d3d...)

>>102477814
Microsoft by officially adapting SPIR-V will effectively make HLSL the shading language for both Vulkan, DirectX and anything else...

>>102477814
dropping windows support would immediately fix the "mess"

>>102477927
Not really, because we still are generating GLSL shaders with sprintf... which is annoying. Dropping Windows would only help to remove one transcribe step.

>>102478171
>would only help to remove one transcribe step.
sounds good to me.

>>102478529
open PR

File: cyMze.png (2 KB, 407x58)

2 KB PNG

Be honest.

Do you need more?

>>102477536
>We need d3d12 backend
yeah sure... right after the Metal backend...

BWAHAHAHAHAHAH

>>102480119
sounds good to me

>>102480403
never happening winbabby/macbabby

>>102480422
You know what, the Metal backend would be more useful than the Vulkan one. Windows uses DirectX, macOS uses Metal, and Android uses OpenGL. Why does mpv need Vulkan? Linux can use OpenGL just fine, and it's known to work better than Vulkan in many cases.

>>102480484
this is some next level babyduck cope

>>102480525
Why AAA game developers does not invest in Vulkan support?

>>102480781
They do though? There have been plenty of AAA games that offer a Vulkan backend on shitblows. And on Linux, dxvk is more performant than dx12 despite being a compatibility layer with overhead.

File: opus.png (124 KB, 1280x948)

124 KB PNG

What's the best tool for converting audio (.flac music, .wav podcasts) to Opus and what settings? ffmpeg with the default settings seems sane from what I see in the documentation, and I'm thinking:
>96k for podcasts
>128k for music
A question I'm more interested in:
What container? Which one has the best device compatibility, and is there any other noteworthy differences (e.g. one has noticeably shittier options for metadata)?
>.opus
>.ogg
>.mka
My understanding is .opus is just a different name for an .ogg container that specifically contains opus audio, and that that .ogg usually implies vorbis. Assuming that's true, is there a good reason to use .ogg for opus instead of the more explicit .opus?

>>102481771
Update:
ffmpeg doesn't work because it can't embed album art for .opus/.ogg (and seemingly .mka as well?).
https://trac.ffmpeg.org/ticket/4448
I used opusenc instead and all the album art & metadata seems to have been copied without issue. But if anyone has further opinions, especially on .mka vs .opus, I'm still curious.

>>102474768

 -ss 10:14.405 -to 11:02.620 -copyts -i "53.mkv" -ss 10:14.405 

faster

 -row-mt 1 

almost always faster and no real downside

madvrtards are currently having a galactic-scale meltdown on the fact that it hasn't been updated for almost more than a year now, and there are people recommending that unhappy users who dont like the timebomb model should buy an Envy

>>102482583
I don't even feel bad for them, madVR has been dead for years. Stop living in the past.

>>102482682
unfortunately most people will just move to MPC-VR which is a considerably more inferior project by every metric

>>102482461
what is -copyts for and why is that better?

>>102482748
i use mpc vr because it supports d3d11va native

>>102483164
I use mpv because it supports d3d11va native

>>102483057
putting -ss and -to in front of -i is faster because it will directly seek to the relevant part
-copyts copies input timestamps to output instead of generating new timestamps starting at zero, it and second -ss after -i are needed for burning in subs because their timings are relative to the original video timestamps

/x/ + mpv collab
https://github.com/mpv-player/mpv/issues/14892

>>102483164
and in exchange you get.. worse gamma, terrible gamut mapping, terrible tonemapping, bad scalers... yeah it's embarrassingly bad.

also, last time i used madvr native api worked fine.

>>102483377
1. user fiddled too much with gpu configuration using the nvidia settings gui
2. windows support is deprecated
closing
closing

is it out yet

>>102484204
no

why do options/audio-file or options/sub-file property show error when I try to show the text?

>>102485453
confirmed.
printed options alone to see all keys.
audio-files and sub-files work.

>>102484204
is it in yet

File: cat-considerate.jpg (84 KB, 640x640)

84 KB JPG

>>102473413
Hey guys. I need your help.
I have a 1080x1080 video. 1:1 aspect ratio.
I want to transcode it to 1082x1080 dimensions stretching the video.

So I don't want to respect the aspect ratio. Just stretch that shit to fill the new dimensions. How can I do this with ffmpeg?

Is there simple history log script for mpv?

>>102486996
with ffmpeg
-s 1082x1080 -aspect 1082:1080

>>102486996
>>102487244
actually -aspect alone might be enough if you don't want to reencode.
so

ffmpeg -i <input_file> -c copy -aspect 1082:1080 <output_file>

>>102487297
Sadly this doesn't seem to work. Output file remains as 1080x1080.

Found the rencode solution with:
-vf "scale=1080:1080,setsar=1/1"

>>102487469
Sorry meant:
-vf "scale=1082:1080,setsar=1/1"

>>102487482
you don't need to re-encode.
using -aspect 1082:1080 should set the aspect ratio on the container level. resolution would stay the same. but the output would be stretched.
you can additionally the aspect ratio on the stream level with e.g. `-bsf:v h264_metadata=sample_aspect_ratio=1082/1080`. replace h264 with the applicable codec.
you can test both options using mpv with `--video-aspect-method=container` and `--video-aspect-method=bitstream`.
test with something like 2000:200 to make the difference clear.

>>102487610
okay. correcting myself again. only h264 and hevc can set sample_aspect_ratio at the bitstream level.
setting at the container level should work anyway.

>>102487610
>>102487649
WOAH that is super awesome thanks!

>>102487610
>>102487721
keep in mind some retarded players might ignore that metadata

pmozil haasn: Hi, could you look at the vulkan dirac hwaccel patch? Here's the link to the mailing list thread:
haasn pmozil: hey, I can take a look a bit later
haasn a bit busy with the STF deadline atm

>>102488759
I miss haasn

new autism just dropped
https://github.com/mpv-player/mpv/pull/14894

>>102489199
lgtm

>>102476669
>target audience is never using it and just cooooooding
fuck troons that put that shit in the OP

>>102489358
cope. ffmpeg is dead

>>102489376
dilate

>>102489358
no one will feed you wintards from now on.
even with mpv, the couple of nu-mpv devs who are pushing support for that crap will one day disappear (as is live in the open source world), the support will rot, and one day it will be dropped altogether.
feel free to troon-shout at the mirror then, while cry-posting how vlc and mpc-hc are superior.

>>102489358
>>102489420
>you're a tranny because im 80 iq and don't know how to compile binaries or something

>>102489654
BASED braphog enjoying larper

>>102489654
>>102489655
kys miserable tinkertrannies

>>102489739
how do you post on /g/ but dont know how to compile your own binaries.

>>102489739
>knowing how to compile C libraries and programs makes you a tinkertranny
nu-/g/ is pretty grim

>>102482682
fix dscale AR and oversample and i'll switch

How is this better than using MPC-HC

>>102489654
>even with mpv, the couple of nu-mpv devs who are pushing support for that crap will one day disappear
What a retarded argument? You could say that about any feature in any software. There is nothing specifically different about one thing and the other. You are trying too hard to push your flawed point of view.

>>102490406
>fix dscale AR
antiringing always results in a reduction of quality. madVR's AR is even worse somehow
>oversample
works fine, the flicker meme has been deboonked recently (also smoothmotion is blurrier and not a real oversample kernel)

>>102490720
>antiringing always results in a reduction of quality. madVR's AR is even worse somehow
current dscale AR is disabled, because of quality issues

>>102490720
>works fine
it really doesn't. the "deboonkers" just lack visual acuity

>>102490803
no shit fagtard

the point is, you have to let filters overshoot for them to reach a given pixel value intensity. antiringing by design is bad, as you're just further blurring the filter kernel and adding aliasing (which is bad because the whole point of using a low-pass filter for upscaling is to reduce aliasing). even if antiringing was magically fixed, no one should use it, as it just reveals their ignorance.

>>102490815
the debookers actually went frame by frame and proved it works fine. you're hallucinating and just regurgitating memes, and have never done your own a-b test. hmm i wonder who i should trust in this exchange?

File: Logo_for_AMD's_FreeSync_t(...).png (36 KB, 343x291)

36 KB PNG

>he still uses oversample in 2024

>>102489739
projecting pedo award

>turn off PC last night
>turn it on this morning
>mpv now takes ~6 seconds to load any video I open regardless of size or location
o-okay

>>102490987
don't use vulkan

>>102490987
stop using windows. this happens on every windows application because it has to get added to the memory cache on a "cold boot". absolutely retarded design.

>>102490999
>>102491046
yes hello im retarded and apparently unplugging my speakers makes it pause to search for audio outputs for six seconds every time you open, fixt ;)

>>102490716
introduce yourself. kasper or nanahi or ...?
>You could say that about any feature in any software.
yes.
>There is nothing specifically different about one thing and the other.
mpv will always have more devs/maintainers who would prefer to remove windows support than properly maintaining it (including daily driving).
a flurry of activity over a couple of years from a couple of windows-daily-driving devs will never change that fact.

>>102490877
>you're hallucinating
fuck you you blind fuck. you can keep your broken ass media player

>>102491143
kek nice meltdown

>>102491122
I just looked up your contributions and you only have 3 commits dude. No need to huff your own farts this much.

>>102491122
>mpv will always have more devs/maintainers who would prefer to remove windows support than properly maintaining it (including daily driving).
What's this weird LARP? Windows support existing doesn't affect any of the core parts of the player.

>>102474768
Thank you

>>102491122
>mpv will always have more devs/maintainers who would prefer to remove windows support
You are the only one saying this. Having multi-platform support is beneficial for the project’s health and longevity. You can focus on supporting your preferred operating system, but this will only reduce the number of mpv users and developers. Both kasper and na-na-hi are fixing many other issues, not just Windows support. Many Windows users are reporting issues and testing mpv. You seem so stubborn that it's blinding you. Feel free to maintain your own player (or even an mpv fork), which will only be used by you and your friends.

How do you make mpv audio sound better

>>102491393
Buy better speakers and/or headphones.

>>102491393
statically build mpv with librempeg for free audio quality boost

>>102491405
They already are.

>>102491422
Then it's time to invest in acoustic treatment of your room.

>>102491393
* use alsa dmix with high quality resampler
* replace mpv `volume` adjustment with `volume-gain` adjustment (volume fixed at 100).
anything beyond that is placebo

>>102491467
>* replace mpv `volume` adjustment with `volume-gain` adjustment (volume fixed at 100).
now that i think about it, why isn't this the default anyways? most programs that do audio resampling usually only allow you to tweak audio on a db scale, and not a weird linear scale

>>102491418
>statically build mpv with librempeg for free audio quality boost
Unironically this may be true if you use some filter

>>102491541
most of the stuff that users commonly use like dynaudnorm haven't received any tweaks AFAIK, but yes ffmpeg audio filters are currently in an unmaintained state so maybe something might change in the future.