how to into hdr?

when i turn on hdr on windows, mpv can play hdr video fine but when i pause the frame gets overblown.
additionally, sdr video stays dark no matter how i adjust the sdr brightness in windows settings

it's over

>>102345500
i found out, looks like it's

target-colorspace-hint=yes

>>102345055
*smooch*

>>102345500
vo=gpu-next

File: 1723543448562606.jpg (81 KB, 1280x720)

81 KB JPG

mpv is minimalist tranny shitware
imagine literally removing all onscreen UI and actively going out of your way to destroy usability because of schizo obsession with MUH MINIMALIZM
you dumb fucks chopped off your UI like trannies chop off their genitals
you can't even tell if this shit is the active window
this is phone app shit mentality on desktop computers
fuck mpv and fuck you mpvniggers for fooling me installing this piece of shit
i want my 2 minutes back

File: __kafuu_chino_and_hoto_co(...).jpg (593 KB, 1040x1199)

593 KB JPG

>>102345055
Chino-chan a cute!

>>102347957
>removing
i love how this retarded hallucination is a thing now.

Dead thread is dead
Even the baitposters are shit

>>102345055
My wife

pedotroon thread

File: maxresdefault.jpg (129 KB, 1280x720)

129 KB JPG

>change speed
>dropped frames

>>102345055
>Next big update:
><github milestone>
lmao.
i'm glad i skimmed OP for a change.

File: 85b324a7bfac6c4a9b650dfde(...).jpg (1.3 MB, 1644x1760)

1.3 MB JPG

install chino

>>102345055
>#g-l script-binding select/select-subtitle-line
This one is really good. It would be even better if I can create a hotkey so I can go back or forward a line.

Link still doesn't work https://gitlab.com/chino-chan/ikatube

>>102351611
working link https://chino-chan.gitlab.io/programs.html#s2

Wrap that rascal with Firejail to prevent anal leakage.

https://wiki.archlinux.org/title/Firejail

https://firejail.wordpress.com/documentation-2/basic-usage/

Never listen to the rascals who give you "security" advice like >>102352124
which involves installing an untrustable privileged setuid binary with a big attack surface, that is if it's not intentionally compromisable.

It's trivial to create your own sandboxed environments with standard tools. Or you can use bubblewrap (without setuid) if you must, as it's small, unprivileged, and doesn't hide anything from you.

- https://github.com/containers/bubblewrap#related-project-comparison-firejail

profile=high-quality

Rate my config.

why do I occasionally get video not sync with the audio when start playing livestream from ikatube? see this on livestream news when people lip not sync with the voice, have to close player then play again to get synced playback.

>>102352643

rebase-start-time=no

should fix it

File: 1725248540324323.jpg (35 KB, 463x260)

35 KB JPG

>>102352524
Utterly pointless.

profile=high-quality

currently has no reason to exist. In the past, gpu-hq provided options such as --dither-depth, --sigmoid-upscaling... etc. which were objectively "high quality" but not enabled by default, and also not readily evident that they were "high quality" to the average layman enduser who doesn't know what a colorspace is. So the old version of this profile provided a convenient flag for people to set-and-forget. But now all of the important options within gpu-hq are already enabled by default in mpv,

profile=high-quality

serves no purpose.

Let's go over each option:

>scale=ewa_lanczossharp
Looks identical to Lanczos 90% of the time, and the other 10% of the time it looks worse than Lanczos due to being less sharp. Considerably slower than Lanczos for no readily apparent benefit.

>hdr-peak-percentile=99.995
The only option within high-quality that is objectively good.

>hdr-contrast-recovery=0.30
Completely subjective whether or not contrast_recovery is necessary for every HDR source. It can lead to blown-out highlights, which is certainly not "high quality".

That is ONE (UNO) option that is objectively "high quality" within

profile=high-quality

. Everything else within it is either questionable or a downgrade from default.

>>102352706
thanks, seems work so far. minor issue with this is the time left display on osc. the value does no seem to reflect to what cache left in seconds. is this expected behavior?

>>102353193
it seems that mpv doesn't add the start time of livestream to total time in this case so the calculated remaining time is wrong

>>102353097
Correct. Throughout the discussion about the profiles, most of the options were enabled by default after evaluating that its performance impact is negotiable currently. The

profile=high-quality

setting enables various useful, though potentially resource-heavy, options. Currently, it doesn't include much, but the idea is that we can add new features to it without requiring users to update their configuration. It’s a "set it and forget it" preset. The

hdr-peak-percentile

setting alone justifies the existence of the HQ profile.

It's important to understand that dynamic peak detection greatly improves HDR tone-mapping quality. However, the histogram-based algorithm is too heavy for smaller iGPUs, which is why we default to using 100%. This allows us to enable dynamic tone-mapping by default without a performance penalty. Of course, there is a tradeoff: a single pixel artifact or a specular reflection can affect the reading, which isn’t ideal in practice. That's why the histogram-based version exists and is enabled in the HQ profile.

Arguably, contrast recovery is also useful. According to haasn, it restores detail and contrast that might otherwise be lost during the tone-mapping process. This is also what madVR does, and everyone seems to like it. Lastly, polar lanczos is less prone to artifacts. Users don't need to understand all these details; they just want to enable the HQ profile and get what it promises.

Default profiles are there to cure cargo-culting options like

scale=ewa_lanczos
scale-blur=0.981251

which most users don't understand. You can adjust all you like, but spoonfeeding by random blogs/guides is considered harmful and this is mpv job to have entry level presets that don't require user to refer to 3rd-party resources.

>>102353560
There are three profiles:
- fast: This profile uses direct sampling without any processing, designed to support any hardware that can output pixel on the screen
- fefault: This is a balanced option for standalone media player, offering good color accuracy and scaling, while being fast enough to support even 8-year-old Intel iGPUs
- high-quality: This option is intended for dGPUs or APUs, enabling features that are known to be more resource-intensive but still improve quality

These profiles are subject to change, but their core purpose and description will remain the same.

>>102353560
>>102353591
All you have to do is remove --scale=ewa_lanczossharp from high-quality and keep the profile only for HDR-specific options. That way the profile has a more clear goal in mind (exposing users to useful HDR options) rather than just being a weird hodgepodge of subjective options that certainly don't fit the "set-and-forget" philosophy that gpu-hq used to fulfill.

>Lastly, polar lanczos is less prone to artifacts.
Blur is an artifact.

>Default profiles are there to cure cargo-culting options like
This is a bad example. For one, what you copypasted is what mpv ITSELF used to recommend when you set ewa_lanczossharp in gpu-next. So it's hardly a "cargocult". Now that ewa_lanczossharp has a filter definition again, what you posted is basically identical to ewa_lanczossharp. There's no harm to having that in your config file.

If I set hdr-contrast-recovery to 0.20, is that a cargocult? Maybe it just looks better to me. Again, the only objectively good option here is --hdr-peak-percentile=99.995. Everything else within the profile is either a downgrade or just subjective.

So mpv has to do two things:
>Disable --hdr-compute-peak by default if it's going to look like shit anyways and provide subpar results
OR
>delete high-quality, add --hdr-peak-percentile=99.995 to default, that way dynamic tonemapping is already good quality by default. Fuck potatos.

As it stands now, there's no benefit to the profile and it fails to fill the niche that

profile=gpu-hq

used to accomplish.

ChatGPT argument.

File: snide.png (158 KB, 512x512)

158 KB PNG

>Use mpv on Windows 10
>Get unavoidable memory leak
Delightful.

>>102354097
>windows
stop larping and just use netflix like a good goy should

>>102353097
ewa_lanczossharp looks better

>>102354588
There's no difference at 1x scale. You're mentally ill.

Lanczos has always been the better reconstruction filter overall. It's faster and it preserves high frequencies better. It also outperforms ewa_lanczossharp in pretty much every arbitrary test and metric.

>>102354737
>It also outperforms ewa_lanczossharp in pretty much every arbitrary test and metric.
Nice one ChatGPT, now provide numbers.

>>102354792
Imagine calling others AI when you just cargocult the "muh ewa_lanczossharp is le sharper because it has sharp in the name!" meme even though it's extremely common knowledge amongst most /mpv/ oldguard that Lanczos will always be sharper. Newfags don't deserve spoonfeeding.

how do I flip a video in mpv? I tried

set vf hflip

. It doesn't work anymore, it used to work.

>>102354919
disable hardware decoding

>>102354848
>Xhe thinks being sharp is the be-all and end-all
LOL!

https://gist.github.com/igv/8a77e4eb8276753b54bb94c1c50c317e
Just use this??

>>102354947
Good news newfag-kun, I have the perfect filter for you. It's called

bicubic

. Zero aliasing, no negative lobes, no ringing, just pure unadulterated blur.

>>102354971
DO NOT REDEEM

>>102354996
Listen faggot, I don't know how else to tell you this but
Blur > aliasing
End of story.

>>102354737
>There's no difference at 1x scale.
YEAH NO SHIT RETARD THERE'S NO UPSCALING DONE AT 1X SCALE KILL YOURSELF FAG

>>102355028
Lanczos is already the best orthogonal filter when it comes to anti-aliasing since it's just Sinc windowed with itself. At all real-world content scenarios, there's no instance where Lanczos will exhibit aliasing. It's not Catrom or Mitchell.

Again, you need to be checked into a mental asylum, because both filters look identical if you're not zooming in and pressing your eyeball against the screen. You tricked yourself into adding more blur because you were groomed into a cargocult. My condolences. Lanczos is for knowers, EWA LanczosSharp is for chumps.

>>102355143
1X scale in this instance means no zooming in when A-B testing... newfag retard. You can do an A-B test or a double blind test with two different images with no zoom (1x scale).

Also, even if I was referring to the actual scale factor of the upscale, you're still upscaling chroma by 2X because most videos are 4:2:0. Seething newfag.

I see the schizos are loose again.

I got this error

Cannot load libcuda.so.1
[ffmpeg/video] h264: Hardware does not support image size 3840x2160 (constraints: width 0-2048 height 0-2048).
[ffmpeg] AVHWDeviceContext: Cannot load libcuda.so.1
[ffmpeg] AVHWDeviceContext: Could not dynamically load CUDA
[ffmpeg/video] h264: Hardware does not support image size 3840x2160 (constraints: width 0-2048 height 0-2048).
[ffmpeg/video] h264: Failed setup for format vaapi: hwaccel initialisation returned error.
Failed to open VDPAU backend libvdpau_va_gl.so: cannot open shared object file: No such file or directory

When trying to open 4K video with fast profile.
I'm using old hardware, so is there something I need to install? or it's just warning message?

>>102355416
>is there something I need to install?
>Hardware does not support image size 3840x2160 (constraints: width 0-2048 height 0-2048).
Yes, you need to install new hardware or disable hwdec.

>>102354928
worked, thanks anon.

>>102355787
Ok, but then it default to SW, Right?
Shouldn't it work as usual?
>>102354919
You can use vappi-copy, or whatever HW decoder you use.

llyyr just got destroyed
https://github.com/mpv-player/mpv/issues/9551#issuecomment-2347058842

hello cake

>>102354737
ewa_lanczossharp looks better to me

>>102357191
see >>102355154

>>102355416
iirc 4K decoding of h264 is for some reason its own separate thing?
i think the [ffmpeg/video] lines are successful hwaccel creations that then fail because your hardware only supports square videos in actual 2K at max (2048x2048) while the two library failures are attempts to dynamically load hwaccel runtimes you don't have installed because it's scanning through all of them looking for a working one

>>102357238
Still looks better to me friend

>>102357347
Seek help. Spend more time watching content and less time looking at 80x80p test images. It's not too late for you buddy.

>>102357366
>lanczos
see aliasing
>lanczossharp
don't see aliasing

>>102357395
Seek help. You're hallucinating. You need to be institutionalized.

>>102357419
why does it upset this much that someone chooses a filter that according to you is "identical to Lanczos 90% of the time" which is presumably your favorite? do you get this upset if someone uses bilinear or some other filter that looks even more different?

>>102357457
No because bilinear is a useful filter. ewa_lanczossharp has no usecase and is inferior to lanczos. Why is there so much dripping hatred in your post? What's going on friend?

>>102357501
>has no usecase
less ringing. usecase settled. glad we could come to an agreement

>>102357297
So do i need to install it then?
https://archlinux.org/packages/extra/x86_64/libvdpau-va-gl/
But I'm using iGPU from like 2nd or third gen.

>>102357517
>less ringing.
Actually it rings more, the filter is simply so blurry that it blurs its own halos, adding large and blurry halos to your image. Look at the filter's radius before you make posts like this friend, Jinc's zero-crossing is at 3.2~.

Also, who said ringing is bad? You do realize that you can't recover and reconstruct high frequency information from the original image without some amount of overshoot, right? Low-passing is good because it anti-aliases the image, but it's also bad because you're blurring the image, which a modicum of ringing helps counteract.

>>102351611
not installing your closed source shitware faggot

File: DWFZtxbMc7aMpbm3eDwv.png (3 KB, 158x148)

3 KB PNG

Let me guess, you need more?

>>102357739
I need less. Delete all of that bloat

>>102357574
no, there's no point, even if they could work as libraries (CUDA is specific to NVIDIA GPUs, vdpau on the other hand is generic unix) they're not going to add hardware capabilties you don't have
hardware accelerated media codecs aren't as flexible as software defined ones
>iGPU
intel's hwdec library is VA-API (which is also generic they just have the main version) which did load so you've got everything you would potentially need installed
if the error's actually preventing you from playing videos you're going to need to disable hwdec

https://trac.ffmpeg.org/ticket/11179
Who the fuck is Balling?

>>102358273
valzapod and balling are the same people

>>102357589
>Also, who said ringing is bad?
ringing looks le bad

>>102358273
>Who the fuck is Balling?
If you don't know Balling you are missing out. They are active on ffmpeg trac and have severe mental issues. I believe they are banned on mpv repos, they wanted to ban them on ffmpeg too, but they are not reading the trac anyway so no harm is being made.

>>102358273
>Also haasns swscale work is funded by STF currently
they took haasn from us

>>102345055
Doyakonga

>>102359221
valzapod has several alts on the mpv repo (he even made an issue not too long ago). he's not very subtle

is it necessary to use the yt-dlp custom ffmpeg build with it?

File: ffmpeg-yt-dlp_patches.png (14 KB, 830x480)

14 KB PNG

>>102360161
only if you're Chinese

File: 15a1b49c26f5c684807a2f0b8(...).jpg (410 KB, 1900x2500)

410 KB JPG

>>102345055
>If you're having performance problems with console.lua or stats.lua, use console:
how is libass drawing text slower than terminals?

>>102362389
because libass is high quality and terminals are not

Just when you thought it couldn't get any worse, mpv maintainers fucking jerk off to each other's GitHub profile pictures occasionally whenever they see each other in GitHub issues instead of actually paying attention to whatever the issue creator wrote. Can mpv be saved at this point?

dumbfuck tourist here, is there a windows version with all the goodies in one zip or exe?

>>102363483
https://github.com/stax76/mpv-hero/releases/download/v1.0/mpv-hero-v1.0.zip

What is the best player for iOS?

File: Buff Bagwell.jpg (248 KB, 800x1200)

248 KB JPG

>>102360226
>Bagwell

>>102363483

https://github.com/shinchiro/mpv-winbuild-cmake/releases

how did ilyyr go from being useless to gaining the knowledge to fix dxvk and WSI bugs overnight

>>102365435
Useless fixes

>https://github.com/mpv-player/mpv/pull/14808
lgtm
>https://github.com/mpv-player/mpv/pull/14815
close
>https://github.com/mpv-player/mpv/pull/14831
close
>https://github.com/mpv-player/mpv/pull/14832
close and ban
>https://github.com/mpv-player/mpv/pull/14834
close

>>102365435
you didn't fix anything.

>>102365756
Yeah no shit faggot why would I contribute to lincucks toy OS for free?

>>102365761
talking about yourself in 3rd person is a mental illness you know

>>102365784
holy schitzo

>>102365800
>schitzo
only llyyr spells it like this, yep, basically confirmed. seek help asap

What makes videos from here, Youtube and Gelbooru look like shit?, is there a way to fix them on MPV? (Sorry i have zero knowledge of video-related technology)

https://github.com/mpv-player/mpv/pull/8130
close

>>102345055
the chinowife

>mpv devs seriously considering using fragile commandline programs + lua scripts for clipboard support
no wonder how mpv becomes shit with a bunch of script kiddies

>>102366903
Example?
If the videos are garbage quality (compression artifacts, low resolution), there's not much to be done. Garbage in, garbage out.

>>102363613
>>102364366
I'll check them out. Thanks anons.

File: file.png (5 KB, 644x73)

5 KB PNG

>>102350907

>>102368657
It's compression artifacts i think, every YouTube video has it even if the source is extremely high quality

File: wifey.webm (1.25 MB, 360x640)

1.25 MB WEBM

Can I get some advice on improving VP9 encoding and reaching the 3 MB limit with decent quality?
This is what I'm using

ffmpeg -i "input" -pix_fmt yuv420p -an -c vp9 -rc_mode 1 -crf 44 -cpu-used 4 -aq-mode 2 -tune-content 1 -g 350 -deadline 0  "output".webm

>>102371203
i'll tell you when you tell me who you copied
you're clueless. the limit isn't even 3, it's 4 MiB

>>102371203

ffmpeg -i 'input' -map 0:v:0 -pix_fmt yuv420p -c:v libvpx-vp9 -row-mt 1 -cpu-used 2 -crf 30 -b:v <tbr>k -pass 1 -f null -
ffmpeg -i 'input' -map 0:v:0 -pix_fmt yuv420p -c:v libvpx-vp9 -row-mt 1 -cpu-used 2 -crf 30 -b:v <tbr>k -pass 2 'output.webm'

calculate the target bitrate (tbr) according to the duration: tbr (kbps) = 0.95 * 2^25 / 1000 / duration (seconds)
use the scale and fps filters to reduce the resolution/fps of high resolution/fps sources
don't make the webm too long, 1:30 is already quite long, only so much you can do within the 4MB limit

how do i use openal-soft htrf audio on mpv? i use shinchiro's build, have downloaded and placed

 soft-aol.dll 

both in System32 and mpv's folder just in case, enabled htrf in alsoft-config, but it doesn't work.
what am i doing wrong?

here's how you fix profile=high-quality: add --scale-antiring

>>102371203
https://desuarchive.org/g/thread/101428428/

>>102372083
>The Definitive Webm Guide For Retards
I love it, bookmarked

>>102371394
From someone in these threads, but when I tested it, it was bad.
>>102371606
Thanks, I'll try two pass with your config.
Why use map though?
And was wondering is pix format necessary?
>row-mt
No wonder encoding was so slow.
>1:30 is already quite long
I know, but on average it's good enough, since the limit is 300 sec.
>>102372083
Thanks senpai.

>>102372083EVEN THOUGH the the tangent about scaling is wrong: if the encoder is bit starved enough, feeding it lower resolution will make the resulting webm higher quality than if it was encoded unscaled and scaled after
the issue is that there isn't really a DeFiNiTiVe guide for resolutions per bitrate ranges. the most retardproof would be to always downscale to 720p. a long video will at least look watchable at that resolution
you can also use -speed/-cpu-used 0 (default is 1) to squeeze a tiny more quality at the expense of it taking longer

Not quite mpv but I don't see a yt-dl thread up right now.
What youtube-dl fork do you use? I was using yt-dlp but it's now insanely refusing to download porn for me. https://github.com/yt-dlp/yt-dlp/blob/master/yt_dlp/extractor/unsupported.py
(I will watch these videos with mpv once they are downloaded)

>>102372512
It's open source, you could just remove the blacklist yourself.

Also, porn is for degens. Based yt-dlp.

Which cscale for CG?

Goddamn where did it all go so wrong?

>>102372648
Explain right now sister.

https://github.com/mpv-player/mpv/milestone/7

is it possible to downscale a video (1080p source with

--vf=scale=1280:720

) and then upscale it with

--glsl-shader

?

>>102345055
I look like this

>>102372999
https://github.com/Jaded-Encoding-Thaumaturgy/vapoursynth-descale

>>102373026
chinochan...

>>102372227
yeah, but then you get those portrait videos.
I swear encoders don't know how to handle those.
Simply rotating it to landscape I get better file sizes.

File: waterfall_1280x720p50_YUV(...).webm (3.8 MB, 1280x720)

3.8 MB WEBM

>>102372200
>Why use map though?
gets rid of subtitles in case there are any, though you can also use -sn if you want
>And was wondering is pix format necessary?
maybe not
4chan used to block other formats in the past, nowadays at least yuv420p10le definitely works (trying yuv444p12le... webm related) and that's probably the only other one you might realistically come across
still, yuv420p is the most widely used and supported format

Is there any real reason the use MPV on Android?

>>102373645
use case for watching movies on a tiny phone screen with terrible color reproduction?

enshittification just got merged by brapfan5

https://github.com/mpv-player/mpv/pull/14731

now you cant set blur-compat=no anymore without breaking subtitles

DO NOT UPDATE AND/OR REVERT

>>102373932
> <Oneric> Any more opinions on https://github.com/mpv-player/mpv/pull/14731 ?
> <sfan5> let's merge it?
> <llyyr> lgtm
guess who is behind this again

>>102374558
who is funding him?

>>102373645
what else is there?

>>102373775
last time I checked, most high end phones had really good color reproduction (provided you use the "natural" color setting or whatever it's called)

>>102374705
phones will never be ideal for videos

>>102375505
PRODUCES MUSTARD GAS DONT CLICK

>>102375505
Which one of you are into this?

mpv --show-profile=high-quality
 scale=ewa_lanczossharp
 scale-antiring=0.6
 hdr-peak-percentile=99.995
 hdr-contrast-recovery=0.30
 blend-subtitles=yes

ship it for v0.39.0

>>102375768
you're a skibidi ohio cringe L numpy troon or something if you aren't into that stinking ass

>>102373932
What option do I use to fix crunchy rips now? blur-compat=no fixed some wonky typesetting.

>>102374558
Report issue before it's too late

>>102377795
>>102377958
It's over (but unironically). Again, just don't update. You can't fix CR softsubs without breaking aspect ratio now.

>>102377795
>>102377958
>>102377987
post a file that it broke

>>102378030
cant be arsed, already reverted it on my end.

>>102377795
>>102377987
isn't --sub-ass-use-video-data=aspect-ratio the same as --sub-ass-vsfilter-blur-compat=no?

>>102378807
yes, they're retards who can't read. you can't disable aspect-combat without disabling blur, which is a nonsense combination of options. you can still disable blur-compat without disabling aspect-compat.

>having less options is... le good!
wow another riveting bikeshed update from mpv, where another core functionality of the player has been needlessly deprecated in lieu of an inferior option.

reminder that all of the sub settings have also been needlessly changed and the old aliases and settings can no longer be replicated.

i have no time to read hundreds of comments.
any major happenings/drama in the last week?
(including ffmpeg, librempeg, ...)

"patches are welcome" always sound cringe, not sure why

>>102380047
it may or may not be cringe in other cases. but
there is one case where it's 100% warranted. it's when the reporter/arguer starts saying stuff like "why aren't WE doing this and that"

>>102380047
because it's polite rejection to feature that no one wants, it's a trouble when someone actually sends a patch

>https://github.com/mpv-player/mpv/pull/14839#issuecomment-2350816349
Was it cringe?

remind me again why sub-ass-use-video-data=aspect-ratio (sub-ass-vsfilter-blur-compat=no) is not set by default?

File: izzy.webm (3.81 MB, 1080x1920)

3.81 MB WEBM

>>102371606
Thanks man.
This works great, though I'm getting like 1.5 frame/second during encoding, I guess this is the limit of my hardware (2nd gen intel)

>>102382094
more of her. or who is she.

>>102381376
Instead of mpv providing information about the video for sub rendering, now libass has to guess instead. And the guess can be wrong.

File: izzyygreennxox_[C_oMFQHSMHV].webm (1.68 MB, 360x640)

1.68 MB WEBM

>>102382107
filename my man.
Another question, when doing scaling for the video with two passes
Should you add the arguments to both pass or only the second pass?

>>102382094
>>102382804
she needs some more meat, she's all bones

File: 569272_v2.gif (1.95 MB, 350x196)

1.95 MB GIF

>>102382907
>more meat

nextsisters? https://github.com/mpv-player/mpv/issues/14442

>>102383108
broken subs

>>102383108
Does blend-subtitles=yes work on gpu-next?

>>102383232
works fine

>>102383374
Then I don't care, don't see the point in video.

>>102376126
lgtm, ship it

>>102383530
it's the only way for high quality subtitle scaling if you watch dvd/bluray with vosub/pgs image subtitles
without blend-subtitles=video you're stuck with bilinear with subtitles

>>102383586
ogey

>>102383586
and you need more than bilinear because...? they're fucking subtitles. imagine wanting nasty lanczos ringing on your vobsubs

>>102383586
also it doesn't rape your cpu if you watch anime with complex typesetting on 4k screen
>>102383639
>and you need more than bilinear because...?
nasty aliasing with downscaling (it's not correct-downscale)
>imagine wanting nasty lanczos ringing on your vobsubs
i use catrom

>>102383684
catrom is bad for upscaling

>>102383746
https://github.com/jsummers/imageworsener/blob/master/technical.txt

By default, IW uses a Catmull-Rom ("catrom") filter for both upscaling and
downscaling. Why?

For one thing, I don't want to default to a filter that has any inherent
blurring. A casual user would expect that when you "resize" an image without
changing the size, it will not modify the image at all. This requirement
eliminates mitchell, gaussian, etc.

The "echoes" produced by filters like lanczos(3) are too weird, I think; and
they can be too severe when using proper gamma correction.

When upscaling, hermite, triangle, and pixel mixing just don't have acceptable
quality. That really only leaves catrom and lanczos2. I somewhat arbitrarily
chose catrom over lanczos2 (they are almost identical).

When downscaling, the differences between various algorithms are much more
subtle. Hermite and pixel mixing are both reasonable candidates, and are nice
in that they have no ringing at all. But they're not quite as sharp as catrom,
and can do badly with images that have thin lines or repetetive details.

>>102383805
>Jason Summers
This guy is a hack who literally just decides what looks "good" based on his feefees. He completely omits:
>sigmoidal colorspaces
>aliasing
>moiré rtifacts
From his """"technical"""" """"""''discussion"""""""

Link an actual mathematician and image processing expert like Robidoux.

File: perhaps.png (282 KB, 512x512)

282 KB PNG

Will dscale AR ever be fixed?

File: 1704043881034219.png (21 KB, 888x849)

21 KB PNG

>>102347957
>le terminal options... IS SOMEHOW LE TINKERTRANNY!
You have to end your life projecting, hijra. Right now.
And it's not because your parents would miss you (they won't), but for the good of everyone around you. KILL YOURSELF, child grooming pedo tranny. I mean it!

i have llyyr fatigue

are chink builds still shit

File: __puru_two_and_banagher_l(...).jpg (138 KB, 800x600)

138 KB JPG

>>102383805
>>102383864
>https://github.com/jsummers/imageworsener/blob/master/technical.txt
Terrible read. Read this instead:

https://legacy.imagemagick.org/Usage/filter/nicolas/

>When resizing, I (Nicolas Robidoux) generally stick to either a tensor (orthogonal, or 2-pass) Resize (-resize) Operator filter, or an EWA (Elliptical Weighted Averaging) Distort Resize (-distort Resize) operator.

>If you want to use one single filter for everything and would rather avoid complications like color space manipulations and fancy parameters, use the LanczosSharp EWA (Elliptical Weighted Averaging) Filter. It produces images that are slightly blurry, with no jaggies to speak of and reasonably mild halos.

>For much sharper results, use the tensor (orthogonal or 2-pass resize) Lanczos Filter.

>Of course, halo and blur are not the whole story: Most importantly, there are also moiré, tonal drift and "jaggies" to worry about.

>On the other hand, I do not recommend upsampling (enlarging) through linear light. That is: You generally should not use standard linear light Colorspace Correction when enlarging images, at least when using filters with significant negative lobes (any filter that introduces haloing, which means most of them). Compare the RGB and XYZ (linear light) enlargements to the sRGB (not color corrected) and sigmoidized (custom color corrected) enlargements shown in the Enlarge with sRGB, RGB, LAB, LUV, XYZ, sigmoidal...? ImageMagick Forum thread. The linear light enlargements (RGB and XYZ) are almost always significantly worse. The sigmoidized elargements generally look best, although enlargements through Lab are sometimes even better. Enlargements through linear light often look so bad that you are better off ignoring color space issues and filtering sRGB pixel values directly (using commands like the above).

0.39 is getting closer

>>102384332
blocked by >>102376126

will 0.39 live up to 0.29 in quality and stability?

>>102384397
Yes.

>>102384397
only when wm4 rises from his grave and gives his his seal of approval

>>102384397
with abandoned libplacebo as dependency, no

>>102384513
0.29.1 has d3d11 and vulkan support which is proto-placebo

>>102347957
>you dumb [curse word] chopped off your UI [needless drivel]
press the up and down arrows to adjust the volume press the left and right buttons to adjust the time. they will show up
>you can't even tell if this [curse word] is the active window
who cares

nothing complaints

>>102381376
What the fuck is even the point of this option? Who cares about some ancient vsfilter shit?

>>102385162
what a stupid statement

>>102385311
I'd rather fix any broken ancient subtitles myself than bother with any of this garbage.

>>102385359
that's not what the option does at all, which is why the libass devs refactored the option and reworded it. your small tard brain only sees "vsfilter" and the "compat" and thinks that it's some kind of backwards compatibility thing, which you have an irrational hatred towards (incelish tendencies?)

try again

>>102385416
Is this some incredibly petty semantic argument about how you can make a broken file today or some shit? The option was originally introduced as a backward compatibility hack ages ago for ancient VSFilter behavior and broken files. "sub-ass-vsfilter-blur-compat" was misleading because it did more than just "blur" retard. stop gaslighting me.

>>102385529
it also did more than just "vsfilter" "compatibility", retard. every issue on the tracker regarding blur-compat proves that blur-compat=yes is the accurate and closer to source behavior, while blur-compat=no is not.

some subtitles are just broken, simple as. if you have fun fixing them yourself then knock yourself out. you cant expect users to have the same autistic fervor when they just want to watch a video file and expect it to work

>>102385596
>every issue on the tracker regarding blur-compat proves that blur-compat=yes is the accurate and closer to source behavior, while blur-compat=no is not.
Literally the first one I randomly click shows exactly the opposite. What a complete waste of my time. Kill yourself.
https://github.com/mpv-player/mpv/issues/6517

File: file.png (134 KB, 1206x601)

134 KB PNG

>Goes on a months long autistic mission to create chroma shaders from the white papers
>Doesn't recommended either of his shaders
Was it autism?

>>102385638
HAHAHA you fucking illiterate retard, read the issue. the subtitle file is literally broken and blur-compat=no fixes it. just proved my point

im actually talking to a chatgpt bot holy shit

>>102385658
>Was it autism?
Always has been.
ArtCNN chroma is actually quite interesting in what it does, but at the same time totally pointless. Bicubic for chroma is enough, go lanczos if you want sharper, with antiringing.

>>102385708
>Bicubic for chroma is enough
enjoy your chroma blur and chroma smear.

at the very least recommend catrom instead...

>>102385708
>>102385728
I personally use:

cscale=ewa_lanczos4sharpest

>>102385748
retarded

>>102385804
I have a 3080, the timings barely change when upscaling to 4K and my GPU usage stays under 10%, who cares.

>>102385828
it has nothing to do with that. you can use ewa filters on a radeon hd 7850 and they work just fine. the filter just sucks.

>>102385699
Are you ESL? That is clearly an example of where "blur-compat=yes" is NOT the accurate and closer to the source behavior.
>the subtitle file is literally broken
Yes exactly. It should be fixed instead of needing hacks. That's the whole point.

>>102385872
hmm, Who do I believe, haasn and kasper or a random anon...

Are hwdec=vulkan and gpu-api=vulkan always the best options on modern hardware? I usually set them to auto.

>>102385886
>That is clearly an example of where "blur-compat=yes" is NOT the accurate and closer to the source behavior.
its reading the sub file correctly. the sub file is broken, hence why you have to disable video data/params so that libass inserts bogus data and makes the sub file agnostic to video data. this is clearly an instance where somebody just copied over a sub file from some kind of anamorphic video and just slapped it onto a bluray without adjusting the sub params.

>Yes exactly. It should be fixed instead of needing hacks. That's the whole point.
blur-compat=no IS the hack you fucking stupid retard. go back to discord

Both FSRCNN and RAISR(RAVU) were published 8 years ago, has nothing better has come out? Hard to believe, specially now that almost all GPUs have deep learning accelerators.

>>102383864
>>102384206
Even Robidoux calls Catmull-Rom "pretty good":
>You can eliminate the second halo completely using sigmoidized -distort Resize and -resize with Keys cubics. These filters are discussed further in the downsampling section below. I recommend them more for downsampling than upsampling. Nonetheless, they are easy to adjust to taste, and give pretty good results. Start with tensor Mitchell (-filter Mitchell -resize) and EWA Robidoux (-distort Resize), and give a try to EWA RobidouxSharp (-filter RobidouxSharp -distort Resize) and tensor Catmull-Rom (-filter CatRom -resize) if you want something sharper. (Only use -resize with HDRI.)
And his discussion is specific to image resizing with an ancient version of ImageMagick. He preferably recommends EWA resizing because at the time, the tensor resizing implementation in the default build of ImageMagick had shit quality. He makes no mention of things not available in ImageMagick (e.g. NNEDI3). He doesn't discuss performance, which is important for realtime video resizing.

>>102385892
haasn is retarded (he can't even see very obvious color banding artifacts anymore after he got a gf) and kasper doesn't like or recommend ewa_lanczos4sharpest for all of the reasons that this thread already concluded ages ago. the antiring strength is set too high and the filter has more aliasing than ewa_lanczossharp.

>>102385963
>its reading the sub file correctly
Nobody disputed that. But in this case it is not accurate because the person who made the subtitle file pretty obviously fucked up. It's not complicated.
>blur-compat=no IS the hack
That's what I was saying the entire time.

>>102385998
>he can't even see very obvious color banding artifacts anymore after he got a gf
New theory, semen retention increases visual acuity

>>102385914
vulkan hwdec is not stable at all for me, but I use Windows.

>>102385914
Vulkan is not that stable, depends on hardware and drivers. It is good fallback if others like vaapi or d3d11 hwdec fails for some reason, but not a first choice.

>>102385914
>>102386070
>>102386088
it is also not lossless, compare a png SHA-256 with swdec va hwdec, and they have different hashes (disable dithering and deband obviously)

d3d, vaapi, nvdec are all lossless and have the same file hash swdec vs hwdec

>>102386118
This should like a bug. Did you compare the images itself? Is the difference on the borders? I suspect 1920x1088 might have dirty border on the bottom. Else I Vulkan shouldn't really produce different result.

>>102386164
it seems to be an issue with how gpu-api=vulkan handles built-in shader LUTs, because vulkan with d3d11va-copy also produces a different SHA-256 for hwdec vs swdec

File: mpv-shot0001.png (2.01 MB, 1440x1080)

2.01 MB PNG

>>102386194
proofs that there's some kind of lossiness with shaders and p010/nv12 format conversion with vulkan

mpv --no-config --vo=gpu-next --gpu-api=vulkan --dither-depth=no --screenshot-format=png --screenshot-high-bit-depth=no --hwdec=no

sha-256

f402699e27a94dce1d4a911e57ed0a8fd63c9268fe1290adbed8f42e34474713

File: mpv-shot0002.png (2.01 MB, 1440x1080)

2.01 MB PNG

>>102386573
native vulkan video

mpv --no-config --vo=gpu-next --gpu-api=vulkan --dither-depth=no --screenshot-format=png --screenshot-high-bit-depth=no --hwdec=vulkan

sha-256

ec69b1fe40edbfa0c8ca4ebbdf7ecb8ebbd9b005d32ba75f7601fbafb948014e

>>102386585
d3d copyback

mpv --no-config --vo=gpu-next --gpu-api=vulkan --dither-depth=no --screenshot-format=png --screenshot-high-bit-depth=no --hwdec=d3d11va-copy

sha-256

ec69b1fe40edbfa0c8ca4ebbdf7ecb8ebbd9b005d32ba75f7601fbafb948014e

(can't post image here since it's identical to vulkan video hash)

Do most other video players like VLC use bilinear scaling?

>>102386810
vlc uses bicubic with b=0/c=0.6

File: 86.webm (1.82 MB, 640x360)

1.82 MB WEBM

Will this be fixed with the latest webtorrent-cli update?
https://github.com/noctuid/mpv-webtorrent-hook

>>102387090
Just use https://miguapp.pages.dev

File: 1726328655350856.webm (3.82 MB, 540x960)

3.82 MB WEBM

>>102371606
Based anon.
I've been using this

ffpb -i "$1" -pix_fmt yuv420p -an -c vp9 -rc_mode 1 -crf 44 -cpu-used 4 -aq-mode 2 -tune-content 1 -g 350 -deadline 0  "$2".webm

And it was shit.

>>102387132
244 KG is crazy

>>102387108
Why?
I want mpv.

>>102387147
Its mpv.
For mobile.

>>102387132
jesus, it's so fucking over for america

>>102387100
>>102387108
buy an ad you faggots

>>102387141
It's American KG
So it's like 110 KG for the rest of the world.
Still double the normal weight of a woman.
>>102387203
It is, and this will be the norm, and what's worst the rest of the world will follow.
And you will even be called a pedophile if you suggested you like women of healthy weight.

File: 1449249291274.gif (1.06 MB, 500x450)

1.06 MB GIF

How bad is this?

ffmpeg  -hide_banner -loglevel error  -i "input" -threads 4 -c:v libvpx -crf 26 -qmin 5 -qmax 30 -b:v 1M -maxrate 1M -bufsize 1M -vf "scale=-1:480" -an -auto-alt-ref 1 -pass 1 -f webm /dev/null
ffmpeg  -hide_banner -loglevel error -i "input" -threads 4 -c:v libvpx -crf 26 -qmin 5 -qmax 30 -b:v 1M -maxrate 1M -bufsize 1M -vf "scale=-1:480" -an -auto-alt-ref 1 -pass 2 -f webm "output".webm

>>102372083
What video is this?

>>102385708
Woah you weren't kidding, this feels like putting your glasses on: https://slow.pics/c/90JgmLDV
Look at Kana's hair or at her clothes, jesus christ.

>>102387526
looks identical..

>>102387297
Why /g/ still allow only webm? It is so bad...

>>102387656
as opposed to what exactly?

>>102387656
Good night saar, go to sleep

>>102387681
av1/vvc/hevc/h264

>>102387695
Bloody bc I told you to go to bed

>>102387695
>hevc/h264
worse compression than vp9
>av1
already deprecated and abandoned
>vvc
proprietary and no hardware decoding support

>>102387695
>av1
You want this instead of "webm"?

>>102371606
>>102387132
>>102371606
I just encode with crf multiple times until I hit the target file size.

>>102387656
VP9 is just as efficient as HEVC.
Hiro should enable AV1 support though.

File: 1720778107745450.png (453 KB, 753x576)

453 KB PNG

How to remove osc bloat from the bottom right?

>>102387297
>

-vf "scale=-1:480"

Unironically use libplacebo for scaling, much better quality.

My command for encoding WebMs:

-vf "libplacebo=w=iw/2:h=-1:downscaler=hermite" -c:v libvpx-vp9 -lag-in-frames 25 -arnr-maxframes 5 -arnr-strength 1 -deadline good -cpu-used 0 -enable-tpl 1 -auto-alt-ref 2 -tile-columns 0 -tile-rows 0 -aq-mode 0 -g 300 -row-mt 1 -tune-content default -b:v 0 -crf 41 -pass 1 -an -f null -

I adjust the key frame interval according to video frame rate.

>>102387807
>hermite
omg i love blur

>>102387840
hermite doesn't blur, but if you don't like it just change it to mitchell or catmull_rom

>>102387854
>hermite doesn't blur
LMAO retard detected

mitchell is even blurrier than hermite btw. what a dumb recommendation.

>>102387873
ogey, just change it to whatever you prefer, enjoy your rings.

Why is my font blurry and not crisp?
Is it font issue? this is what I used

force_style='Fontsize=12,Alignment=2,MarginV=10,Fontname=DejaVuSans

>>102387894
you can have sharp downscaling without ringing, it's called bilinear. but you're just a cargoculting retard so you wouldnt know

>>102387947
enjoy your aliasing retard

>>102387947
>recommending BILINEAR
HAHAHAHAHAHAHAAHAHAHAHAHAHA
Ophthalmologist, now.

>>102387958
>>102387970
bilinear doesn't alias unless you go above a 2x downscale factor, so for 1080p-540p or 2160p-1080p downscaling its the most ideal filter that doesnt smudge high frequency information. everything else is either blurry or will ring horribly.

you can also use a box filter and that provides similarish results. retains source sharpness without any artifacts from using meme shit like hermite or mitchell

you cargoculters are subhuman

>>102388002
hmm, who to trust, haasn or a random anon spewing shit on 4chan

>>102387297
Why use both crf and b:v?

>>102388019
haasn has literally never made any comments about this. i dont know why you keep saying this retard tier shit.

btw hermite has moire issues when you go above a 2x downscale factor as well because its only radius=1. so you either get sharpness and moire, or blurriness and moire. pick your poison.

They should make some sort of thing that can remove ringing, like a de-ring? Call it like --de-ring or something. You should just be able to clip the pixels really.

>>102388049
it already exists

vf=pp:dr

>>102388049
Imagine this working for dscale.

>>102387807
>All those fancy settings
How much of a difference does it make?

>>102388071
it adds more blur. thats the only difference

>>102388030
You shouldn't?

hmm haasn told me to gobble a billion cocks, so i think ill go and do that right now. not a cargoculter btw

>>102385998
ewa_lanczossharp is great

>>102388116
t. CounterPillow

https://github.com/mpv-player/mpv/pull/14847

NEW SUB ENSHITTIFICATION

DO NOT MERGE OR I WILL START DISTRIBUTING MY OWN FORK OF MPV

File: 1712119840757220.jpg (2.51 MB, 3840x5760)

2.51 MB JPG

Do you think she uses mpv?

>>102387341
the filename is the youtube ID
>>102388071
cargoculted nonsense that they don't even grasp
trust occams razor and follow >>102372083
look up each option in the manual

>>102388234
Affects 0.001% of mpv userbase

>>102388529
>erm no one watches dvds anymore chud

File: 1662827442281743.gif (2.55 MB, 307x307)

2.55 MB GIF

wtf happened to ffmpeg mvp and the other one
Anyway, I have 2 TBs of H.264 streamer VODs I would like to encode to NVEnc H.265 using handbrake, my main worry is the loss of quality by encoding. How do I set it up so that 99.9% of the video is retained and what do I use to double check, something like a difference map

>>102388599
>99.9%
impossible

>>102388599
buy more storage, don't bother encoding all that.

>>102387807
>>102387854
mental retardation (<100 IQ):

dscale=hermite

>knows literally nothing about resampling, pretends like he's an expert
>doesn't even know what moire is
>"b-b-but it doesn't ring!! muh halos!"
>probably got into encoding in the 2020s

normal (100-120 IQ)
>doesn't care

knowers (120-130 IQ):

dscale=mitchell

>knows some theory behind resampling, has enough hubris to admit that he's not an expert
>knows that mitchell is the best filter in linear light
>knows what moire is, and why it should be avoided in video content (causes temporal instability)

enlightened (130-140 IQ):

dscale=lanczos
linear-downscaling=no

>knows that linear light is just as much of a meme as gamma light (robidoux self-admitted to it)
>knows that trying to pursue source accuracy is futile because there are too many random video files on the internet that haven't been gamma-corrected in the first place
>uses whichever kernel preserves the most high-frequency information, because it makes the most sense perceptually
>friends with knowers

the chosen ones (140-??? IQ)

dscale=bilinear
linear-downscaling=no
correct-downscaling=no

>knows that this is literally what every single other application and website uses
>"why bother?"
>fast
>all high-frequency information, no blur
>"just get a bigger monitor"

>>102389159
>

dscale=hermite

Yep, I still don't know why haasn pushed it, I think the resoning was that it is good enough, because there is not difference when downscaling anyway, but it is faster, because of fancy non-negative weights optimization.

>>102389578
most of us are using catmull_rom probably

File: glute.webm (1.84 MB, 324x576)

1.84 MB WEBM

In ffmpeg, how to get these subtitle style?

https://github.com/mpv-player/mpv/issues/14849

babe wake up new waymeme shitmess

>>102345055
How in the hell you manage to make it NOT AUTO CLOSE when finishing playback?

>>102389578
>because of fancy non-negative weights optimization.
hermite_fast in libplacebo isn't even implemented in mpv

>>102389768
--keep-open

>>102388664
Considering h265 can be lossless I don't really believe that, I'm just looking for a step below it

>>102389904
if you do lossless h265 i can guarantee your filesize will be higher than original lol. again, don't bother

>>102388030
to not unnecessarily bloat really short or easy to encode clips
it's called constrained quality mode

>>102388071
half of the settings are redundant because they're the default

so is CatRom and approximation of Lanczos2 or is Lanczos2 and approximation of CatRom?? who came first

>>102390032
>it's called constrained quality mode
I know, but why not just do single pass crf for those? And two pass for when it is harder to hit target bitrate?

>>102390167
I don't see the benefit of that.
First pass of 2pass is quite fast. Best case, 1pass crf isn't much faster. Worst case, it turns out too big and you run 2pass vbr anyway, making the overall process much slower.
Btw, 1pass crf is worse than 2pass crf and again, since the first pass of 2pass is quite fast, 2pass crf would be almost always preferable (only applies to libvpx).

>>102345055
when I'm watching a video in windowed mode the video looks blurry, is there a way to keep the video sharp when mpv is not in fullscreen?

>>102391601

dscale=bilinear
correct-downscaling=no

https://github.com/mpv-player/mpv/issues/8254

>>102391895
akemi is right

>>102391983
akemi is right on 99% of things (the other 1% for buying Apple)

https://github.com/mpv-player/mpv/issues/13082

>>102392046
>this is an insult to those who also contributed massively to this project
yikes

>>102391842
>>102389159
>>102389837

Where do you put all that stuff? command prompts?

>>102392108
mpv.conf

>>102392108
up my ass where it feels good

im running linux and use mpv to watching youtube. SOMETIMES if i close a video then i can't open a new one unless i do this:

pkill -9 mpv

does anyone else experience this?

File: Config missing.png (9 KB, 189x264)

9 KB PNG

>>102392114
would you believe that my version of MPV doesn't have it? in fact i tried to download the newest one and didn't had it.

>>102393424
install gentoo

dynaudnorm or loudnorm?

>>102393680
replaygain

>>102393424
would you believe you can make files yourself?

https://github.com/mpv-player/mpv/pull/14778#issuecomment-2351398204 wife

>>102394519
kill

>>102347957
VLC is more bad in playing videos file of mkv can't even remember the subtitles files don't have a thumbnail hover don't have a dark mode for windows for these many years vlc don't have dark mode in windows VLC can't have heavy 4k files even having better processor.
Mpv video player can do all this things so easily without breaking video files

File: 1706185937855000.png (33 KB, 552x139)

33 KB PNG

is there an option to disable this popup when changing system volume on Windows while mpv is playing media? it's very annoying

something like #hardware-media-key-handling flag in Chromium or media.hardwaremediakeys.enabled option in Firefox

let's do something different:
bicubic_fast or spline16 as a fast scaler choice?
note: i like sharpness.

>>102395687
it doesn't get blurrier than bicubic_fast
creative bump

>>102395687
I have bilinear in my lowresource profile, scale=lanczos/cscale=spline16 in my slightly less lowresource profile.

>>102395047
media-controls=no

>>102395752
yes. my question was more about whether spline16 is particularly bad at anything that it should be avoided.

>>102360226
What's the point of making a fork for just one extremely specific case?

>>102396395
It's useful for them and ffmpeg will never accept a patch for that.

>>102395687
>>102396157
I don't think Spline16 is particularly fast?
>whether spline16 is particularly bad at anything that it should be avoided
It's almost the exact same as Lanczos2 and Catmull-Rom.

Is this the yt-dlp and FFmpeg thread too?

Should I use mbedtls instead of openssl or gnutls?

>>102396854
it's a slow, limited TLS implementation that breaks on every second minor update because the maintainers don't even now how compatibility is spelled

Can MPV play audio files (.wav and .mp3) while displaying .vtt or .srt subtitle files?
Currently use MPC-HC but it doesn't do it so I'm trying to find a good alternative player.

>>102397256
mpv --idle --force-window [some audio]

>>102396854
>>102396980
is it possible to build ffmpeg with libressl nowadays?

>>102388348
>trust occams razor and follow
>not crf
Shitter doesn't know anything kek

>>102389159
So you are just shit posting now, concession accepted, hermite wins.

>>102398516
can you disprove it?

>>102391842
>

correct-downscaling=no

igv?
Anyways you are braindead.

>>102398495
>>102398516
>>102398543
mentally iill samefag

>>102398543
try it yourself instead of spouting nonsense, faggot. excellent sharpness and no ringing.

>>102397724
probably

new thread doko

>>102398753
the openbsd ffmpeg port is still on 4.4