What exactly is 2.4 or 2.2 pure power curve gamma? Is when it looks like pic related, right?
When the gamma response is flat at 2.4 across all the luminance output, right?
yes and no. in a perfect world with perfect display devices, yes. in reality, no
[...]
In reality, with a pure power gamma curve, true gamma approaches 0 as you near the limits of the display's static contrast ratio
https://www.desmos.com/calculator/zbz34fm1ql
I am asking because I calibrated my TV using HCFR and a rpi pattern generator. I used "2.4 power law" on HCFR to calibrate the gray scale because that's what AVS forum boomers seem to recommend nowadays. So I was thinking if I should set mpv to target-trc=gamma2.4.
I already set:
target-prim=bt.709
target-peak=120
target-contrast=4290
I would recommend calibrating to BT.1886 response, not pure power gamma response. Pure power curves tend to crush blacks at the low end, e.g. seen here (curves tuned to your exact contrast)
Except that most "applications" aren't exactly targeting sRGB, as 99% of the devs don't even know what sRGB is.
Even when sRGB is the expected EOTF it's likely that the dev was using a 2.2 display. Few companies give a shit about colour accuracy and the most prominent one, Apple, isn't calibrating to sRGB...
According to the Light Illusion (creators of PGenerator and other cal software) guy, BT.1886 causes lifted shadows:
https://www.avsforum.com/threads/who-prefers-a-higher-gamma-than-2-2.1758185/post-28991073
And according to these boomers the ITU is using 2.4 since 2012:
?si=ME7GB6rcLqCW0xfC&t=1698
btw I have the TV in a complexity dark room (no windows).
[...]
[...]
[...]
I am asking because I calibrated my TV using HCFR and a rpi pattern generator. I used "2.4 power law" on HCFR to calibrate the gray scale because that's what AVS forum boomers seem to recommend nowadays. So I was thinking if I should set mpv to target-trc=gamma2.4.
I already set:
target-prim=bt.709
target-peak=120
target-contrast=4290
But I am having doubts about the TRC setting.
I think if the display is HW calibrated you don't need to specify any of those parameter on the conf file, I mean the display is calibrated, just let mpv send the video as is.
But that's just my assumption, correct me if am wrong.
3 weeks ago
Anonymous
What about HDR to SDR tone-mapping, those parameters are probably useful for that.
lifted shadows is a feature, not a bug, you dolt
look at the black crush in
I would recommend calibrating to BT.1886 response, not pure power gamma response. Pure power curves tend to crush blacks at the low end, e.g. seen here (curves tuned to your exact contrast)
besides, if you're only worrying about color accuracy in mpv, then you should realize that will perform black point compensation one way or the other, your only question here should be "which curve gives me the best numerical precision"
you could calibrate to gamma 5.0 if you want to, mpv wouldn't give a shit, because the function is inverted inside the display code anyway. It will just lead to more noticeable banding/dithering noise near black
3 weeks ago
Anonymous
hmm I kind of get it.
I will recalibrate to BT.1886, I assume mpv will adjust the curve with the target peak of 120nits and the contrast of 4290(or whatever result I get once I recalibrate), right?
Calibrate the display with HCFR to BT.1886, calibrate it again with DisplayCAL but leave everything to "As measured", load the resulting ICC profile into mpv.
yes and no. in a perfect world with perfect display devices, yes. in reality, no
In reality, with a pure power gamma curve, true gamma approaches 0 as you near the limits of the display's static contrast ratio
https://www.desmos.com/calculator/zbz34fm1ql
[...]
[...]
[...]
I am asking because I calibrated my TV using HCFR and a rpi pattern generator. I used "2.4 power law" on HCFR to calibrate the gray scale because that's what AVS forum boomers seem to recommend nowadays. So I was thinking if I should set mpv to target-trc=gamma2.4.
I already set:
target-prim=bt.709
target-peak=120
target-contrast=4290
But I am having doubts about the TRC setting.
added also BT.1886 and sRGB-style curves for shits and giggles
https://www.desmos.com/calculator/hk1vdj6fov
Why isn't AV1 support enable yet? According to Mozilla is already supported on all browsers:
https://developer.mozilla.org/en-US/docs/Web/Media/Formats/Containers#webm
Is it really because of phonefags? They can't play 12-bit VP9 either so I don't see the issue of enabling another incompatible format.
just the other day i was gonna watch Drive (2011), 2 minutes in of not being able to fix the subtitles I launched it in VLC and everything worked OOTB.
nah, sadly not, haasn was too lazy to implement that.
He didn't even bother making a manual or wiki for it.. he just links to the source:
https://code.videolan.org/videolan/libplacebo/-/blob/master/src/options.c#L877
Why does this fucking shit does not work. I have a scripts folder and the skin file in there wirks finez but trackselect.lua fucking does nothing. trackselect.onf is in the script-opts folder in the same location. The GitHub repository is gone. I just want for it to not select the SDH subs which for some reason is before the normal subs.
you're gonna want keep-open=always and keep-open-pause=no in your mpv.conf, then map the media key to seek to the beginning in input.conf.
thanks dude, works like a charm.
But if you map the media key, when you pause, won't it also seek for the beginning?
you can also use an auto profile if you want to use whatever binding you have set for pause to restart playback
keep-open
[replay]
profile-cond=eof_reached
profile-restore=copy
loop
why the fuck is there a massive repo called mpv-config? how do people make this shit so complicated
Lmao some of the comments are longer than what's in the manual. And it seems most of the options are commented out anyway?
[HQ]
scale=ewa_hanning
cscale=bilinear
dscale=catmull_rom
scale-antiring=0.8
dscale-antiring=0.8
linear-upscaling=no
sigmoid-upscaling=yes
correct-downscaling=yes
linear-downscaling=no
deband=no
>enough
Using hungy resource shaders to upscale chroma plane is a meme, doing it with built in scalers is not unless you're barely scaling x1.5 or some shit like that. If not, get yourself and optometrist.
Krig is great actually especially with a luma shader. I used to be anti-Krig but its actually pretty cool.
3 weeks ago
Anonymous
Well, I agreed with Krig being the only chroma shader who matters. Still using all that computing power to upscale chroma is nonsense (even given the case when the source has hypersaturated colours).
3 weeks ago
Anonymous
It really depends, the only other built-in scaler I would use for chroma would probably be one of the ewa_lanczos variants, but they're about as computationally "heavy" as KrigBilateral for me. Maybe a non-ringing scaler such as blackman sinc is unironically the way to go if you're worried that much about wasting computing power, but don't want to use something ugly like bilinear.
3 weeks ago
Anonymous
I upscale sometimes 480p DVDs to 2160p. Using Krig there gets me the same frames like for example using ravu-lite-ar-r3 twice. For 720p to 1080p-1440p may still be usable.
what the fuck, just get a better display if you're genuinely forced into a position where you need to use your phone to watch stuff, this guy must have a washed out LCD display from the 00s or something
Wouldn't that require a shitload of bandwidth? Or you'd just be compressing it again, at which point it really doesn't matter if you used your meme shader or not, you're fucking retarded then anyway.
I've always felt like NTSC was good enough for everything except text editing. Everything above that for film or whatever is just a waste of bandwidth IMO.
Container and video dimension mismatch? mpv should probably be using whatever mediainfo is to obtain the information.
I wonder if this is a regression of if it has always been like this.
>What's SIMD
Okay, imagine you're a monkey and you have a bunch of colorful balls in front of you. Each ball represents a number or a piece of information. Now, let's say you want to do the same thing to all those balls at once, like maybe you want to paint them all red.
SIMD stands for "Single Instruction, Multiple Data." It's like having a special magic brush that lets you paint all the balls at the same time. Instead of painting them one by one, which would take a long time, you can use this magic brush to paint them all with a single stroke. It's super fast and efficient!
In computer terms, SIMD is a way for your computer's processor to perform the same operation on multiple pieces of data simultaneously. Just like our magic brush helps you change the color of many balls at once, SIMD helps the computer process lots of data quickly, making tasks like graphics, video processing, and other calculations much faster and more efficient. So, it's like a superpower for your computer!
Each core in your computer has a large number of ALUs in it. Pipelining, hyperthreading, and even out-of-order execution does not really take full advantage of them. SIMD instructions allow you to explicitly fill fat registers with lots of data and operate on them all in one cycle. If your CPU is at 100% load but its frequency is really high, it's not actually doing much. Might even be just waiting on memory half the time. If it's under 100% load and the frequency is lower, that usually means SIMD is bringing it to its knees in terms of power or thermals, which is a good thing. AVX512 is kinda broken because it's all fragmented and has retarded power draw (yes even on Zen 4) causing downclocks, and invokes a much larger latency penality, but AVX2 has about 90% adoption these days and still manages to make each core about 8x "faster" under ideal conditions.
don't have pic but Lynne looks irl like a man with long hair and chubby belly, and the typical tranny strained "trying too hard" voice
fat in all the wrong places, male voice, male bone structure, etc.
>it just werkz
Whats up with this shit?
mpv.com --no-config --vo=gpu-next "[Beatrice-Raws] Tenki no Ko [BDRip 3840x2160 HEVC HDR DTSHD].mkv >ib4 shit release/video/movie/file
doesn't matter should still play fine..
in VLC it does..
>mpv doesn't show hwdec on stats
It has, for years now
3 weeks ago
Anonymous
>using display-resample with [MTBB] meme subs
you are a brave man
3 weeks ago
Anonymous
vo=gpu also sharts the bed: >mpv --no-config --vo=gpu --gpu-api=d3d11 --window-scale=0.4 "[Beatrice-Raws] Tenki no Ko [BDRip 3840x2160 HEVC HDR DTSHD].mkv"
Never have dropped frames, toaster-kun.
3 weeks ago
Anonymous
To add more, I get some ffmpeg errors:
[ffmpeg/video] hevc: CABAC_MAX_BIN : 7
[ffmpeg/video] hevc: The cu_qp_delta -1094995529 is outside the valid range [-32, 31].
AV: 00:00:00 / 01:52:20 (0%) A-V: 0.000
[ffmpeg/video] hevc: CABAC_MAX_BIN : 7
[ffmpeg/video] hevc: The cu_qp_delta 1094995529 is outside the valid range [-32, 31].
AV: 00:00:00 / 01:52:20 (0%) A-V: 0.000
[ffmpeg/video] hevc: The cu_qp_delta -53 is outside the valid range [-32, 31].
AV: 00:00:00 / 01:52:20 (0%) A-V: 0.000
[ffmpeg/video] hevc: The cu_qp_delta -40 is outside the valid range [-32, 31].
AV: 00:00:00 / 01:52:20 (0%) A-V: 0.000
[ffmpeg/video] hevc: The cu_qp_delta -35 is outside the valid range [-32, 31].
AV: 00:00:00 / 01:52:20 (0%) A-V: 0.000
[ffmpeg/video] hevc: CABAC_MAX_BIN : 7
[ffmpeg/video] hevc: The cu_qp_delta 1094995529 is outside the valid range [-32, 31].
AV: 00:00:00 / 01:52:20 (0%) A-V: 0.000
[ffmpeg/video] hevc: The cu_qp_delta -60 is outside the valid range [-32, 31].
AV: 00:00:01 / 01:52:20 (0%) A-V: 0.000
[ffmpeg/video] hevc: The cu_qp_delta 86 is outside the valid range [-32, 31].
AV: 00:00:01 / 01:52:20 (0%) A-V: 0.000
[ffmpeg/video] hevc: The cu_qp_delta 36 is outside the valid range [-32, 31].
3 weeks ago
Anonymous
>Never have dropped frames, toaster-kun.
5800X+6950XT. It's not uncommon to get like 100+ delayed/mistimed frames, never said anything about dropping them.
3 weeks ago
Anonymous
>delayed/mistimed
If you aint where you are supposed to be when your time to be there has come, you getting dropped,.
vo=gpu also sharts the bed: >mpv --no-config --vo=gpu --gpu-api=d3d11 --window-scale=0.4 "[Beatrice-Raws] Tenki no Ko [BDRip 3840x2160 HEVC HDR DTSHD].mkv"
Never have dropped frames, toaster-kun.
>pictures show obvious decoding errors
yeah gee I wonder why changing something that happens after decoding won't affect the output
It's probably a fallout of the bug that it turns out vsync display interval was being trucated. dmg fixed that by changing to double, but there are possibly lots of other places that could benefit from increased precision.
Most likely that PR is going to lead to bugs because of some variable somewhere that wasn't updated to use the scaling ratio. I also don't really see the benefit of moving vsync base from micros to nanos, the display clock likely isn't even that high res.
>What changes make you feel this way
The one that was just posted changing to nanos everywhere. I'll bet 1 mpv bug report that it will cause an issue within display sync
I feel this is OK since it's not _too_ much bloat (it re-uses existing mechanism), but what's Kasper's master plan with this? Why has he been spending 3 months on cropping related things. Is cropping your video the new meta?
>This script uses the lavfi cropdetect filter to automatically >insert a crop filter with appropriate parameters for the currently playing video
What's the difference between this and just zooming/panning the video? Doesn't it achieve the same effect?
I have no idea, I usually only grab autoload.lua from the repo. But is there any reason why you would want to get rid of blackbars anyways? The only application I can think of is "widescreen" 4:3 DVDs to get rid of the blackbars on the side.
No I get that. But if you can manually pan/zoom, what benefits does the new --video-crop option give? It seems like anything you can do with one you can do with the other?
But video pan/scale/zoom works perfectly with hwdec as well. I'm using it right now.
3 weeks ago
Anonymous
Like I said, I have no idea. But I'm PRETTY sure the main purpose of cropping is for anamorphic content, assuming the encoder hasn't already done their own cropping.
>you might have reencoded
I didn't, I cut it with lossless cut
The clip still doesn't work for me.
3 weeks ago
Anonymous
not him, but possibly different version of ffmpeg used? Can you both post version?
3 weeks ago
Anonymous
are you using ffmpeg/mpv from shinchiro's auto builder? can you download ffmpeg windows from ffmpeg's official site and try to play it with that version of ffplay?
3 weeks ago
Anonymous
>play it with that version of ffplay?
See
ffplay also shits itself in the exact same way (green screen), cmd output:
https://pastebin.com/raw/z6ejr6PD
>ffplay version 2023-09-04-git-f8503b4c33-full_build-www.gyan.dev Copyright (c)
3 weeks ago
Anonymous
yeah I donno what's up, it works for me on linux with the exact same version of ffmpeg
3 weeks ago
Anonymous
looking at the lavc code, it seems to take different code paths specifically for vaapi/vulkan/vdpau/nvdec, so can you try nvdec or vulkan hwdec and see if it fixes it?
3 weeks ago
Anonymous
Tried it with another shitty release of the movie:
"[DMG&VCB-Studio] Tenki no Ko"
That plays but has random skips and other artifacts, very bizarre.. you think this is a ffmpeg bug?
3 weeks ago
Anonymous
For a second I thought I finally understood why this "best release" on bbt is so fucked up without any comments saying so.
But then I played it in vlc and still fucked.
3 weeks ago
Anonymous
I tried this release on the debian build (0.35.1) and the flatpak build (0.36), both with the same config seen here
While I'm pretty happy with image and scaling quality, I haven't looked into HDR, tonemapping, calibration and all that fun stuff.
vo=gpu-next
gpu-api=vulkan
profile=gpu-hq
glsl-shaders="~~/shaders/compute/ravu-zoom-ar-r3-rgb.hook"
video-sync=display-resample
interpolation
tscale=oversample
hwdec=auto
I have reached nirvana as far as playback is concerned. With amdgpu.freesync_video=1 and xrandr.lua, my display seamlessly switches to the appropriate refresh rate. You could achieve similar results on a fixed refresh rate display with custom modelines for 48Hz and whatnot, but changing refresh rate will cause the screen to blank for a couple seconds when opening and closing mpv. Idk if there are wayland compositors that allow for something similar.
[...] >skylake
So you have vulkan support, good. Try using vo=gpu-next and gpu-api=vulkan, It's a bit more efficient than OpenGL and should help performance.
No issues on either aside from the flatpak one not changing the refresh rate.
https://files.catbox.moe/m4s519.png
3 weeks ago
Anonymous
The issue is an ffmpeg regression introduced on May of this year.
3 weeks ago
Anonymous
Did that regression somehow make it into a stable release?
3 weeks ago
Anonymous
no, but the github windows builders build literally everything off git HEAD
Hey llyyr now that you are here, tell haasn to fix the Hermite parameters:
https://code.videolan.org/videolan/libplacebo/-/blob/master/src/filters.c#L478
They should be B=0 and C=0, see:
https://www.imagemagick.org/Usage/filter/#cubic_bc
you found the issue, so you should send the patch. It's just changing one number anon
3 weeks ago
Anonymous
See
>GitHub CI does
I don't have a GitHub account
3 weeks ago
Anonymous
I don't know what that means. I'm saying that all the Windows mpv builds include a ffmpeg built off git HEAD.
You can get the binary from https://github.com/shinchiro/mpv-winbuild-cmake without needing a github account
3 weeks ago
Anonymous
I don't have neither a gitlab nor a github account to submit the hermite patch.
3 weeks ago
Anonymous
just send haasn an email
3 weeks ago
Anonymous
Fix the values, commit it locally, use git format-patch -n1 to generate a patch and email it to haasn
Be sure to attach a bunch kissy faces into your e-mail as well.
3 weeks ago
Anonymous
Fix the values, commit it locally, use git format-patch -n1 to generate a patch and email it to haasn
3 weeks ago
Anonymous
3 weeks ago
Anonymous
You can always install ffmpeg through WinGet, it's much easier than installing and updating manually. I do that for all my programs on Windows, it's nice.
https://www.gyan.dev/ffmpeg/builds/
3 weeks ago
Anonymous
doesn't matter since the mpv build statically links to ffmpeg, your system ffmpeg has nothing to do with it
3 weeks ago
Anonymous
Right, I forgot that's how things tend to be done on windows. I'll have to check if mpv.net is built against a stable ffmpeg.
3 weeks ago
Anonymous
Last I checked, mpv.net just uses libmpv provided by shinchiro and zhongfly.
yeah I donno what's up, it works for me on linux with the exact same version of ffmpeg
Ok something weird is going on my system, if I print ffmpeg -version it says ffmpeg version 4.2.3
Let me check my Path shit
3 weeks ago
Anonymous
>ffmpeg version 4.2.3
lmao
3 weeks ago
Anonymous
he took the mpv 0.29 meme too literally, and decided to use an ancient ffmpeg as well.
[...]
Ok something weird is going on my system, if I print ffmpeg -version it says ffmpeg version 4.2.3
Let me check my Path shit
Unless you somehow managed to link with that version of ffmpeg, I doubt it would be the issue?Just do `mpv -v`
3 weeks ago
Anonymous
he took the mpv 0.29 meme too literally, and decided to use an ancient ffmpeg as well.
[...]
Unless you somehow managed to link with that version of ffmpeg, I doubt it would be the issue?Just do `mpv -v`
IT WAS FUCKING IMAGE MAGICK
where ffmpeg
c:program filesimagemagick-7.1.0-q16-hdriffmpeg.exe
3 weeks ago
Anonymous
This doesn't make much sense to me, I thought mpv was compiled with ffmpeg on Windows? Why would an external ffmpeg binary interfere with mpv?
3 weeks ago
Anonymous
Yep, still doesn't work on neither mpv nor ffplay:
ffplay -version
ffplay version N-111995-g5f8c113b75-20230909 Copyright (c) 2003-2023 the FFmpeg developers
built with gcc 13.1.0 (crosstool-NG 1.25.0.196_227d99d)
I'm trying to encode a source with a hard telecined field blend pattern that goes ff Bf Bf fB fB.
I realized that this pattern leaves unblended fields for frame A and C and one unblended field for both B and D.
I'm trying to reconstruct an unblended 2:3 pulldown pattern by replacing the blended fields with unblended fields.
Since I'm missing a field for B and D, I'm using QTGMC to interpolate the missing fields.
This works pretty well for removing junk frames from decimation, but it creates this vertical shake effect due to frames B and D being interpolated.
Running QTGMC with InputType=1 after IVTC and Decimation works pretty well for curbing the vertical shake, but the output looks kind of like it has motion smoothing.
Here's the deblending code of my Vapoursynth script
def interpolate_fields(clip, TFF=True):
clip = core.std.Interleave(clips=[clip, clip])
clip = core.std.DoubleWeave(clip, tff=TFF)
clip = core.std.SelectEvery(clip, cycle=2, offsets=0)
This is exactly the sort of thing when you’re 14 and dream about.
“Aww dood it wood be so kewl if I wake up in front of big tee vee and play all the vee gays I want!”
My back hurts just looking at this photo. Do you just sit down for 8 hours in your comfy recliner and code? Is it productive code or is this just like one of those sys ad type guys that thinks he needs more monitors so he can pretend to monitor resources
>Is it productive code or is this just like one of those sys ad type guys that thinks he needs more monitors so he can pretend to monitor resources
tfw I've been working the past 5 years of my life professionally using only a single monitor, all because my secondary monitor broke and I was too lazy to replace it, so I eventually realized I don't even need it
Hey. I listen through mpv a radio station (in celluloid) and the program save even 4 houres of the recording in the background/cache. How can I save it on hdd or where I can find the file cuz I can't find and an api documentation also don't help
While I'm pretty happy with image and scaling quality, I haven't looked into HDR, tonemapping, calibration and all that fun stuff.
vo=gpu-next
gpu-api=vulkan
profile=gpu-hq
glsl-shaders="~~/shaders/compute/ravu-zoom-ar-r3-rgb.hook"
video-sync=display-resample
interpolation
tscale=oversample
hwdec=auto
I have reached nirvana as far as playback is concerned. With amdgpu.freesync_video=1 and xrandr.lua, my display seamlessly switches to the appropriate refresh rate. You could achieve similar results on a fixed refresh rate display with custom modelines for 48Hz and whatnot, but changing refresh rate will cause the screen to blank for a couple seconds when opening and closing mpv. Idk if there are wayland compositors that allow for something similar.
[...]
Skylake, yes it works but GPU usage is quite high
I thought vaapi should work better, apparently not
>skylake
So you have vulkan support, good. Try using vo=gpu-next and gpu-api=vulkan, It's a bit more efficient than OpenGL and should help performance.
vo=gpu is same as not using vaapi on command line, it has much higher CPU overhead
I thought VAAPI should be used with Intel integrated graphics?
in any case I get: "Using hardware decoding (vaapi)."
???
then why OSD works, but it doesn't work when vo=vaapi is selected?
>Note that there is absolutely no reason to use this, other than compatibility. This is low quality, and has issues with OSD. We strongly recommend that you use --vo=gpu with --hwdec=vaapi instead.
The OSD does work with vo=vaapi, but breaks as soon as you resize the window or go fullscreen.
What kind of toaster are you using? Even my sandy bridge laptop can use vo=gpu with simpler shaders like FSR or CAS without dropping frames.
imagine needing a giant vertical screen to constantly have a manual of a video player open, just to watch something.
Real men just use windows media player
https://github.com/mpv-player/mpv/issues/12373 >Works as expected
Doesn't actually work as expected because behavior changed, hence the bug report. Starting to hate kasper.
Thanks to whichever anon reported it.
>Works as expected this are true native dimensions of the video frame.
it's expected behavior to violate the spec and report an incorrect resolution? those extra 8 pixels are literally garbage padding data added by the encoder and are never meant to be seen or known by the user
According to the encoder he:
"For this release, I converted Dolby Vision Profile 7 BL + EL + RPU from BD to Profile 8 BL + RPU."
Don't know if it is related to the issue.
works for me. As other anon said, must be ffmpeg. Linux with latest stable builds of all deps except libplacebo which is latest git, and mpv git.
Might try a much older build from around March.
Like the other anon said, you can try to bisect with ffplay which shouldn't take too long. Gyan keeps all of their builds here:
https://github.com/GyanD/codexffmpeg/tags
af=sofalizer="C:/mpv/sofalizer/ClubFritz6.sofa":type=time:gain=-8
http://sofacoustics.org/data/database/clubfritz/
Loads of files available, some might work better for you. It's depended on your head. ClubFritz6 is cargocult, 4 is used quite a bit as well but I think that downmixes to 44.1kHz instead of 48kHz like 6. https://www.sofaconventions.org/mediawiki/index.php/Files
type=time is slower according ffmpeg, no clue if it's better. gain=-8 I did because it clipped on some of my files originally and I have an analogue amplifier anyway.
You can also try this, though it's less sophisticated. I'm pretty sure I just modified the 5.1 for 6.1 and 7.1. Based on https://superuser.com/questions/852400/properly-downmix-5-1-to-stereo-using-ffmpeg#1410620
For 5.1:
af=lavfi=[pan=stereo|FL=0.5*FC+0.707*FL+0.707*SL+0.5*LFE|FR=0.5*FC+0.707*FR+0.707*SR+0.5*LFE]
For 7.1:
af=lavfi=[pan=stereo|FL=0.5*FC+0.707*FL+0.707*SL+0.707*BL+0.5*LFE|FR=0.5*FC+0.707*FR+0.707*SR+0.707*BR+0.5*LFE]
And in case you encounter 6.1:
af=lavfi=[pan=stereo|FL=0.5*FC+0.707*FL+0.707*SL+0.5*BC+0.5*LFE|FR=0.5*FC+0.707*FR+0.707*SR+0.5*BC+0.5*LFE]
This wouldn't include 2.1 though. Sofalizer does. There are some other things you can try. Like af=lavfi=[pan=stereo|c0=c2+0.30*c0+0.30*c4|c1=c2+0.30*c1+0.30*c5] Which somebody mixed and called "nightmode dialogue", to boost dialogue but not have too much volume during the night. Sounds like shit to me.
Some reference links. Not required.
https://www.reddit.com/r/mpv/comments/11cr5u9/is_it_possible_to_use_the_headphone_filter/
https://gist.github.com/kevinlekiller/9fd21936411d8dc5998793470c6e3d16
https://ffmpeg.org/ffmpeg-filters.html#sofalizer
You could use that script I guess, but I just have this profile.:
[multichannel]
profile-cond=(p["audio-params/channel-count"] > 2)
af=sofalizer="C:/mpv/sofalizer/ClubFritz6.sofa":type=time:gain=-8
A full month is going to be a lot of commits for ffmpeg.
You can fork shinchiro, add
GIT_REMOTE_NAME origin
GIT_TAG main
GIT_RESET ad417eb5
to ffmpeg.cmake, disable the included patch, and hope the CI builds without needing other modifications. I used a short hash to avoid spam detection, might not work with git reset.
3 weeks ago
Anonymous
>GIT_TAG main
Copy/paste mistake. ffmpeg still uses master. Just replace the PATCH_COMMAND line with those three lines and hopefully it will build.
3 weeks ago
Anonymous
Bro I am not going to compile ffmpeg, I don't watch Shinkai shit to begin with, I only downloaded the movie to check if
>it just werkz
Whats up with this shit?
mpv.com --no-config --vo=gpu-next "[Beatrice-Raws] Tenki no Ko [BDRip 3840x2160 HEVC HDR DTSHD].mkv >ib4 shit release/video/movie/file
doesn't matter should still play fine..
in VLC it does..
was posting fud or not.
You are welcome to file the bug report to ffmpeg's tracker, the test clip is right here
Here:
https://pastebin.com/FtX7AVcX
According to the encoder he:
"For this release, I converted Dolby Vision Profile 7 BL + EL + RPU from BD to Profile 8 BL + RPU."
>I am not going to compile ffmpeg
You don't, GitHub CI does. Useful if anyone else has the issue because sasukefag insists on using git for nearly every dependency.
Hey llyyr now that you are here, tell haasn to fix the Hermite parameters:
https://code.videolan.org/videolan/libplacebo/-/blob/master/src/filters.c#L478
They should be B=0 and C=0, see:
https://www.imagemagick.org/Usage/filter/#cubic_bc
>If you have media files with exactly the same codec and codec parameters you can concatenate them
Can I still concoct two files with different bitrates (but same crf settings)?
Does ffmpeg -f concat -safe 0 -i mylist.txt -c copy output.wav also work if the files I want to merge (both mkv) have one video and audio stream? Or should I use the concat protocol in that case?
I tried vo=gpu-next on my fairphone 4 with the f-droid build and couldn't play anything. The snapdragon 750g it uses does support vulkan. No idea if the ROM it came with is any different.
>it just werkz
Whats up with this shit?
mpv.com --no-config --vo=gpu-next "[Beatrice-Raws] Tenki no Ko [BDRip 3840x2160 HEVC HDR DTSHD].mkv >ib4 shit release/video/movie/file
doesn't matter should still play fine..
in VLC it does..
my original issue, can't believe llyyr actually fixed something the SIMD trans queen messed up, go figure.
that angle makss it look like the chair is toddler sized lmao
wintoddlers btfo
i doubt thats optimal sub positioning
profile=gpu-hq
Rate my config.
pretty cool as always, very nice
but have a look at mine:
>good pointing device
>nontiling WM
not great, not terrible
What exactly is 2.4 or 2.2 pure power curve gamma? Is when it looks like pic related, right?
When the gamma response is flat at 2.4 across all the luminance output, right?
>power curve
>flat
I can't into mathemagics 🙁
https://theailearner.com/2019/01/26/power-law-gamma-transformations/
I am asking because I calibrated my TV using HCFR and a rpi pattern generator. I used "2.4 power law" on HCFR to calibrate the gray scale because that's what AVS forum boomers seem to recommend nowadays. So I was thinking if I should set mpv to target-trc=gamma2.4.
I already set:
target-prim=bt.709
target-peak=120
target-contrast=4290
But I am having doubts about the TRC setting.
I would recommend calibrating to BT.1886 response, not pure power gamma response. Pure power curves tend to crush blacks at the low end, e.g. seen here (curves tuned to your exact contrast)
alternatively, you could calibrate to an sRGB contrast and get "accurate" graphics in applications other than mpv, then set target-trc=srgb
Except that most "applications" aren't exactly targeting sRGB, as 99% of the devs don't even know what sRGB is.
Even when sRGB is the expected EOTF it's likely that the dev was using a 2.2 display. Few companies give a shit about colour accuracy and the most prominent one, Apple, isn't calibrating to sRGB...
According to the Light Illusion (creators of PGenerator and other cal software) guy, BT.1886 causes lifted shadows:
https://www.avsforum.com/threads/who-prefers-a-higher-gamma-than-2-2.1758185/post-28991073
And according to these boomers the ITU is using 2.4 since 2012:
?si=ME7GB6rcLqCW0xfC&t=1698
btw I have the TV in a complexity dark room (no windows).
>btw I have the TV in a complexity dark room (no windows).
Your mom's basement?
I think if the display is HW calibrated you don't need to specify any of those parameter on the conf file, I mean the display is calibrated, just let mpv send the video as is.
But that's just my assumption, correct me if am wrong.
What about HDR to SDR tone-mapping, those parameters are probably useful for that.
lifted shadows is a feature, not a bug, you dolt
look at the black crush in
besides, if you're only worrying about color accuracy in mpv, then you should realize that will perform black point compensation one way or the other, your only question here should be "which curve gives me the best numerical precision"
you could calibrate to gamma 5.0 if you want to, mpv wouldn't give a shit, because the function is inverted inside the display code anyway. It will just lead to more noticeable banding/dithering noise near black
hmm I kind of get it.
I will recalibrate to BT.1886, I assume mpv will adjust the curve with the target peak of 120nits and the contrast of 4290(or whatever result I get once I recalibrate), right?
It should but I don't really know.
Calibrate the display with HCFR to BT.1886, calibrate it again with DisplayCAL but leave everything to "As measured", load the resulting ICC profile into mpv.
yes and no. in a perfect world with perfect display devices, yes. in reality, no
In reality, with a pure power gamma curve, true gamma approaches 0 as you near the limits of the display's static contrast ratio
https://www.desmos.com/calculator/zbz34fm1ql
added also BT.1886 and sRGB-style curves for shits and giggles
https://www.desmos.com/calculator/hk1vdj6fov
Now get rid of the huge 4k TV's and replace them with a Valve Index 2.
>Now get rid of the huge 4k TV's and replace them with a projector and a madVR Envy
ftfy
Why isn't AV1 support enable yet? According to Mozilla is already supported on all browsers:
https://developer.mozilla.org/en-US/docs/Web/Media/Formats/Containers#webm
Is it really because of phonefags? They can't play 12-bit VP9 either so I don't see the issue of enabling another incompatible format.
I like mpv but what should I use to listen to my albums? Clementine? I need to have a fast search and comfy UI. Using KDE Plasma btw.
foobar
I use that on windows currently but I am migrating to Linux wholeheartedly
>wholeheartedly
Anon, if someone is pointing a gun to your head, please give us a sign.
Strawberry
Artoriuz updated the charts with performance
--d3d11-sync-interval=0 does that shit actually do smth or its placebo?
just the other day i was gonna watch Drive (2011), 2 minutes in of not being able to fix the subtitles I launched it in VLC and everything worked OOTB.
Is there a way to print all the libplacebo-opts parameters? I tried --libplacebo-opts=help and --libplacebo-opts-append=help and nothing printed out.
nah, sadly not, haasn was too lazy to implement that.
He didn't even bother making a manual or wiki for it.. he just links to the source:
https://code.videolan.org/videolan/libplacebo/-/blob/master/src/options.c#L877
>mid size speaker on floor without risers
>sub in corner
can't consoom a brain i guess
It's not a hard rule, but sub in a corner is often a good spot for many rooms.
he probably just isn't an audiotard
Why does this fucking shit does not work. I have a scripts folder and the skin file in there wirks finez but trackselect.lua fucking does nothing. trackselect.onf is in the script-opts folder in the same location. The GitHub repository is gone. I just want for it to not select the SDH subs which for some reason is before the normal subs.
There's Indian code in mpv?
ded thread
ded project
I have never felt jealous of someone's setup until now.
Damn.
meme setup thread?
im curious how you watch anything like this
It's a standing setup.
?si=LISOUOQ-EpQrvvys
The TV is way too high up, rip viewing angles and necks.
yo how can i make mpv start the video again by pressing the play button just like mpc?
you're gonna want keep-open=always and keep-open-pause=no in your mpv.conf, then map the media key to seek to the beginning in input.conf.
thanks dude, works like a charm.
But if you map the media key, when you pause, won't it also seek for the beginning?
you can also use an auto profile if you want to use whatever binding you have set for pause to restart playback
keep-open
[replay]
profile-cond=eof_reached
profile-restore=copy
loop
It's happening
https://github.com/mpv-player/mpv/pull/12368
>dokur0chan was the catrom shitposter
catrom isn't a shitpost
Unlikely that it will get merged considering the fact that section of the manual hasn't been updated ever since wm4 wrote it.
just make haasn take the catrom pill
Why are chinc mpv dev are butthurted so easily https://github.com/dyphire/mpv-config/discussions/33#discussioncomment-6956347
>dude just translate your outdated cargocult config for me
he's not wrong though
why the fuck is there a massive repo called mpv-config? how do people make this shit so complicated
extremely high doses of cargoculting
...did you really need to ask that? of course they do
Do people really copy confs from the internet?
Lmao some of the comments are longer than what's in the manual. And it seems most of the options are commented out anyway?
[HQ]
scale=ewa_hanning
cscale=bilinear
dscale=catmull_rom
scale-antiring=0.8
dscale-antiring=0.8
linear-upscaling=no
sigmoid-upscaling=yes
correct-downscaling=yes
linear-downscaling=no
deband=no
New cargocult just dropped.
why do you need antiringing with ewa_hanning? its already soft as fuck
>cscale=bilinear
you're plain retard (or have an iGPU)
its enough actually
scale=spline36
cscale=bilinear
dscale=catmull_rom
sigmoid-upscaling=yes
correct-downscaling=yes
linear-downscaling=yes
>enough
Using hungy resource shaders to upscale chroma plane is a meme, doing it with built in scalers is not unless you're barely scaling x1.5 or some shit like that. If not, get yourself and optometrist.
Krig is great actually especially with a luma shader. I used to be anti-Krig but its actually pretty cool.
Well, I agreed with Krig being the only chroma shader who matters. Still using all that computing power to upscale chroma is nonsense (even given the case when the source has hypersaturated colours).
It really depends, the only other built-in scaler I would use for chroma would probably be one of the ewa_lanczos variants, but they're about as computationally "heavy" as KrigBilateral for me. Maybe a non-ringing scaler such as blackman sinc is unironically the way to go if you're worried that much about wasting computing power, but don't want to use something ugly like bilinear.
I upscale sometimes 480p DVDs to 2160p. Using Krig there gets me the same frames like for example using ravu-lite-ar-r3 twice. For 720p to 1080p-1440p may still be usable.
I assume they're using krigbilateral or something together with that profile, but I'm not untangling what profile uses what.
how big are those screens?
if that thing reclined into 180 degree position you'd literally never have to leave save for work/shitting
just take a shit in your chair and clean it up afterwards
Why are redditors like this?
On a 7inch phone screen you can't tell the difference between 720p and 1080p anyway.
what the fuck, just get a better display if you're genuinely forced into a position where you need to use your phone to watch stuff, this guy must have a washed out LCD display from the 00s or something
Wouldn't that require a shitload of bandwidth? Or you'd just be compressing it again, at which point it really doesn't matter if you used your meme shader or not, you're fucking retarded then anyway.
who the fuck uses such a heavy upscaler on a smartphone, do people really be this retard in real life?
I've always felt like NTSC was good enough for everything except text editing. Everything above that for film or whatever is just a waste of bandwidth IMO.
animejanai + RIFE
It appears BDs and streams use mod16 1920x1088 but mediainfo reports 1920x1080. This is rather annoying to see on 90% of content.
Container and video dimension mismatch? mpv should probably be using whatever mediainfo is to obtain the information.
I wonder if this is a regression of if it has always been like this.
Can anyone tell what lua script this chad had open?
What's SIMD
>What's SIMD
Okay, imagine you're a monkey and you have a bunch of colorful balls in front of you. Each ball represents a number or a piece of information. Now, let's say you want to do the same thing to all those balls at once, like maybe you want to paint them all red.
SIMD stands for "Single Instruction, Multiple Data." It's like having a special magic brush that lets you paint all the balls at the same time. Instead of painting them one by one, which would take a long time, you can use this magic brush to paint them all with a single stroke. It's super fast and efficient!
In computer terms, SIMD is a way for your computer's processor to perform the same operation on multiple pieces of data simultaneously. Just like our magic brush helps you change the color of many balls at once, SIMD helps the computer process lots of data quickly, making tasks like graphics, video processing, and other calculations much faster and more efficient. So, it's like a superpower for your computer!
Thanks for the explanation but
made more sense to me.
it's where you sacrifice your soul to satan to make your video player faster
Single Instruction Multiple Data - subset of cpu instructions that process a lot of vectorized data at once
Each core in your computer has a large number of ALUs in it. Pipelining, hyperthreading, and even out-of-order execution does not really take full advantage of them. SIMD instructions allow you to explicitly fill fat registers with lots of data and operate on them all in one cycle. If your CPU is at 100% load but its frequency is really high, it's not actually doing much. Might even be just waiting on memory half the time. If it's under 100% load and the frequency is lower, that usually means SIMD is bringing it to its knees in terms of power or thermals, which is a good thing. AVX512 is kinda broken because it's all fragmented and has retarded power draw (yes even on Zen 4) causing downclocks, and invokes a much larger latency penality, but AVX2 has about 90% adoption these days and still manages to make each core about 8x "faster" under ideal conditions.
will this make my gmail arrive faster? Because I heard I need fast interwebs and good processor for fast gmail
No.
yes sir
AVX512 is stupid because of dennard scaling
SuckIn My Dick
Ver2 is called Suck In More Dick
I love her so much...
>her
haasn is cuter
>her
Lynne 100% doesn't pass, have you seen that abomination IRL?
show pic
so, use vlc?
don't have pic but Lynne looks irl like a man with long hair and chubby belly, and the typical tranny strained "trying too hard" voice
fat in all the wrong places, male voice, male bone structure, etc.
so like the average linux user and or ffmpeg/mpv dev?
Just tell me, are there vulns?
Is it safe to stream?
wat?
>it just werkz
Whats up with this shit?
mpv.com --no-config --vo=gpu-next "[Beatrice-Raws] Tenki no Ko [BDRip 3840x2160 HEVC HDR DTSHD].mkv
>ib4 shit release/video/movie/file
doesn't matter should still play fine..
in VLC it does..
try mpv 0.29
I'm on git master..
that's your problem right there
what is this gui/menu
looks like modernx
modernx
>using outdated software
>using broken software
does modernx show you cached range?
i dunno i just know what it is, uosc is the only good GUI
>i dunno i just know what it is
I've seen pictures of it, so I know what the custom osc is. I don't know every little detail about it
yes
does --gpu-api=vulkan fix it?
no issue = fake news = FUD
you know the rules
Well, he is not posting fud
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
so what broken hardware decoding are you using
>so what broken hardware decoding
Do you see any hwdec on that image? you blind retard
mpv doesn't show hwdec on the stats screen. Or did some recent change add that?
>mpv doesn't show hwdec on stats
It has, for years now
>using display-resample with [MTBB] meme subs
you are a brave man
vo=gpu also sharts the bed:
>mpv --no-config --vo=gpu --gpu-api=d3d11 --window-scale=0.4 "[Beatrice-Raws] Tenki no Ko [BDRip 3840x2160 HEVC HDR DTSHD].mkv"
Never have dropped frames, toaster-kun.
To add more, I get some ffmpeg errors:
[ffmpeg/video] hevc: CABAC_MAX_BIN : 7
[ffmpeg/video] hevc: The cu_qp_delta -1094995529 is outside the valid range [-32, 31].
AV: 00:00:00 / 01:52:20 (0%) A-V: 0.000
[ffmpeg/video] hevc: CABAC_MAX_BIN : 7
[ffmpeg/video] hevc: The cu_qp_delta 1094995529 is outside the valid range [-32, 31].
AV: 00:00:00 / 01:52:20 (0%) A-V: 0.000
[ffmpeg/video] hevc: The cu_qp_delta -53 is outside the valid range [-32, 31].
AV: 00:00:00 / 01:52:20 (0%) A-V: 0.000
[ffmpeg/video] hevc: The cu_qp_delta -40 is outside the valid range [-32, 31].
AV: 00:00:00 / 01:52:20 (0%) A-V: 0.000
[ffmpeg/video] hevc: The cu_qp_delta -35 is outside the valid range [-32, 31].
AV: 00:00:00 / 01:52:20 (0%) A-V: 0.000
[ffmpeg/video] hevc: CABAC_MAX_BIN : 7
[ffmpeg/video] hevc: The cu_qp_delta 1094995529 is outside the valid range [-32, 31].
AV: 00:00:00 / 01:52:20 (0%) A-V: 0.000
[ffmpeg/video] hevc: The cu_qp_delta -60 is outside the valid range [-32, 31].
AV: 00:00:01 / 01:52:20 (0%) A-V: 0.000
[ffmpeg/video] hevc: The cu_qp_delta 86 is outside the valid range [-32, 31].
AV: 00:00:01 / 01:52:20 (0%) A-V: 0.000
[ffmpeg/video] hevc: The cu_qp_delta 36 is outside the valid range [-32, 31].
>Never have dropped frames, toaster-kun.
5800X+6950XT. It's not uncommon to get like 100+ delayed/mistimed frames, never said anything about dropping them.
>delayed/mistimed
If you aint where you are supposed to be when your time to be there has come, you getting dropped,.
broski... how long has it been since you use mpv?
>did some recent change add that?
Yeah, it was added in 95b6bc, 6 years ago
This + last weeks mathematical snaffu, what is going to happen next? Prison?
>pictures show obvious decoding errors
yeah gee I wonder why changing something that happens after decoding won't affect the output
????
Those aren't hardware decoding errors, I can repro them on Windows without hwdec. Something is just royally fucked on shinchiro builds.
I didn't mention hardware decoding
<KEY> af toggle lavfi=[loudnorm]
There is too much difference in volume in the dialogues of the same scene. How to fix this?
I hope this picture is old when deckard comes out this year
https://github.com/mpv-player/mpv/pull/12371
what are the larger implications of this
It's probably a fallout of the bug that it turns out vsync display interval was being trucated. dmg fixed that by changing to double, but there are possibly lots of other places that could benefit from increased precision.
Most likely that PR is going to lead to bugs because of some variable somewhere that wasn't updated to use the scaling ratio. I also don't really see the benefit of moving vsync base from micros to nanos, the display clock likely isn't even that high res.
my setup is like this but instead of the TV on the left i have my thinkpad in a dock lol
4k @ 60 hz / 2560x1440 @ 120hz with T470s
Wow, Kasper has been busy <3
Busy bricking mpv, that is.
No need to be bitter
vsync code is scary, im expecting a brickage either way
busy adding bloat, that is
Explain how one thing that has been merged is a negative
you don't need to crop your videos
That's all you have to say? Very anticlimactic.
What changes make you feel this way? The recent talk didn't raise any alarms.
Its OK I trust kasper to do the right thing, MPC-HC was epic after all
>What changes make you feel this way
The one that was just posted changing to nanos everywhere. I'll bet 1 mpv bug report that it will cause an issue within display sync
Sirs, can you post the URL, sirs?
I feel this is OK since it's not _too_ much bloat (it re-uses existing mechanism), but what's Kasper's master plan with this? Why has he been spending 3 months on cropping related things. Is cropping your video the new meta?
You WILL crop the video
I don't even know what autocrop.lua does, I've never used it...
>This script uses the lavfi cropdetect filter to automatically
>insert a crop filter with appropriate parameters for the currently playing video
What's the difference between this and just zooming/panning the video? Doesn't it achieve the same effect?
I have no idea, I usually only grab autoload.lua from the repo. But is there any reason why you would want to get rid of blackbars anyways? The only application I can think of is "widescreen" 4:3 DVDs to get rid of the blackbars on the side.
Do you really not know the difference between automatic and manual?
No I get that. But if you can manually pan/zoom, what benefits does the new --video-crop option give? It seems like anything you can do with one you can do with the other?
Not him, but from my understanding the new video-crop works with hwdec, that's the main difference.
But video pan/scale/zoom works perfectly with hwdec as well. I'm using it right now.
Like I said, I have no idea. But I'm PRETTY sure the main purpose of cropping is for anamorphic content, assuming the encoder hasn't already done their own cropping.
jfc
>that fucking size
>still can't managed to have windows on top of windows
carpal tunnel
ffplay also shits itself in the exact same way (green screen), cmd output:
https://pastebin.com/raw/z6ejr6PD
And the file works just fine with mpc-hc plus madvr btw, literally Joever
Well, just knowing it works in VLC was good enough. No need to throw the file into garbage bins to see if it gets dirty again.
How the fuck does it break on ffplay and mpv but work fine in VLC and mpc-hc?
ffmpeg bug, post a small sample cuz i'm not downloading the whole thing
Give me a second
I use mpv, madvr is hot garbage but I have it installed for testing.
Here:
https://pastebin.com/FtX7AVcX
According to the encoder he:
"For this release, I converted Dolby Vision Profile 7 BL + EL + RPU from BD to Profile 8 BL + RPU."
Don't know if it is related to the issue.
it works for me, you might have reencoded the video instead of clipping it
but I do see
[ffmpeg/video] hevc: Overread in VUI, retrying from timing information...
[ffmpeg/video] hevc: extended_precision_processing_flag not yet implemented
[ffmpeg/video] hevc: high_precision_offsets_enabled_flag not yet implemented
[ffmpeg/video] hevc: cabac_bypass_alignment_enabled_flag not yet implemented
[ffmpeg/video] hevc: Overread in VUI, retrying from timing information...
[ffmpeg/video] hevc: extended_precision_processing_flag not yet implemented
[ffmpeg/video] hevc: high_precision_offsets_enabled_flag not yet implemented
[ffmpeg/video] hevc: cabac_bypass_alignment_enabled_flag not yet implemented
>you might have reencoded
I didn't, I cut it with lossless cut
The clip still doesn't work for me.
not him, but possibly different version of ffmpeg used? Can you both post version?
are you using ffmpeg/mpv from shinchiro's auto builder? can you download ffmpeg windows from ffmpeg's official site and try to play it with that version of ffplay?
>play it with that version of ffplay?
See
>ffplay version 2023-09-04-git-f8503b4c33-full_build-www.gyan.dev Copyright (c)
yeah I donno what's up, it works for me on linux with the exact same version of ffmpeg
looking at the lavc code, it seems to take different code paths specifically for vaapi/vulkan/vdpau/nvdec, so can you try nvdec or vulkan hwdec and see if it fixes it?
Tried it with another shitty release of the movie:
"[DMG&VCB-Studio] Tenki no Ko"
That plays but has random skips and other artifacts, very bizarre.. you think this is a ffmpeg bug?
For a second I thought I finally understood why this "best release" on bbt is so fucked up without any comments saying so.
But then I played it in vlc and still fucked.
I tried this release on the debian build (0.35.1) and the flatpak build (0.36), both with the same config seen here
No issues on either aside from the flatpak one not changing the refresh rate.
https://files.catbox.moe/m4s519.png
The issue is an ffmpeg regression introduced on May of this year.
Did that regression somehow make it into a stable release?
no, but the github windows builders build literally everything off git HEAD
you found the issue, so you should send the patch. It's just changing one number anon
See
I don't know what that means. I'm saying that all the Windows mpv builds include a ffmpeg built off git HEAD.
You can get the binary from https://github.com/shinchiro/mpv-winbuild-cmake without needing a github account
I don't have neither a gitlab nor a github account to submit the hermite patch.
just send haasn an email
Be sure to attach a bunch kissy faces into your e-mail as well.
Fix the values, commit it locally, use git format-patch -n1 to generate a patch and email it to haasn
You can always install ffmpeg through WinGet, it's much easier than installing and updating manually. I do that for all my programs on Windows, it's nice.
https://www.gyan.dev/ffmpeg/builds/
doesn't matter since the mpv build statically links to ffmpeg, your system ffmpeg has nothing to do with it
Right, I forgot that's how things tend to be done on windows. I'll have to check if mpv.net is built against a stable ffmpeg.
Last I checked, mpv.net just uses libmpv provided by shinchiro and zhongfly.
Works for me without any problems on GNU/Linux.
Ok something weird is going on my system, if I print ffmpeg -version it says ffmpeg version 4.2.3
Let me check my Path shit
>ffmpeg version 4.2.3
lmao
he took the mpv 0.29 meme too literally, and decided to use an ancient ffmpeg as well.
Unless you somehow managed to link with that version of ffmpeg, I doubt it would be the issue?Just do `mpv -v`
IT WAS FUCKING IMAGE MAGICK
where ffmpeg
c:program filesimagemagick-7.1.0-q16-hdriffmpeg.exe
This doesn't make much sense to me, I thought mpv was compiled with ffmpeg on Windows? Why would an external ffmpeg binary interfere with mpv?
Yep, still doesn't work on neither mpv nor ffplay:
ffplay -version
ffplay version N-111995-g5f8c113b75-20230909 Copyright (c) 2003-2023 the FFmpeg developers
built with gcc 13.1.0 (crosstool-NG 1.25.0.196_227d99d)
mpv --version
mpv v0.36.0-274-gbe922231 Copyright © 2000-2023 mpv/MPlayer/mplayer2 projects
built on Sep 10 2023 01:04:13
libplacebo version: v6.292.0-336-gc985725-dirty
FFmpeg version: N-111995-g5f8c113b7
FFmpeg library versions:
libavutil 58.19.100
libavcodec 60.26.100
libavformat 60.11.100
libswscale 7.3.100
libavfilter 9.11.100
libswresample 4.11.100
Thank you. It works on my Windows machine using ffplay just fine.
Yep, this is broken for me as well.
I tried it with hevc_cuvid, it partly plays then, but skips and lags a lot.
works for me with
vo=gpu-next
hwdec=d3d11va
hwdec-codecs=all
d3d11va-zero-copy=yes
but I do see this too
mpv is just pure kusoware. Did you think the madVR chad copypasta was just a meme...?
but vlc (and presumably mpc-hc) use libav too? If this is a decode or demux issue, why aren't they affected.
>dropped frames 114
jesus christ dude
its time for sizumam to return and save the Windows desktop experience
I'm trying to encode a source with a hard telecined field blend pattern that goes ff Bf Bf fB fB.
I realized that this pattern leaves unblended fields for frame A and C and one unblended field for both B and D.
I'm trying to reconstruct an unblended 2:3 pulldown pattern by replacing the blended fields with unblended fields.
Since I'm missing a field for B and D, I'm using QTGMC to interpolate the missing fields.
This works pretty well for removing junk frames from decimation, but it creates this vertical shake effect due to frames B and D being interpolated.
Running QTGMC with InputType=1 after IVTC and Decimation works pretty well for curbing the vertical shake, but the output looks kind of like it has motion smoothing.
Here's the deblending code of my Vapoursynth script
def interpolate_fields(clip, TFF=True):
clip = core.std.Interleave(clips=[clip, clip])
clip = core.std.DoubleWeave(clip, tff=TFF)
clip = core.std.SelectEvery(clip, cycle=2, offsets=0)
clip = haf.QTGMC(clip, Preset='very slow', TFF=TFF, opencl=True)
clip = core.std.SeparateFields(clip[::2], tff=TFF)
return clip[::2], clip[1::2]
def deblend(clip):
clip = core.std.SeparateFields(clip, tff=True)
clip = core.std.SelectEvery(clip, cycle=10, offsets=[0, 1, 3, 5, 6, 8])
a1 = clip[0::6]
a2 = clip[1::6]
b1, b2 = interpolate_fields(clip[2::6], TFF=False)
c2 = clip[3::6]
c1 = clip[4::6]
d1, d2 = interpolate_fields(clip[5::6], TFF=True)
clip = core.std.Interleave(clips=[a1, a2, b1, b2, b1, c2, c1, d2, d1, d2])
clip = core.std.DoubleWeave(clip, tff=True)
clip = core.std.SelectEvery(clip, cycle=2, offsets=0)
return clip
Webm demonstrating the effect. The decimated output was not run through QTGMC.
sorry here at /mpv/ we only have knowledge of upscaling and shaders.
If I believe hard enough, someone from Irrational Encoding Wizardry will see my post and save me. Making a Doom9 account is too straightforward.
post a clip of the source
This is exactly the sort of thing when you’re 14 and dream about.
“Aww dood it wood be so kewl if I wake up in front of big tee vee and play all the vee gays I want!”
My back hurts just looking at this photo. Do you just sit down for 8 hours in your comfy recliner and code? Is it productive code or is this just like one of those sys ad type guys that thinks he needs more monitors so he can pretend to monitor resources
>Is it productive code or is this just like one of those sys ad type guys that thinks he needs more monitors so he can pretend to monitor resources
tfw I've been working the past 5 years of my life professionally using only a single monitor, all because my secondary monitor broke and I was too lazy to replace it, so I eventually realized I don't even need it
Hey. I listen through mpv a radio station (in celluloid) and the program save even 4 houres of the recording in the background/cache. How can I save it on hdd or where I can find the file cuz I can't find and an api documentation also don't help
>update mpv
>now scrolling sets up volume
how the fuck do I change it
https://github.com/mpv-player/mpv/pull/12329
updoot this
why would they commit such a stupid change
lmao this ccp chinc https://github.com/dyphire/mpv-config/issues/34
looking at his profile history, the person who posted that issue is probably equally as schizophrenic
How did you access the profile?
huh? you just click on i--
>aicynide's activity is private
he just privated it kek, his profile showed that he opens like 5+ issues per day
Did po5 ever elaborate on why he got banned?
>his battle station is in a house like a homosexual
ngmi
sir pls python support in mpv when https://github.com/mpv-player/mpv/pull/11574
my config
hwdec=auto
keep-open=yes
thoughts ?
use auto-safe
hwdec=no
>enter thread thinking i'll find some cool scripts to install or functionality
>it's schizo math about optimal gamma values or other shit
https://github.com/mpv-player/mpv/pull/11574#issuecomment-1712768071
Indian fight!
Sirs, please merge.
I noticed aicynide also gave the absolutely most helpful reply in
I've aicynide2 ready
Any anons who reached nirvana on their current setup?
While I'm pretty happy with image and scaling quality, I haven't looked into HDR, tonemapping, calibration and all that fun stuff.
vo=gpu-next
gpu-api=vulkan
profile=gpu-hq
glsl-shaders="~~/shaders/compute/ravu-zoom-ar-r3-rgb.hook"
video-sync=display-resample
interpolation
tscale=oversample
hwdec=auto
I have reached nirvana as far as playback is concerned. With amdgpu.freesync_video=1 and xrandr.lua, my display seamlessly switches to the appropriate refresh rate. You could achieve similar results on a fixed refresh rate display with custom modelines for 48Hz and whatnot, but changing refresh rate will cause the screen to blank for a couple seconds when opening and closing mpv. Idk if there are wayland compositors that allow for something similar.
>skylake
So you have vulkan support, good. Try using vo=gpu-next and gpu-api=vulkan, It's a bit more efficient than OpenGL and should help performance.
Its fake
[vo/vaapi] OSD format not supported. Disabling OSD.
how to make this work?
>how to make this work
By not using vo=vaapi. Can't you use vo=gpu or gpu-next?
vo=gpu is same as not using vaapi on command line, it has much higher CPU overhead
I thought VAAPI should be used with Intel integrated graphics?
in any case I get: "Using hardware decoding (vaapi)."
???
then why OSD works, but it doesn't work when vo=vaapi is selected?
>Note that there is absolutely no reason to use this, other than compatibility. This is low quality, and has issues with OSD. We strongly recommend that you use --vo=gpu with --hwdec=vaapi instead.
I see, vo=gpu is actually default here
if the OSD would work the reason would be a better performance
just use your browser as a video player at that point? Sure, bilinear scaling is fast but at that point there's no reason to be using a video player
The OSD does work with vo=vaapi, but breaks as soon as you resize the window or go fullscreen.
What kind of toaster are you using? Even my sandy bridge laptop can use vo=gpu with simpler shaders like FSR or CAS without dropping frames.
sandy bridge works just fine with spline36 and mitchell for scaling at 1080p
Exactly, I don't see the point of using vo=vaapi when vo=gpu with its default or higher quality scaling works fine on decade old integrated graphics.
Skylake, yes it works but GPU usage is quite high
I thought vaapi should work better, apparently not
The most cringe thing about this picture is that they are using iTunes on Windows. ew
imagine needing a giant vertical screen to constantly have a manual of a video player open, just to watch something.
Real men just use windows media player
https://github.com/mpv-player/mpv/issues/12373
>Works as expected
Doesn't actually work as expected because behavior changed, hence the bug report. Starting to hate kasper.
Thanks to whichever anon reported it.
>Works as expected this are true native dimensions of the video frame.
it's expected behavior to violate the spec and report an incorrect resolution? those extra 8 pixels are literally garbage padding data added by the encoder and are never meant to be seen or known by the user
Have you guys tried an older build?
zhongfly builds?
no-config?
verified/re-checked files?
I tried 0.36 and zhongfly and still fucked
Sample file from
works for me. As other anon said, must be ffmpeg. Linux with latest stable builds of all deps except libplacebo which is latest git, and mpv git.
Might try a much older build from around March.
Well, it looks like a regression, current mpv v0.36.0-274-gbe922231 doesn't work but mpv-x86_64-v3-20230115-git-6cdce9e does work just fine.
use ffplay directly to reduce variables, see if you can bisect it
Like the other anon said, you can try to bisect with ffplay which shouldn't take too long. Gyan keeps all of their builds here:
https://github.com/GyanD/codexffmpeg/tags
rip your eyesight
What are your audio mixing matrix for 5.1/7.1 to headphones?
I just let my AO do its thing. But if you want to get fancy you can use ffmpeg filters to create custom downmixing profiles.
af=sofalizer="C:/mpv/sofalizer/ClubFritz6.sofa":type=time:gain=-8
http://sofacoustics.org/data/database/clubfritz/
Loads of files available, some might work better for you. It's depended on your head. ClubFritz6 is cargocult, 4 is used quite a bit as well but I think that downmixes to 44.1kHz instead of 48kHz like 6. https://www.sofaconventions.org/mediawiki/index.php/Files
type=time is slower according ffmpeg, no clue if it's better. gain=-8 I did because it clipped on some of my files originally and I have an analogue amplifier anyway.
You can also try this, though it's less sophisticated. I'm pretty sure I just modified the 5.1 for 6.1 and 7.1. Based on https://superuser.com/questions/852400/properly-downmix-5-1-to-stereo-using-ffmpeg#1410620
For 5.1:
af=lavfi=[pan=stereo|FL=0.5*FC+0.707*FL+0.707*SL+0.5*LFE|FR=0.5*FC+0.707*FR+0.707*SR+0.5*LFE]
For 7.1:
af=lavfi=[pan=stereo|FL=0.5*FC+0.707*FL+0.707*SL+0.707*BL+0.5*LFE|FR=0.5*FC+0.707*FR+0.707*SR+0.707*BR+0.5*LFE]
And in case you encounter 6.1:
af=lavfi=[pan=stereo|FL=0.5*FC+0.707*FL+0.707*SL+0.5*BC+0.5*LFE|FR=0.5*FC+0.707*FR+0.707*SR+0.5*BC+0.5*LFE]
This wouldn't include 2.1 though. Sofalizer does. There are some other things you can try. Like af=lavfi=[pan=stereo|c0=c2+0.30*c0+0.30*c4|c1=c2+0.30*c1+0.30*c5] Which somebody mixed and called "nightmode dialogue", to boost dialogue but not have too much volume during the night. Sounds like shit to me.
Some reference links. Not required.
https://www.reddit.com/r/mpv/comments/11cr5u9/is_it_possible_to_use_the_headphone_filter/
https://gist.github.com/kevinlekiller/9fd21936411d8dc5998793470c6e3d16
https://ffmpeg.org/ffmpeg-filters.html#sofalizer
You could use that script I guess, but I just have this profile.:
[multichannel]
profile-cond=(p["audio-params/channel-count"] > 2)
af=sofalizer="C:/mpv/sofalizer/ClubFritz6.sofa":type=time:gain=-8
Bisecting: 530 revisions left to test after this (roughly 9 steps)
>9 steps
So what's the bad commit or build? Shouldn't take much more than five minutes to test that.
https://github.com/FFmpeg/FFmpeg/commit/fc429d785e9e24c5520ce716d4bc3b5547e581eb
Not the SIMD queen 🙁
>May 28
Surprised no one noticed until now
So is this the commit that fucked things up?
Yeah something got fucked up between April and May:
April 30 build works perfectly:
https://pastebin.com/raw/k5PMznBt
May 31 build is fucked up with the same artifacts:
https://pastebin.com/raw/9n4msdFM
A full month is going to be a lot of commits for ffmpeg.
You can fork shinchiro, add
GIT_REMOTE_NAME origin
GIT_TAG main
GIT_RESET ad417eb5
to ffmpeg.cmake, disable the included patch, and hope the CI builds without needing other modifications. I used a short hash to avoid spam detection, might not work with git reset.
>GIT_TAG main
Copy/paste mistake. ffmpeg still uses master. Just replace the PATCH_COMMAND line with those three lines and hopefully it will build.
Bro I am not going to compile ffmpeg, I don't watch Shinkai shit to begin with, I only downloaded the movie to check if
was posting fud or not.
You are welcome to file the bug report to ffmpeg's tracker, the test clip is right here
and the probable regression is right here
>I am not going to compile ffmpeg
You don't, GitHub CI does. Useful if anyone else has the issue because sasukefag insists on using git for nearly every dependency.
>GitHub CI does
I don't have a GitHub account
Good morning sirs
llyyr already sending ffmpeg patches huh
and LULZ will seethe yet again
Thanks for fixing it, llyyr
It's literally one anon who's buttmad llyyr told him his issue sucks
Thank you llyyr
Hey llyyr now that you are here, tell haasn to fix the Hermite parameters:
https://code.videolan.org/videolan/libplacebo/-/blob/master/src/filters.c#L478
They should be B=0 and C=0, see:
https://www.imagemagick.org/Usage/filter/#cubic_bc
Can I merge two videos with ffmpeg just by copying or do I have to reencode to make it work?
Use concat:
https://trac.ffmpeg.org/wiki/Concatenate#Concatenatingmediafiles
>If you have media files with exactly the same codec and codec parameters you can concatenate them
Can I still concoct two files with different bitrates (but same crf settings)?
Yeah no problem
Does ffmpeg -f concat -safe 0 -i mylist.txt -c copy output.wav also work if the files I want to merge (both mkv) have one video and audio stream? Or should I use the concat protocol in that case?
Is the default mpv hdr to sdr conversion good?
Sometimes it looks kind of dark
By default what do you mean exactly?
by default it will use vo=gpu, so no. use vo=gpu-next for tonemapping
Use gpu-next
make sure to specify target-peak to your display's value if it's still dark crank up the monitor's brightness and target-peak
I think I discovered a bug with polar antiring.
new
yikes no thanks
how to reduce the time combining planes and color conversion takes?
try gpu-next, the latest and greatest in FAST rendering (if you're using vulkan or d3d11, that is)
gpu-next doesnt work on android
it does tho
I tried vo=gpu-next on my fairphone 4 with the f-droid build and couldn't play anything. The snapdragon 750g it uses does support vulkan. No idea if the ROM it came with is any different.
but vo=gpu does. try that with gpu-api=vulkan and see if it makes things better.
haasn pls fix thanks
https://github.com/mpv-player/mpv/issues/12375
downscale antiring doesn't work on ortho filters btw, so your last screenshot wouldn't have produced anything regardless
don't update to next ffmpeg release
I will, cry about it homosexual.
meds
How fast would it be merged into mpv-git? I want to see if
gets fixed.
whenever a new build is published
https://github.com/zhongfly/mpv-winbuild this updates every day or on every new mpv commit
that fixed
my original issue, can't believe llyyr actually fixed something the SIMD trans queen messed up, go figure.
How is my mpv android config?
What should I change for the best quality?