Describe the use-case better. What does "keep up" mean? Adjust by making it slower or faster?
You might be better served by grabbing the stream to a local file, and playing with:
mpv <(tail -f -c +0 stream_file.ts)
This way, mpv will never close or not know how to buffer, you would still have the stream locally no matter what. You can let it buffer a bit, and then adjust speed to 0.99 (just press '[') if avoiding stops-starts is what you're after. You can go back to 1.00x speed after a while (just press ']').
When you watch an actual live stream (not a recording of one) you usually want to have the lowest latency possible. Events like stream startup or buffering due to transient network hiccups can get you significantly behind. The cache displays, say, a couple of seconds buffered ahead, when it should be 0.
Another way why you could get behind is if the player is playing it slightly slower than the stream source due to clock divergence.
One solution is to speed up playback until you're caught up. Once caught up, playback speed should be set to 1 to avoid buffering again.
You always need a reasonable buffer due to network uncertainties, thereafter it's 1:1 so it's pointless beyond faster playback to catch up to the current stream if e.g. you pause it.
so what version did you use before around the 28th, and what command is it
try n4.4 from here:
https://github.com/BtbN/FFmpeg-Builds/releases/tag/autobuild-2024-04-04-12-57/
It would have been 6,6.1 or 7
I've had this script for atleast a year running nightly before this 8 pixel shit came up
ffmpeg -y -hide_banner -vaapi_device /dev/dri/renderD129 -i "$input_file"
-vf 'format=p010,hwupload' -c:v hevc_vaapi -b:v 0 -qp 22
-rc_mode 1 -compression_level 12 -global_quality 50 -bf 1
-profile:v 2 -tier 1 -async_depth 64 -an "$output_file"
>It would have been 6,6.1 or 7
OK just tell distribution, distribution release and package name at this point
did 4.4 work?
2 weeks ago
Anonymous
4.4 did not work
Im on arch btw
I am -Syu'd so latest?
package name of what?
i use ffmpeg from repo
This green line showed up and my encodes are 1920x1088
2 weeks ago
Anonymous
pacman.log please, need to know from which to which ffmpeg version you updated with
2 weeks ago
Anonymous
i was using 7 but downgraded to 6.1.1-7.1 right when the issues happened.
Going back to 7 dosen't solve the problem either
2 weeks ago
Anonymous
OK so you were using some git master AUR ? because the official ffmpeg package is 6.1 and was never 7.0 from what i can see. 7.0 got released about a week ago
moreover i don't see any relevant changes in the last month that are present in both the ffmpeg 6.1 and 4.4 branch that is relevant to your issue
too long; didn't read: VAAPI issue. arch moment
any pointers to how to get an h264 webcam stream over http to as low latency as possible?
my current setup is this and my latency is around 2 seconds which is not great
ffmpeg -i 0 -listen 1 -c:v libx264 -preset veryfast -tune zerolatency -f h264 -crf 17 http://localhost:8008
looking into it rtp seems like it might be a good solution but straight out of the box the stream got very corrupted, I know it has less error correction but this seems worse than it should be, especially considering it's currently over localhost
it's really boxy and laggy and on the receiving side from ffplay I get h264 errors about corrupted macroblocks, "mb_type 772 in I slice too large" etc
I'm guessing it has to do with the SDP file, I haven't used it before so I threw something together based on samples I found but I'm guessing this isn't good enough
my current pipeline for generating the stream is: (it's using the python wrapper because actually it's not really a simple webcam, I use a python wrapper to communicate with the camera and do some operations of the frames using opencv but it's pretty much the same thing as regular ffmpeg)
ffmpeg
.input('pipe:', format='rawvideo', codec="rawvideo", pix_fmt='bgr24', s='1920x1080', framerate=5, hwaccel="cuda")
.output(
'rtp://localhost:8008',
listen=1,
pix_fmt="yuv420p",
preset="ultrafast",
vcodec="libx264",
f="rtp",
crf=17,
an=None,
tune='zerolatency',
)
.overwrite_output()
.run_async(pipe_stdin=True)
and my SDP file is
v=0
o=- 0 0 IN IP4 127.0.0.1
s=FFmpeg RTP Stream
c=IN IP4 127.0.0.1
t=0 0
m=video 8080 RTP/AVP 96
a=rtpmap:96 H264/90000
a=fmtp:96 packetization-mode=1;
a=control:streamid=1234
a=framerate:5
moron here, How do I get a program to include the header 'mpv/client.h'? Trying to install something through pip which wants that but dunno how to tell it where to find it
>mpv headers are usually installed in a standard system path already, where a compiler should find them without extra frags
Well its certainly not finding it by itself >pkgconfig
how would I use that with installing it through pip in cmd? Also I'm doing this in windows, and am a real moron, so explain to it like I'm 5
1. install gentoo
2. emerge python-mpv
on that note, pympv is pretty dead (including its fork), do python-mpv instead
2 weeks ago
Anonymous
>on that note, pympv is pretty dead (including its fork), do python-mpv instead
I can't, pympv is needed as a dependency for another program I'm trying to install, which is what this is all about. When I try to build that from source I get this error, so I assume its from me not having pympv, so then I tried getting that, and been running in these issues
2 weeks ago
Anonymous
>mpv headers are usually installed in a standard system path already, where a compiler should find them without extra frags
Well its certainly not finding it by itself >pkgconfig
how would I use that with installing it through pip in cmd? Also I'm doing this in windows, and am a real moron, so explain to it like I'm 5
Ok but jokes aside how would I do this in windows?
I compiled mpv on my system (Debian 12) following the instructions that I found in the mpv-build repos and everything works fine, except for AV1 playback. When I run mpv on the terminal, it says that my system doesn't support AV1 hwdec, which was expected, but it isn't falling back to software decoding, I only get a black screen instead. I have the libdav1d-dev dependency installed.
So... before I spend another 15 minutes compiling everything, does anyone know what am I missing? I noticed that ffmpeg's configure file mentions the flag --enable-libdav1d, should I assume that isn't set by default? I'm not familiar with this stuff, so I can't quite tell what I'm looking at. Also, if that's the case, is there any other flag that I should probably enable? mpv-build only gives some examples for encoding, which I'm not interested.
>it says that my system doesn't support AV1 hwdec, which was expected, but it isn't falling back to software decoding, I only get a black screen instead
that's weird
try --no-hwdec or set --hwdec-codecs
you can check if dav1d is available by running mpv --vd=help
Running mpv --vd=help, the only line that mentions AV1 says:
>av1 - Alliance for Open Media AV1
I just finished rebuilding with --enable-libdav1d added to ffmpeg_options and it did indeed fix the issue, and --vd=help also lists dav1d as it should. I also added --enable-nonfree and --enable-libfdk-aac since I saw a pull request mentioning those.
basically this thinly veiled mpv general with off-topic pictures is made by a troony called llyyr who enters a bigger mental breakdown when someone else makes the thread and mpv isn't listed as first in the OP
without ffmpeg and yt-dlp in the OP, the thread dies unless it necrobumps the thread
it and the other trannies use this thread to talk shit to each other anonymously because they don't have the balls (no pun intended) to do it with their names on
They're really only making this to talk shit to each other. Takes not long time to identify them. I remember llyyr posted something on a github issue 15 seconds after it was posted in here.
basically this thinly veiled mpv general with off-topic pictures is made by a troony called llyyr who enters a bigger mental breakdown when someone else makes the thread and mpv isn't listed as first in the OP
without ffmpeg and yt-dlp in the OP, the thread dies unless it necrobumps the thread
it and the other trannies use this thread to talk shit to each other anonymously because they don't have the balls (no pun intended) to do it with their names on
not possible on mpv without 3rd party malware. use madVR
2 weeks ago
Anonymous
I may be misunderstanding, but I don't want to have like 60fps interpolation and such, I just don't want judder. My ideal I think would be having the display's refresh ratch lowered to match the media.
2 weeks ago
Anonymous
its not "60 fps" motion interpolation, it will blend the frames where there's judder so that you dont perceive any jitter during panning
2 weeks ago
Anonymous
I'm interested to try. Does madVR work with yt-dlp like how mpv does? Point is I've never used madVR, so I need to know how to use it to stream things
2 weeks ago
Anonymous
madVR is a video renderer, MPC-HC is a media player. and yes, MPC-HC works with yt-dlp the same way that mpv does
2 weeks ago
Anonymous
Actually I'm using a Thinkpad with Intel Iris, no dedicated GPU. Glance through the doom9 madvr thread suggests I'm toast
use vf to resample to 24hz
you won't be able to spot the duplicated frame
especially not in anime which already has tons of duplicated frames
there is like, a 1% chance the frame might be duplicated while panning which you then could tell
2 weeks ago
Anonymous
>just rape your video bro
concession accepted
2 weeks ago
Anonymous
moronic Black person
2 weeks ago
Anonymous
Very nice
2 weeks ago
Anonymous
I'm
Whenever the "camera" pans in 23.976 fps anime the judder effect doesn't look good on my 60hz display. What config to compensate?
, how can I do >use vf to resample to 24hz
Sorry if this should be obvious, I'm checking the documentation and stuff, it's just there's a lot of stuff -vf can do, it seems. I want to know how you'd do it
2 weeks ago
Anonymous
Wait, why would I want to resample the video to 24hz? The video is already 23.976, and the problem is my display is 60hz.
2 weeks ago
Anonymous
it won't
the solution is to stop using 60Hz, make yourself a 72Hz (or 72/1.001Hz) mode
2 weeks ago
Anonymous
I'll give this a shot. Is CRU the best way to do that?
2 weeks ago
Anonymous
i don't use windows so i don't have any recommendations for it
2 weeks ago
Anonymous
use display-sync in mpv, it will adjust the video speed ever so slightly to line the video up with your display
2 weeks ago
Anonymous
anyone who unironically recommends display-resample in these threads should get gangraped. enjoy your video and audio quality reduction
i was suggesting display resample for 24/1.001 > 24, that's a 0.1% difference
you reckon you can recognise audio playing 0.1% faster than intended?
the pitch is adjusted/accounted for
This is why madVR will always be superior. It can achive near perfect enough framepacing without having to reduce audio quality and change video speed. You're watching the movie exactly how the creators intended. madVR just has a superior vsync system to mpv, which has always been very jittery and stuttery even if you do use an integer multiple refreshrate. (and also oversample actually works)
2 weeks ago
Anonymous
>without having to reduce audio quality and change video speed
oh, you are so wrong
2 weeks ago
Anonymous
>without having to reduce audio quality and change video speed
madVR does not resample video or audio. Try again shill.
2 weeks ago
Anonymous
You are right, mpv is superior in its video speed adjustment ability. The thing that people have to use reclock externally with madVR.
2 weeks ago
Anonymous
>mpv is superior in it's ability to ruin videos and audio
This isn't a good flex chief. Also ReClock LMAO what is this, 2009? madVR's default vsync is superior everything else in terms of frametimings and low jitter.
2 weeks ago
Anonymous
You have the choice between the following: >audio/video desync >changing video timing >resampling audio
What you describe would achieve the first one.
2 weeks ago
Anonymous
Are you moronic? Do you think audio mode resamples audio and video too? You do realize you can have vsync without resampling right anon? You certainly can't be this moronic right anon...? Are mpvcels really this deprived of a good vsync mode that they think display-resample is the status quo? Have you ever thought that maybe mpv has flawed jittery vsync...?
2 weeks ago
Anonymous
If you don't resample audio, audio is your clock and will affect video speed and frame spacing. Unless you let them desync.
2 weeks ago
Anonymous
This homie doesn't even realize that all sensible vsync modes leave audio untouched (for audio quality and for exact bitstreaming) and repeat frames every 5 hours or so to maintain a/v sync ohnonono the absolute state of mpvcels. This is why no one uses your garbage, you are all extremely ill informed.
2 weeks ago
Anonymous
This just isn't true, mpv can do the same. It just requires that your audio clock is very close to the video clock.
2 weeks ago
Anonymous
This is very true actually, mpvcel. mpv is incapable of having the same vsync as madVR. In fact, if you have something like a 120hz screen, you will get frame repeats 5 (FIVE) times more often than someone who uses madVR because the audio mode doesn't know anything about your display, and only syncs based on video framerate. Now imagine if you have a 240hz screen and you're watching 23.976fps video, now you're getting frame repeats 10 times more than madVR! Absolute state of mpv. This is why you people cope by ruining your videos and audios through resampling.
2 weeks ago
Anonymous
Again, not true. You don't seem to know what the frick you're talking about.
2 weeks ago
Anonymous
kekkers you're the one who doesn't know what you're talking about. this is all very true. calculate the amount of time you think it would take for mpv to generate a repeat frame on your display. For example, on my uncalibrated 120ish hz display, it's about 40 seconds per repeat frame, (pretty bad but manageable), but for mpv you have to divide that by 5, so it's actually 8 seconds per repeat frame (jesus christ... The absolute state). Very easy to record and validate yourself. madVR is the only video renderer that is actually aware of your display refresh rate and will compensate for clock drifts in a sensible way.
Again, just cope. mpv is still eons behind madVR.
2 weeks ago
Anonymous
You convinced me to try madVR. Where can I download the Linux version?
2 weeks ago
Anonymous
>Loonix
My condolences. madVR is Windows-only kinoware. Many such cases. Just ask mpv devs to implement sensible vsync instead of copying what VLC does by default.
2 weeks ago
Anonymous
madvr endgame settings need for tranime.
2 weeks ago
Anonymous
>mpv has good A/V sy-- ACK
Step aside.
>at least mpv has smoothmot-- ACK
Step aside.
https://github.com/mpv-player/mpv/issues/13108 >mpv has good antiringing thou-- ACK
Step aside.
https://github.com/haasn/libplacebo/commit/9855b663f1da050c36dd2278ceb372044c44fc73 >mpv has good dithering thou-- ACK
Step aside.
https://github.com/mpv-player/mpv/issues/13881 >AT LEAST LIBPLACEBO ISN'T ABANDONWARE THOU-- ACKKKKKK
Step aside.
https://github.com/haasn/libplacebo/commits/master/
Massive MadVR sweep.
2 weeks ago
Anonymous
mpc-hc or mpc-be? Also provide config.
2 weeks ago
Anonymous
just buy a 5000$ box to play videos instead of using your computer to play videos, goy
kys shill
Is there a way to talk to some yt-dlp dev that isn't fricking discord? I have an issue but I'm not sure if I should turn it into a github issue. They seem to be struggling enough to deal with the ones they already have (1000+ open).
Is it normal for mpv to drop 1 frame once in a while? In the "output", not the "decode" section
Like once every 20 minutes if any at all
My goyfig is just
hwdec=yes
Why does A/V for display-vdrop show a number like +0.01 or whatever, but video-sync=audio always shows an A/V value of 0? I thought the way both of these compensate for A/V sync is the same, so why does vdrop have desynced video and audio? What's the point of using display-vdrop then if it has worse A/V sync than display-vdrop?
I want to to make a video with the video stream from source #0 and audio from source #1, cut it with -shortest and apply fade in and fade out to the first and last 3 sec of the audio.
>see a post claiming some tweaks can improve rendering quality significantly >check screenshot >can't see any difference >there's subtitle in screenshot
I'm surprised that people are autistic enough to notice and care about tiny quality difference while reading subtitles.
>Sorry I was not born in your country. Should I use "how can I ..." instead?
Shit excuse, you can learn proper English outside of the US and England. moron.
>Exactly. You could help, instead of bullying me for ESL...
i wasn't bullying you, i was explaining to someone else why he sees questions formulated like that so often. if anything i was supporting your post by claiming it was likely a genuine question and not a robot
The real problem is that your question doesn't make sense at all. You don't need to be a grammar nazi to express yourself.
>Sorry I was not born in your country. Should I use "how can I ..." instead?
Shit excuse, you can learn proper English outside of the US and England. moron.
Maybe you should learn Japanese to prove your point.
2 weeks ago
Anonymous
>Maybe you should learn Japanese to prove your point.
日本語は知るよ
2 weeks ago
Anonymous
>The real problem is that your question doesn't make sense at all.
of course it does you autistic moron >how to open all images from a webpage in a mpv as a playlist?
this is literally what he said. dissect how it DOESN'T make sense
>Sorry I was not born in your country. Should I use "how can I ..." instead?
this is how i'd phrase it;
"how can i make a playlist in mpv of all the pictures from a web page?"
though your post is comprehensible, some people just see esl-like patterns and get angry over it whether it makes sense or not. i get it, i too wonder why you come here, to english websites, i never post on websites that use a language i don't know. is it just because english is more popular, so there's more available information?
>though your post is comprehensible, some people just see esl-like patterns and get angry over it
exactly
>Sorry I was not born in your country. Should I use "how can I ..." instead?
this is how i'd phrase it;
"how can i make a playlist in mpv of all the pictures from a web page?"
though your post is comprehensible, some people just see esl-like patterns and get angry over it whether it makes sense or not. i get it, i too wonder why you come here, to english websites, i never post on websites that use a language i don't know. is it just because english is more popular, so there's more available information?
2 weeks ago
Anonymous
thank you teacher
2 weeks ago
Anonymous
>, i too wonder why you come here, to english websites, i never post on websites that use a language i don't know. is it just because english is more popular, so there's more available information?
They think their fellow native language speakers are all subhuman morons so they come here to spew esl babble because they deserve to talk to smarter people.
you're literally adjusting video speed and compromising video quality. let me guess do you also like +4% PAL video speed-ups? >you can't hear any difference
display-resample confirmed to be used by earlets.
when is ffmpeg getting an audio normalization filter similar to what YouTube and Spotify do? dynaudnorm is garbage and the audio loudness transitions are way too noticeable as scenes change
YouTube pre-measures loudness and stores it in metadata, so it's only a matter of applying an offset to normalize. I suppose Spotify does something similar.
I have been thinking about doing some Lua script that measures the loudness at runtime, but I'm not sure how to handle audio read-ahead.
At this point, I'm too lazy to try it, but maybe someone can take that idea and make it work.
Note that this concept would only work on locally stored files. Media over the network would only work if the stream contains any parsable loudness metadata, or if you cache the whole thing and measure it before starting playback.
Why does applying eq brightness/contrast/saturation filters increases the output filesize? Does the order of filters affect it, i.e. applying it before or after scaling?
>pixel clipper
you dont need it
cscale=ewa_lanczossharp
scale-antiring=0.6
That's what I tried (--video-sync=display-resample --tscale=mitchell --interpolation=yes) and it didn't fix anything and made everything blurry to boot. Any other known remedies? Thanks though, good to know that's the actual keyword
[...]
And you wont ever be a fish. Or useful.
>That's what I tried (--video-sync=display-resample --tscale=mitchell --interpolation=yes) and it didn't fix anything and made everything blurry to boot. Any other known remedies? Thanks though, good to know that's the actual keyword
I use this and happy with it.
video-sync=display-resample
tscale=sphinx
tscale-blur=0.65
GPU-next is trash. I only ever use it for HDR content and that's about it. Jitter, tearing, and other issues always pop up when I use GPU-next
does gpu-next still use ewa_lanczossharp for cscale? I can't see it anymore
also why no hwdec? and is vulkan better than vaapi?
Vulkan should generally have better GPU utilization than vaapi
why no hwdec? isn't graphics acceleration better?
I might be wrong about this, but i'm pretty sure mpv uses the GPU even without acceleration being enabled. You can get a little bit more use out of hardware decoding,
mpvisters, why does my anime look stuttery when it's panning sideways or upwards? Really takes the joy out of beautiful landscapes and the like. 240hz at 24 fps. Interpolation is gay but would probably help
That's what I tried (--video-sync=display-resample --tscale=mitchell --interpolation=yes) and it didn't fix anything and made everything blurry to boot. Any other known remedies? Thanks though, good to know that's the actual keyword
Omitting interpolation and sphinx instead of mitchell makes it slightly better but still headache inducing and unenjoyable. The stills look great though..
I recall it has only ever been satisfactory when I copied some riced out config that forced it to run at 240 fps when I was just starting out, but I'd rather avoid that
>pixel clipper
you dont need it
cscale=ewa_lanczossharp
scale-antiring=0.6
[...] >That's what I tried (--video-sync=display-resample --tscale=mitchell --interpolation=yes) and it didn't fix anything and made everything blurry to boot. Any other known remedies? Thanks though, good to know that's the actual keyword
I use this and happy with it.
video-sync=display-resample
tscale=sphinx
tscale-blur=0.65
Thanks I'll try that next
2 weeks ago
Anonymous
I also use this script, hope u can fix the issue anon.
https://github.com/CogentRedTester/mpv-changerefresh
2 weeks ago
Anonymous
>windows only
Wont work for me sadly, but the tscale-blur seems to be the right track, thank you!
2 weeks ago
Anonymous
there is xrandr script too on gitlab or github.
2 weeks ago
Anonymous
Thanks for making me check.
https://gitlab.com/lvml/mpv-plugin-xrandr
Works as advertised I suppose (set refresh to 143.xxxxxxxxxx and reset it properly after quitting, sadly the judder is still there. I'll play with the blur as that was the biggest step in the right direction so far
2 weeks ago
Anonymous
This script and
video-sync=audio
tscale=sphinx
tscale-blur=0.65
are satisfactory. Any obvious idiocies committed with that? Thanks for the help, I will now enjoy
2 weeks ago
Anonymous
if you dont use display-* sync method interpolation wont work anon.
2 weeks ago
Anonymous
Idiocy counter: 1
I will make a commented note but leave the config as is. Didn't even enable interpolation either. Whichever component it is, it works now.
2 weeks ago
Anonymous
ok moron tscale wont do anything if u dont use display-resample kek
2 weeks ago
Anonymous
That's very cool and I'll add a commented note to that as well. However it works now. Guess the script is the heavy hitter than. During testing I had all kinds of things enabled or flagged at launch and I am not motivated to cut down the irrelevant options now.
by smaller you bean the 16 version? I mean I run RAVU fine. what's the difference between the 16, 32, and 64 versions?
use dicky its better and faster
2 weeks ago
Anonymous
Do I have to use a toggle or are shaders enabled by default if I put in my conf >glsl-shader="~~/mpv/shaders/dicky-6x16-DS.glsl"
I have never heard of dicky. how can I tell if it's better than FSRCNNX or ArtCNN? those at least had measurements
https://artoriuz.github.io/blog/mpv_upscaling.html
ArtCNN is deprecated, use dicky as everyone else is saying.
what does the first number mean? I know the 2nd is # of filters.
1st is # of conv layers
so I tried dicky as everyone was saying and neither filter could do it at 60Hz. maybe when I get a 2nd RAM stick for dual channel things would improve a bit. but right now ravu-lite-ar-r4 is still the best
2 weeks ago
Anonymous
Don't fall for the meme.
2 weeks ago
Anonymous
what gpu do you have
if its really old maybe try 1x8 in fp16 directory
2 weeks ago
Anonymous
just an integrated Vega 6 (from the 4650U). single channel memory. what is fp16?
Don't fall for the meme.
I mean I wouldn't buy a dedicated desktop PC with a 4090 just for meme filters, just seeing what my own hardware could do.
I see. what does chroma even do? make the colors better? doesn't cscale do the same thing? I have ewa_lanczossharp (default for profile=high-quality and gpu-next)
neither. just use display-vdrop. why would you ever use display-resample-vdrop? it's literally the worst of both worlds >frame repeats AND fricked up video speed
The shaders you downloaded are probably the dp4a versions that require vo=gpu-next and gpu-api=vulkan. And if you're on old hardware they won't be faster anyway. Try the ones in the mpv/fp16 or results/fp16 directories instead.
2 weeks ago
Anonymous
>Try the ones in the mpv/fp16
that werked
2 weeks ago
Anonymous
oh I'm not using vulkan as my api yeah thanks that was it
2 weeks ago
Anonymous
??? display-vdrop is literally less CPU intensive since resampling isn't required.
any winblows users know if I can install something like Anki without it conflicting with my mpv install? it seems like Anki comes with its own mpv build
dicky is not showing up for me on mpv for some reason?? every other shader including artcnn and ravu show up in the stats menu but not dicky, this is on the same source btw, what am I doing wrong?
[...]
The shaders you downloaded are probably the dp4a versions that require vo=gpu-next and gpu-api=vulkan. And if you're on old hardware they won't be faster anyway. Try the ones in the mpv/fp16 or results/fp16 directories instead.
haasn did nothing wrong
Didn't he rape a kid?
that's a good thing
>https://github.com/mpv-player/mpv/issues/13862
I thought that Chinese people where supposed to be smart
Give me a script that keeps up with live streams by adjusting playback speed slightly.
Describe the use-case better. What does "keep up" mean? Adjust by making it slower or faster?
You might be better served by grabbing the stream to a local file, and playing with:
mpv <(tail -f -c +0 stream_file.ts)
This way, mpv will never close or not know how to buffer, you would still have the stream locally no matter what. You can let it buffer a bit, and then adjust speed to 0.99 (just press '[') if avoiding stops-starts is what you're after. You can go back to 1.00x speed after a while (just press ']').
When you watch an actual live stream (not a recording of one) you usually want to have the lowest latency possible. Events like stream startup or buffering due to transient network hiccups can get you significantly behind. The cache displays, say, a couple of seconds buffered ahead, when it should be 0.
Another way why you could get behind is if the player is playing it slightly slower than the stream source due to clock divergence.
One solution is to speed up playback until you're caught up. Once caught up, playback speed should be set to 1 to avoid buffering again.
You always need a reasonable buffer due to network uncertainties, thereafter it's 1:1 so it's pointless beyond faster playback to catch up to the current stream if e.g. you pause it.
just ask GPT-sama
bonus: give it the youtube link
s/ffmpeg/librempeg
Nice fingers Rajesh
>mpv ffmpeg yt-dlp
>yt-dlp
Doesn't seem to work properly:
ffmpeg is adding 8 pixels to my encodes as of around he 28th i've found.
The command has not been altered since before then.
Where do i start?
Binary diff.
Bisecting ffmpeg.
Cant go back,to many things require new versions of ffmpeg libs. Id have to downgrade the entire system partially and thats even more of a mess
halp?
*rapes u*
static
https://github.com/BtbN/FFmpeg-Builds/
issue still presents its self with a static build from 9 hours ago
master or n7?
master
i linked static builds so you could downgrade until the regression(?) disappears
try n7 and so on
its present in 5,6,7 branches
so what version did you use before around the 28th, and what command is it
try n4.4 from here:
https://github.com/BtbN/FFmpeg-Builds/releases/tag/autobuild-2024-04-04-12-57/
It would have been 6,6.1 or 7
I've had this script for atleast a year running nightly before this 8 pixel shit came up
ffmpeg -y -hide_banner -vaapi_device /dev/dri/renderD129 -i "$input_file"
-vf 'format=p010,hwupload' -c:v hevc_vaapi -b:v 0 -qp 22
-rc_mode 1 -compression_level 12 -global_quality 50 -bf 1
-profile:v 2 -tier 1 -async_depth 64 -an "$output_file"
>It would have been 6,6.1 or 7
OK just tell distribution, distribution release and package name at this point
did 4.4 work?
4.4 did not work
Im on arch btw
I am -Syu'd so latest?
package name of what?
i use ffmpeg from repo
This green line showed up and my encodes are 1920x1088
pacman.log please, need to know from which to which ffmpeg version you updated with
i was using 7 but downgraded to 6.1.1-7.1 right when the issues happened.
Going back to 7 dosen't solve the problem either
OK so you were using some git master AUR ? because the official ffmpeg package is 6.1 and was never 7.0 from what i can see. 7.0 got released about a week ago
moreover i don't see any relevant changes in the last month that are present in both the ffmpeg 6.1 and 4.4 branch that is relevant to your issue
too long; didn't read: VAAPI issue. arch moment
amd encoder? thats a hardware bug
https://gitlab.freedesktop.org/mesa/mesa/-/issues/9185
i literally have hundreds of copies using the same script where it didnt do this,and its not av1 though?
nkdsi
profile=high-quality
Rate my config.
any pointers to how to get an h264 webcam stream over http to as low latency as possible?
my current setup is this and my latency is around 2 seconds which is not great
ffmpeg -i 0 -listen 1 -c:v libx264 -preset veryfast -tune zerolatency -f h264 -crf 17 http://localhost:8008
just don't encode it bro
I need some kind of encoding as a requirement
looking into it rtp seems like it might be a good solution but straight out of the box the stream got very corrupted, I know it has less error correction but this seems worse than it should be, especially considering it's currently over localhost
it's really boxy and laggy and on the receiving side from ffplay I get h264 errors about corrupted macroblocks, "mb_type 772 in I slice too large" etc
I'm guessing it has to do with the SDP file, I haven't used it before so I threw something together based on samples I found but I'm guessing this isn't good enough
my current pipeline for generating the stream is: (it's using the python wrapper because actually it's not really a simple webcam, I use a python wrapper to communicate with the camera and do some operations of the frames using opencv but it's pretty much the same thing as regular ffmpeg)
ffmpeg
.input('pipe:', format='rawvideo', codec="rawvideo", pix_fmt='bgr24', s='1920x1080', framerate=5, hwaccel="cuda")
.output(
'rtp://localhost:8008',
listen=1,
pix_fmt="yuv420p",
preset="ultrafast",
vcodec="libx264",
f="rtp",
crf=17,
an=None,
tune='zerolatency',
)
.overwrite_output()
.run_async(pipe_stdin=True)
and my SDP file is
v=0
o=- 0 0 IN IP4 127.0.0.1
s=FFmpeg RTP Stream
c=IN IP4 127.0.0.1
t=0 0
m=video 8080 RTP/AVP 96
a=rtpmap:96 H264/90000
a=fmtp:96 packetization-mode=1;
a=control:streamid=1234
a=framerate:5
any idea what to do better?
What wrapper lib is that?
ffmpeg-python pip package
>libx264
Use h264_nvenc instead it's faster
moron
Read
If http is not a requirement, use rtp.
mpv yt-dlp dwm dmw bmw fsb nsdap homosexual st mpd
Pedotroon thread
TRVTHNVKE
We ARE girls, you may like it or not it doesn't matter.
>We
Who are you calling "We"
Go back to
If she/they were talking about the whole mpv general + contributors (which is basically the same).
How is it still alive?
troony thread
Yes
List top bug that should be fixed in mpv.
lack of full repo access for username wm4
Ask JEEB
How do I install AV1 codec on Windows?
moron here, How do I get a program to include the header 'mpv/client.h'? Trying to install something through pip which wants that but dunno how to tell it where to find it
mpv headers are usually installed in a standard system path already, where a compiler should find them without extra frags (e.g. in `/usr/include`).
In any case, `pkg-config --cflags mpv` should print the compile flags you need. `pkg-config --libs mpv` will print the link ones.
>mpv headers are usually installed in a standard system path already, where a compiler should find them without extra frags
Well its certainly not finding it by itself
>pkgconfig
how would I use that with installing it through pip in cmd? Also I'm doing this in windows, and am a real moron, so explain to it like I'm 5
1. install gentoo
2. emerge python-mpv
on that note, pympv is pretty dead (including its fork), do python-mpv instead
>on that note, pympv is pretty dead (including its fork), do python-mpv instead
I can't, pympv is needed as a dependency for another program I'm trying to install, which is what this is all about. When I try to build that from source I get this error, so I assume its from me not having pympv, so then I tried getting that, and been running in these issues
Ok but jokes aside how would I do this in windows?
Could anyone help me solve this issue?
I compiled mpv on my system (Debian 12) following the instructions that I found in the mpv-build repos and everything works fine, except for AV1 playback. When I run mpv on the terminal, it says that my system doesn't support AV1 hwdec, which was expected, but it isn't falling back to software decoding, I only get a black screen instead. I have the libdav1d-dev dependency installed.
So... before I spend another 15 minutes compiling everything, does anyone know what am I missing? I noticed that ffmpeg's configure file mentions the flag --enable-libdav1d, should I assume that isn't set by default? I'm not familiar with this stuff, so I can't quite tell what I'm looking at. Also, if that's the case, is there any other flag that I should probably enable? mpv-build only gives some examples for encoding, which I'm not interested.
>it says that my system doesn't support AV1 hwdec, which was expected, but it isn't falling back to software decoding, I only get a black screen instead
that's weird
try --no-hwdec or set --hwdec-codecs
you can check if dav1d is available by running mpv --vd=help
Running mpv --vd=help, the only line that mentions AV1 says:
>av1 - Alliance for Open Media AV1
I just finished rebuilding with --enable-libdav1d added to ffmpeg_options and it did indeed fix the issue, and --vd=help also lists dav1d as it should. I also added --enable-nonfree and --enable-libfdk-aac since I saw a pull request mentioning those.
Thanks for the reply though!
what is this thread about
all your wildest desires
basically this thinly veiled mpv general with off-topic pictures is made by a troony called llyyr who enters a bigger mental breakdown when someone else makes the thread and mpv isn't listed as first in the OP
without ffmpeg and yt-dlp in the OP, the thread dies unless it necrobumps the thread
it and the other trannies use this thread to talk shit to each other anonymously because they don't have the balls (no pun intended) to do it with their names on
They're really only making this to talk shit to each other. Takes not long time to identify them. I remember llyyr posted something on a github issue 15 seconds after it was posted in here.
>llyyr
found this guy's reddit account and apparently he's a massive zionist. wtf? why would they make this thread?
Why do these programs need a general?
>Why do these programs need a general?
has akemi's backend recovered yet?
likely not, it's been wrapping pointers day in and day out.
Whenever the "camera" pans in 23.976 fps anime the judder effect doesn't look good on my 60hz display. What config to compensate?
it's over. just give up. if anyone says display-shartssample, just ignore them. it's completely broken currently.
install MPC-HC+madVR, then enable SmoothMotion. Or alternatively use modeline selection to drop your refreshrate to 48
>modeline selection
I'm on Windows, how can I drop fps to 48?
not possible on mpv without 3rd party malware. use madVR
I may be misunderstanding, but I don't want to have like 60fps interpolation and such, I just don't want judder. My ideal I think would be having the display's refresh ratch lowered to match the media.
its not "60 fps" motion interpolation, it will blend the frames where there's judder so that you dont perceive any jitter during panning
I'm interested to try. Does madVR work with yt-dlp like how mpv does? Point is I've never used madVR, so I need to know how to use it to stream things
madVR is a video renderer, MPC-HC is a media player. and yes, MPC-HC works with yt-dlp the same way that mpv does
Actually I'm using a Thinkpad with Intel Iris, no dedicated GPU. Glance through the doom9 madvr thread suggests I'm toast
Built-in interpolation I guess, though don't expect much unless you tweak it or use vapoursynth interpolation
get a 120hz display you fricking poorgay. 120hz is the MINIMUM these days
this doesn't fix the issue, brainlet-kun
yet it does poorlet baka
24fps fits perfectly into 120hz
120 / 23.97602397602398 = 5.0049999999999991700041666666668
wow... perfect!
use vf to resample to 24hz
you won't be able to spot the duplicated frame
especially not in anime which already has tons of duplicated frames
there is like, a 1% chance the frame might be duplicated while panning which you then could tell
>just rape your video bro
concession accepted
moronic Black person
Very nice
I'm
, how can I do
>use vf to resample to 24hz
Sorry if this should be obvious, I'm checking the documentation and stuff, it's just there's a lot of stuff -vf can do, it seems. I want to know how you'd do it
Wait, why would I want to resample the video to 24hz? The video is already 23.976, and the problem is my display is 60hz.
it won't
the solution is to stop using 60Hz, make yourself a 72Hz (or 72/1.001Hz) mode
I'll give this a shot. Is CRU the best way to do that?
i don't use windows so i don't have any recommendations for it
use display-sync in mpv, it will adjust the video speed ever so slightly to line the video up with your display
This is why madVR will always be superior. It can achive near perfect enough framepacing without having to reduce audio quality and change video speed. You're watching the movie exactly how the creators intended. madVR just has a superior vsync system to mpv, which has always been very jittery and stuttery even if you do use an integer multiple refreshrate. (and also oversample actually works)
>without having to reduce audio quality and change video speed
oh, you are so wrong
>without having to reduce audio quality and change video speed
madVR does not resample video or audio. Try again shill.
You are right, mpv is superior in its video speed adjustment ability. The thing that people have to use reclock externally with madVR.
>mpv is superior in it's ability to ruin videos and audio
This isn't a good flex chief. Also ReClock LMAO what is this, 2009? madVR's default vsync is superior everything else in terms of frametimings and low jitter.
You have the choice between the following:
>audio/video desync
>changing video timing
>resampling audio
What you describe would achieve the first one.
Are you moronic? Do you think audio mode resamples audio and video too? You do realize you can have vsync without resampling right anon? You certainly can't be this moronic right anon...? Are mpvcels really this deprived of a good vsync mode that they think display-resample is the status quo? Have you ever thought that maybe mpv has flawed jittery vsync...?
If you don't resample audio, audio is your clock and will affect video speed and frame spacing. Unless you let them desync.
This homie doesn't even realize that all sensible vsync modes leave audio untouched (for audio quality and for exact bitstreaming) and repeat frames every 5 hours or so to maintain a/v sync ohnonono the absolute state of mpvcels. This is why no one uses your garbage, you are all extremely ill informed.
This just isn't true, mpv can do the same. It just requires that your audio clock is very close to the video clock.
This is very true actually, mpvcel. mpv is incapable of having the same vsync as madVR. In fact, if you have something like a 120hz screen, you will get frame repeats 5 (FIVE) times more often than someone who uses madVR because the audio mode doesn't know anything about your display, and only syncs based on video framerate. Now imagine if you have a 240hz screen and you're watching 23.976fps video, now you're getting frame repeats 10 times more than madVR! Absolute state of mpv. This is why you people cope by ruining your videos and audios through resampling.
Again, not true. You don't seem to know what the frick you're talking about.
kekkers you're the one who doesn't know what you're talking about. this is all very true. calculate the amount of time you think it would take for mpv to generate a repeat frame on your display. For example, on my uncalibrated 120ish hz display, it's about 40 seconds per repeat frame, (pretty bad but manageable), but for mpv you have to divide that by 5, so it's actually 8 seconds per repeat frame (jesus christ... The absolute state). Very easy to record and validate yourself. madVR is the only video renderer that is actually aware of your display refresh rate and will compensate for clock drifts in a sensible way.
Again, just cope. mpv is still eons behind madVR.
You convinced me to try madVR. Where can I download the Linux version?
>Loonix
My condolences. madVR is Windows-only kinoware. Many such cases. Just ask mpv devs to implement sensible vsync instead of copying what VLC does by default.
madvr endgame settings need for tranime.
>mpv has good A/V sy-- ACK
Step aside.
>at least mpv has smoothmot-- ACK
Step aside.
https://github.com/mpv-player/mpv/issues/13108
>mpv has good antiringing thou-- ACK
Step aside.
https://github.com/haasn/libplacebo/commit/9855b663f1da050c36dd2278ceb372044c44fc73
>mpv has good dithering thou-- ACK
Step aside.
https://github.com/mpv-player/mpv/issues/13881
>AT LEAST LIBPLACEBO ISN'T ABANDONWARE THOU-- ACKKKKKK
Step aside.
https://github.com/haasn/libplacebo/commits/master/
Massive MadVR sweep.
mpc-hc or mpc-be? Also provide config.
just buy a 5000$ box to play videos instead of using your computer to play videos, goy
kys shill
vlc handbrake youtube.com
https://github.com/mpv-player/mpv/issues/13880
it's so over, mac sisters how do you cope?
akemi's backend needs correction
why would one use mpv over quicktime
Is there a way to talk to some yt-dlp dev that isn't fricking discord? I have an issue but I'm not sure if I should turn it into a github issue. They seem to be struggling enough to deal with the ones they already have (1000+ open).
Is it normal for mpv to drop 1 frame once in a while? In the "output", not the "decode" section
Like once every 20 minutes if any at all
My goyfig is just
hwdec=yes
meant "decoder" not "decode"
no, it's not normal. But if your config is hwdec=yes you should set profile=fast
So if I get rid of hwdec=yes the occasional frame drop will disappear?
profile=fast sounds like it'll look like shit even if that does fix it
profile=fast unironically looks better than defaults, which just adds a bunch of shit artifacts to the videos
cscale should probably go back to bilinear but lanczos is fine for scale.
if you want a non-ringing chroma scaler just use mitchell
I miss it
Why does A/V for display-vdrop show a number like +0.01 or whatever, but video-sync=audio always shows an A/V value of 0? I thought the way both of these compensate for A/V sync is the same, so why does vdrop have desynced video and audio? What's the point of using display-vdrop then if it has worse A/V sync than display-vdrop?
wmv was better and you know it
I want to to make a video with the video stream from source #0 and audio from source #1, cut it with -shortest and apply fade in and fade out to the first and last 3 sec of the audio.
<code>ffmpeg -y -hide_banner -i video.mp4 -i audio.opus -c:v copy -map 0:v:0 -c:a opus -map 1:a:0 -max_interleave_delta 1G -shortest -filter_complex "[1:a]afade=d=3,areverse,afade=d=3,areverse" -strict -2 -movflags +faststart output.mp4</code>
How do I do audio fade in and fade out before -shortest is called? filter_complex would cut out the fadeout part if the source #0 video is shorter.
>see a post claiming some tweaks can improve rendering quality significantly
>check screenshot
>can't see any difference
>there's subtitle in screenshot
I'm surprised that people are autistic enough to notice and care about tiny quality difference while reading subtitles.
Another babyduck cope post
how to open all images from a webpage in a mpv as a playlist?
Why do people ask questions in this format on the internet these days? Every bot site phrases questions like this with the ? and it pisses me off
it's just esl's
Meta gaming ESL search queries + Indian and AI written slop article titles (not like there's much of a difference between the two).
Sorry I was not born in your country. Should I use "how can I ..." instead?
Exactly. You could help, instead of bullying me for ESL...
https://github.com/noctuid/gallery-dl-view works ok
>Sorry I was not born in your country. Should I use "how can I ..." instead?
Shit excuse, you can learn proper English outside of the US and England. moron.
Apparently I cannot.
>Exactly. You could help, instead of bullying me for ESL...
i wasn't bullying you, i was explaining to someone else why he sees questions formulated like that so often. if anything i was supporting your post by claiming it was likely a genuine question and not a robot
The real problem is that your question doesn't make sense at all. You don't need to be a grammar nazi to express yourself.
Maybe you should learn Japanese to prove your point.
>Maybe you should learn Japanese to prove your point.
日本語は知るよ
>The real problem is that your question doesn't make sense at all.
of course it does you autistic moron
>how to open all images from a webpage in a mpv as a playlist?
this is literally what he said. dissect how it DOESN'T make sense
>though your post is comprehensible, some people just see esl-like patterns and get angry over it
exactly
>Sorry I was not born in your country. Should I use "how can I ..." instead?
this is how i'd phrase it;
"how can i make a playlist in mpv of all the pictures from a web page?"
though your post is comprehensible, some people just see esl-like patterns and get angry over it whether it makes sense or not. i get it, i too wonder why you come here, to english websites, i never post on websites that use a language i don't know. is it just because english is more popular, so there's more available information?
thank you teacher
>, i too wonder why you come here, to english websites, i never post on websites that use a language i don't know. is it just because english is more popular, so there's more available information?
They think their fellow native language speakers are all subhuman morons so they come here to spew esl babble because they deserve to talk to smarter people.
anyone who unironically recommends display-resample in these threads should get gangraped. enjoy your video and audio quality reduction
video quality reduction how? audio may be resampled, but you can't hear any difference
you're literally adjusting video speed and compromising video quality. let me guess do you also like +4% PAL video speed-ups?
>you can't hear any difference
display-resample confirmed to be used by earlets.
i was suggesting display resample for 24/1.001 > 24, that's a 0.1% difference
you reckon you can recognise audio playing 0.1% faster than intended?
very easy to notice the pitch difference.
the pitch is adjusted/accounted for
it is not, dumbass. display-tempo does that but it degrades audio quality even more because the filter isn't designed for small changes in pitch.
why ewa lanczossharp it's so blurry
by design
slurping display-resample smooths here
My reds are orange
how
you are so 2023
when is ffmpeg getting an audio normalization filter similar to what YouTube and Spotify do? dynaudnorm is garbage and the audio loudness transitions are way too noticeable as scenes change
What's wrong with loudnorm?
low quality and feels tinny. i also heard that single-pass loudnorm in general is not ideal.
YouTube pre-measures loudness and stores it in metadata, so it's only a matter of applying an offset to normalize. I suppose Spotify does something similar.
I have been thinking about doing some Lua script that measures the loudness at runtime, but I'm not sure how to handle audio read-ahead.
At this point, I'm too lazy to try it, but maybe someone can take that idea and make it work.
Note that this concept would only work on locally stored files. Media over the network would only work if the stream contains any parsable loudness metadata, or if you cache the whole thing and measure it before starting playback.
Why does applying eq brightness/contrast/saturation filters increases the output filesize? Does the order of filters affect it, i.e. applying it before or after scaling?
The change in the profiles really caused the splineschizo to lose the few braincells he had didn't it?
endgame?
>vo=gpu-next
>gpu-api=vulkan
>hwdec=no
>profile=high-quality
>deband=no
>pixel clipper
>adaptative sharpen or another sharpening filter when watching low budget blurry ecchi anime
gpu-next and vulkan is meme.
>pixel clipper
you dont need it
cscale=ewa_lanczossharp
scale-antiring=0.6
>That's what I tried (--video-sync=display-resample --tscale=mitchell --interpolation=yes) and it didn't fix anything and made everything blurry to boot. Any other known remedies? Thanks though, good to know that's the actual keyword
I use this and happy with it.
video-sync=display-resample
tscale=sphinx
tscale-blur=0.65
does gpu-next still use ewa_lanczossharp for cscale? I can't see it anymore
also why no hwdec? and is vulkan better than vaapi?
why no hwdec? isn't graphics acceleration better?
GPU-next is trash. I only ever use it for HDR content and that's about it. Jitter, tearing, and other issues always pop up when I use GPU-next
Vulkan should generally have better GPU utilization than vaapi
I might be wrong about this, but i'm pretty sure mpv uses the GPU even without acceleration being enabled. You can get a little bit more use out of hardware decoding,
mpvisters, why does my anime look stuttery when it's panning sideways or upwards? Really takes the joy out of beautiful landscapes and the like. 240hz at 24 fps. Interpolation is gay but would probably help
ywnbaw
yeah that is common problem that is called judder, display-resample and tscale=sphinx would help
That's what I tried (--video-sync=display-resample --tscale=mitchell --interpolation=yes) and it didn't fix anything and made everything blurry to boot. Any other known remedies? Thanks though, good to know that's the actual keyword
And you wont ever be a fish. Or useful.
Omitting interpolation and sphinx instead of mitchell makes it slightly better but still headache inducing and unenjoyable. The stills look great though..
I recall it has only ever been satisfactory when I copied some riced out config that forced it to run at 240 fps when I was just starting out, but I'd rather avoid that
Thanks I'll try that next
I also use this script, hope u can fix the issue anon.
https://github.com/CogentRedTester/mpv-changerefresh
>windows only
Wont work for me sadly, but the tscale-blur seems to be the right track, thank you!
there is xrandr script too on gitlab or github.
Thanks for making me check.
https://gitlab.com/lvml/mpv-plugin-xrandr
Works as advertised I suppose (set refresh to 143.xxxxxxxxxx and reset it properly after quitting, sadly the judder is still there. I'll play with the blur as that was the biggest step in the right direction so far
This script and
video-sync=audio
tscale=sphinx
tscale-blur=0.65
are satisfactory. Any obvious idiocies committed with that? Thanks for the help, I will now enjoy
if you dont use display-* sync method interpolation wont work anon.
Idiocy counter: 1
I will make a commented note but leave the config as is. Didn't even enable interpolation either. Whichever component it is, it works now.
ok moron tscale wont do anything if u dont use display-resample kek
That's very cool and I'll add a commented note to that as well. However it works now. Guess the script is the heavy hitter than. During testing I had all kinds of things enabled or flagged at launch and I am not motivated to cut down the irrelevant options now.
is spynx better than oversample?
profile=high-quality
vo=gpu-next
gpu-api=vulkan
hwdec=vulkan
glsl-shader="~~/shaders/ravu-lite-ar-r4.hook"
scale-antiring=0.6
dscale=catmull_rom
tscale=oversample
tscale-radius=16.0
tscale-antiring=0.8
tscale-clamp=1.0
interpolation
video-sync=display-resample
sigmoid-upscaling
correct-downscaling
dither-depth=auto
are my settings good?
are my settings good?
kys tinkertrannoid
read the entire manual, the archives, test each and every setting and then you wouldn't need to ask this question
tscale-blur=0.6991556596428412
did you try?
Of course. Nothing less than the entire number space between 0 and 1
how much better are FOXCNN and FSRCNNX than RAVU? I want to cope and see what I'm missing out on
use dicky
what the frick is foxcnn? also the best shader is this
I meant ArtCNN sorry
https://artoriuz.github.io/blog/mpv_upscaling.html
How did you get fox from art lmao.
https://slow.pics/c/9NAf763N
damn artCNN looks much better than RAVU. sucks I can't handle it?
also what's the best RAVU version? I heard ravu-lite-ar-r4 is the best one followed by ravu-zoom-ar-r3
is ravu even good?
To be honest, if you can't handle the smaller artcnn you should probably not be running any shaders.
by smaller you bean the 16 version? I mean I run RAVU fine. what's the difference between the 16, 32, and 64 versions?
It's just the number of filters (which consequently also increases the size of the filters), higher = better but also slower.
use dicky its better and faster
I have never heard of dicky. how can I tell if it's better than FSRCNNX or ArtCNN? those at least had measurements
https://artoriuz.github.io/blog/mpv_upscaling.html
dicky has results too look result folder
so I tried dicky as everyone was saying and neither filter could do it at 60Hz. maybe when I get a 2nd RAM stick for dual channel things would improve a bit. but right now ravu-lite-ar-r4 is still the best
Don't fall for the meme.
what gpu do you have
if its really old maybe try 1x8 in fp16 directory
just an integrated Vega 6 (from the 4650U). single channel memory. what is fp16?
I mean I wouldn't buy a dedicated desktop PC with a 4090 just for meme filters, just seeing what my own hardware could do.
https://github.com/dickyplapper/dicky/tree/master/results
ArtCNN is deprecated, use dicky as everyone else is saying.
low-budget pastel-colored desynched blurry ecchi tranime
kys spaniard spammer
wtf is that overshapened shit?
agp ftm hrt
Do I have to use a toggle or are shaders enabled by default if I put in my conf
>glsl-shader="~~/mpv/shaders/dicky-6x16-DS.glsl"
what does the first number mean? I know the 2nd is # of filters.
1st is # of conv layers
these new numbers are weird to get used to
sex sex sex with mpv sisters
sex.conf
tummy pero pero
I haven't updated my mpv in 2 years. What has changed?
are chroma scalers and downscalers worth it? I use RAVU or upscaling, CFL for chroma, and SSimDownscaler for downscaling
No, the only chroma shader worth using is ArtCNN chroma but that's a massive resource hog for what you get: https://slow.pics/c/c2YnEJGo
I see. what does chroma even do? make the colors better? doesn't cscale do the same thing? I have ewa_lanczossharp (default for profile=high-quality and gpu-next)
This question makes no sense, the c in cscale IS chroma.
I see. so ewa_lanczossharp does a good enough job for cscale? and what's a good downscaler?
Anything works for dscale, the defaults are ok.
hermite is trash. use mitchell. both are blurry as shit but at least mitchell has no aliasing
I see. measurements say catmull with scale-anti ring (I set mine to 0.6) is best.
Catrom AR is and has always been a good choice, but this doesn't work on mpv currently without user shaders.
really? why not? because of dscale-antiring not working? scale-antiring works for both scale and dscale
and what is tscale for?
scale-antiring only affects downscaling if you're using a polar filter.
true, but a bug in gpu-next makes it so dscale-antiring doesn't work and scale-antiring affects both scale and dscale
dicky all in
Is this format new?
Is youtube serving flac quality audio now?
never seen it before
stats for nerds in the youtube player for that video doesn't go higher than 251 for me
why is display-vdrop smoother than audio? expain?
because display- is smoother
This doesn't really explain anything though. What's different about audio and display-vdrop's implementation that would make the latter smoother?
display-resample vs display-resample-vdrop. NOW
neither. just use display-vdrop. why would you ever use display-resample-vdrop? it's literally the worst of both worlds
>frame repeats AND fricked up video speed
I need interpolation
interpolation works fine on display-vdrop
Way heavier than display resample tho
The shaders you downloaded are probably the dp4a versions that require vo=gpu-next and gpu-api=vulkan. And if you're on old hardware they won't be faster anyway. Try the ones in the mpv/fp16 or results/fp16 directories instead.
>Try the ones in the mpv/fp16
that werked
oh I'm not using vulkan as my api yeah thanks that was it
??? display-vdrop is literally less CPU intensive since resampling isn't required.
any winblows users know if I can install something like Anki without it conflicting with my mpv install? it seems like Anki comes with its own mpv build
>anki comes with its own mpv build
???
it's statically built (like most windows programs), so what's the confusing part?
i didn't realize anki became so bloated.
anki has always had mpv as dependency though?
i bet you could make bank by repackaging mpv as india video player and pakistani video player
It's called Plex, and wm4 almost succeeded
Who will make the logo?
saved
Any recommendations for tscale?
oversample is alright, but some panning scenes still look pretty odd for my taste
dicky is not showing up for me on mpv for some reason?? every other shader including artcnn and ravu show up in the stats menu but not dicky, this is on the same source btw, what am I doing wrong?
Same here
meant for this
HELP
how do i get mpv like shortcuts on kodi? 1s rewinding while holding shift, being able to go frame by frame etc.
Or how do i use mpv as a media center. i want it to boot straight into it's interface and not require a mouse.
https://github.com/mpv-player/mpv/issues/13887
over it is
refresh_rate);
wl->current_output->make, wl->current_output->model,
wl->current_output->id, wl->scaling, wl->current_output->refresh_rate);
holy saarcode
why are we letting pajeet shitcode in mpv?
sirs that's just printing properties
doesn't make it not pajeetcode
hi llyyr
>merged
the absolute state of mpv in 2024...
so is vulkan decode usable and faster now or should I keep vaapi?
try it yourself
depends on your gpu/drivers, probably the same at best
https://github.com/mpv-player/mpv/issues/13889
plex staff ignoring issue templates? very unproffessional
wy dosen't
'vf toggle hflip'
work with gpu-next and vulkan?
Because gpu-next and vulkan are trash.
ban andarjeetwinux from the repo
Why?
reaction spamming and chatgpt tier responses
is oversample good for tscale?
kys spammer
>WHY EWA_LANCZOSSHARP IT'S SO BLURRY
WHY EWA_LANCZOSSHARP IT'S SO BLURRY
WHY EWA_LANCZOSSHARP IT'S SO BLURRY
and that's a good thing
NO WHY IS THAT A GOOD SHIT moron I WANT A CLEAR IMAGE
wtf is up with all the different dicky folders
magpie
mpv
results
which one do i use??!?!
kys spammer
are you using magpie? no.
are you using results? no.
are you using mpv? yes.
don't reply to esl spammer in genuine earnest
How do I edit video metadata
ffmpeg -metadata
How is dicky better than artcnn? Who is dickyplapper? Would you let him plap your dicky? Is it over mpvbros?
I'm switching to VLC. ^_^
>Who is dickyplapper?
me, I plap the dicky
scale=ewa_lanczossharp
why is my dicky blurry????
WHY EWA_LANCZOSSHARP IT'S SO BLURRY
WHY EWA_LANCZOSSHARP IT'S SO BLURRY
WHY EWA_LANCZOSSHARP IT'S SO BLURRY
epic discord raid sisters, lets go back to JETcord
Thanks dickyplapper
for breaking my config. Move your dicky not files in repo.
It was a necessary sacrifice to save me from saying the same thing 10 more times.
Ebr