[Ffmpeg-devel-irc] ffmpeg.log.20180407

[burek]

[00:02:03 CEST] <alonex> durandal_1707 interesting
[00:02:05 CEST] <alonex> codec
[00:02:17 CEST] <alonex> how it's hard for cpu ...
[00:04:16 CEST] <alonex> if u will code this it will be very cool
[00:04:43 CEST] <alonex> becouse i can't see alternative for Camstudio techsmith with ffmpeg
[00:04:44 CEST] <alonex> ;(
[00:05:37 CEST] <alonex> camstudio is ok, but.. almost200;
[00:05:40 CEST] <alonex> ;
[00:05:42 CEST] <alonex> $
[00:09:36 CEST] <alonex> http://www.infognition.com/ScreenPressor/speed.png
[00:10:06 CEST] <alonex> is that meaning same cpu usage as camstudio?
[00:10:12 CEST] <alonex> pk9,94ik90r4]\
[00:11:23 CEST] <alonex> well, any wayi guess if you will build encoder it would be GREAT!!!
[00:11:31 CEST] <alonex> thank u
[00:15:19 CEST] <alonex> my goal capturing 4 monitors with good cpu usage
[00:15:28 CEST] <alonex> in 4 files ;)
[00:15:44 CEST] <alonex> i dont know how it's in possible with camstudio
[00:15:48 CEST] <alonex> ;)
[00:15:56 CEST] <alonex> with ffmpeg not a problem..
[00:26:11 CEST] <blap> you guys know if anyone has done neural network upscaling of video / movies to higher res?  lik with https://www.wikihow.com/Use-Waifu2x
[00:28:12 CEST] <JEEB> FFmpeg has the classic NNEDI
[00:28:22 CEST] <JEEB> which bases on the vapoursynth version of nnedi3
[00:28:40 CEST] <JEEB> which as far as I remember is a much less filtering thing than waifu2x :P
[00:28:43 CEST] <JEEB> it just didn't have a funky name
[00:29:13 CEST] <JEEB> and the original nnedi3 is by now almost 10 years old https://forum.doom9.org/showthread.php?t=147695
[00:29:38 CEST] <JEEB> do note that you need the weights data for nnedi3 since that's pretty much the biggest part of it
[00:29:59 CEST] <JEEB> which you can download @ https://github.com/sekrit-twc/znedi3/blob/master/nnedi3_weights.bin
[00:29:59 CEST] <blap> ty, interesting
[00:32:32 CEST] <blap> is it worth playing with?
[00:32:40 CEST] <blap> i see it has a glsl shader too
[00:33:44 CEST] <JEEB> it is
[00:34:11 CEST] <blap> the sample pics i find are all bitrotted away
[00:34:56 CEST] <JEEB> if you think waifu2x is worth playing with, this stuff was what got used in slow deinterlacers etc
[00:36:13 CEST] <blap> got any samples of it at work?
[00:36:37 CEST] <blap> ah there's an online one
[00:36:50 CEST] <blap> http://semapho.re/nnedi3/
[00:40:11 CEST] <durandal_1707> nnedi is deinterlacer
[00:41:19 CEST] <JEEB> yes, but it just happens to be pretty good at 2x'ing things
[00:46:28 CEST] <JEEB> https://forum.doom9.org/showthread.php?t=129953
[00:46:33 CEST] <JEEB> and yes, even the original author noted it
[00:46:45 CEST] <JEEB> This filter turned out to be pretty good for resizing as well (limited to powers of 2 enlargement). Using it for resizing is pretty easy... pointresize the height to 2x, use nnedi, rotate left or right, pointresize again, use nnedi a second time, etc...
[00:46:57 CEST] <JEEB> (that's the original thread for NNEDI version 1)
[00:50:04 CEST] <blap> these guys seem to have soemthing http://www.infognition.com/VideoEnhancer/
[00:50:14 CEST] <blap> using more than one frame for analysis is a good idea
[00:50:57 CEST] <JEEB> yea, the fancier scripts utilizing NNEDI within themselves did pretty much that
[00:51:02 CEST] <JEEB> if you look at stuff like QTGMC
[00:53:14 CEST] <JEEB> although generally I've found is that you just want temporal stability, although I don't remember NNEDI3 having issues with that. EEDI2 on the other hand... (or at least I think that's what I used when a production company asked me for something stupid, and I was stupid enough to say "sure")
[00:53:29 CEST] <blap> :)
[00:54:01 CEST] <JEEB> pretty stills, but in motion you could see the temporal instability which wasn't visible in the source
[00:54:07 CEST] <blap> ah
[00:55:05 CEST] <JEEB> but yea, nnedi* also got heavily utilized in fancy deinterlacing filter chains :)
[00:55:08 CEST] <JEEB> http://avisynth.nl/index.php/QTGMC
[00:55:38 CEST] <JEEB> (I never looked at the source code of that one, but it did give alright results)
[00:58:33 CEST] <blap> thanks for getting me started
[00:59:12 CEST] <blap> i didn't realize deinterlacing was useful for this
[00:59:33 CEST] <JEEB> true interlacing happens to have half-height fields
[00:59:41 CEST] <JEEB> so true deinterlacing has to scale that 2x
[00:59:47 CEST] <JEEB> (with some quirks etc)
[01:00:47 CEST] <JEEB> so getting a good scaler there is important
[01:02:04 CEST] <JEEB> (interlacing is pretty much a container where you only know that there can be up to full rate of fields in there, but what's actually contained in the signal is not necessarily separate fields)
[01:02:11 CEST] <JEEB> which is why I talk of "true" interlacing
[01:02:21 CEST] <JEEB> as in, the content is actually interlaced within the signal
[01:45:56 CEST] <wfbarksdale> Does anyone know how, "ISO 14496 part 12: ISO base media file format" relates to "container types" ? is this the spec for any valid "mp4" container?
[01:47:11 CEST] <JEEB> 14496-12 is ISOBMFF, which is basically what you'd consider MP4, although I think there is 14496-14 which calls itself the "mp4 file format" and is a derivative of 14496-12
[01:47:37 CEST] <JEEB> generally it goes so that you have 14496-12, and then sub-specs that specify specific things in MP4
[01:47:54 CEST] <JEEB> like 14496-15 being NAL unit based video formats in 14496-12
[01:48:45 CEST] <JEEB> (also it seems like 14496-14 is effectively dead, since it was last published in 2003)
[01:49:19 CEST] <JEEB> so 14496-12 and 14496-15 and then whatever additional specs you need for whatever you need to put into mp4 should be it?
[01:50:00 CEST] <wfbarksdale> gotchya, that makes sense
[02:10:41 CEST] <wfbarksdale> I see " This specification defines MP4 as an instance of the ISO Media File format [ISO/IEC 14496-12
[02:11:42 CEST] <wfbarksdale> instance of is programmer word
[02:24:27 CEST] <furq> blap: there's already avs/vs stuff for using nnedi3 for po2 scaling
[02:24:35 CEST] <furq> look for nnedi3_rpow2
[02:25:04 CEST] <blap> ty
[02:25:29 CEST] <furq> also yeah qtgmc is amazing for sd stuff
[02:25:39 CEST] <blap> yeah i have sd stuff
[02:27:03 CEST] <blap> http://avisynth.nl/index.php/QTGMC  and http://forum.doom9.org/attachment.php?attachmentid=15838&d=1491145773   that is what i want for qtgmc for non-interlaced SD sources (linux)?
[02:27:59 CEST] <furq> https://forum.doom9.org/showthread.php?t=166582
[02:28:01 CEST] <furq> this is what you want on *nix
[02:28:12 CEST] <furq> avs is windows-only and also the multithreaded forks are very crash-prone
[02:28:17 CEST] <furq> and you definitely don't want to run this singlethreaded
[02:29:16 CEST] <JEEB> yeh, a lot of stuff got ports to vapoursynth, which is cross-platform
[02:29:22 CEST] <furq> qtgmc is no good for scaling btw
[02:29:48 CEST] <blap> what
[02:29:56 CEST] <furq> i meant it's amazing as a deinterlacer
[02:29:58 CEST] <JEEB> yea, it's a deinterlacer
[02:30:04 CEST] <furq> since that was mentioned
[02:30:07 CEST] <blap> my source is not interlaced
[02:30:12 CEST] <JEEB> then just plain nnedi3
[02:30:20 CEST] <furq> yeah you just want to find an implementation of nnedi_rpow2 then
[02:30:42 CEST] <JEEB> nnedi_rpow2 is rather simple (chroma offsets being the only thing that pops to mind)
[02:30:55 CEST] <furq> you could probably just find the script and port it to lavfi
[02:31:12 CEST] <JEEB> that didn't take long https://forum.doom9.org/showthread.php?t=172652
[02:32:20 CEST] <JEEB> https://gist.github.com/YamashitaRen/020c497524e794779d9c
[02:32:31 CEST] <JEEB> has all of the rpow2s
[02:32:42 CEST] <furq> nice, that looks more up to date
[02:33:07 CEST] <JEEB> yea, last updated in 2017
[02:33:44 CEST] <furq> https://github.com/HomeOfVapourSynthEvolution/VapourSynth-NNEDI3CL
[02:33:45 CEST] <furq> oh man
[02:34:05 CEST] <furq> this might turn out to be a big deal
[02:34:14 CEST] <JEEB> I think that's the thing which the zimg author then decided to troll back
[02:34:21 CEST] <JEEB> and make a faster CPU version
[02:34:41 CEST] <JEEB> https://github.com/sekrit-twc/znedi3
[02:34:47 CEST] <JEEB> I haven't tested, of course
[02:35:08 CEST] <furq> i don't think i have either of those
[02:36:40 CEST] <JEEB> but yea, twc added AVX512 and stuff
[04:43:58 CEST] <wfbarksdale> Whats the general perception of AV1 in the industry? I haven't heard of any hardware vendors supporting it... seems likely to fall to HEVC...
[04:52:37 CEST] <furq> it's not even out yet
[04:52:58 CEST] <furq> bye
[05:40:06 CEST] <fella> what does h264 mmco: unref short error mean?
[05:41:16 CEST] <fella> http://www.ffmpeg-archive.org/quot-mmco-unref-short-failure-quot-since-ffmpeg-x264-syntax-change-td3592555.html
[05:42:24 CEST] <fella> ^^ that seems to be almost exactly the same as with my encodings
[05:52:50 CEST] <blap>  https://youtu.be/AhT2mRwrNno  mechanical fourier analysis
[07:44:10 CEST] <Whiskey> i use ffmpeg to convert flac to mp3
[07:44:34 CEST] <Whiskey> i started yesterday at 01:30 time is now 07:45 and its around 2000 files and it still not done, is not thet very slow?
[07:45:10 CEST] <Whiskey> Its on an i7-9600k and M.2 disk
[07:48:51 CEST] <Foaly> depends on the file sizes
[07:49:07 CEST] <Whiskey> well you know size of FLAC
[07:49:28 CEST] <Whiskey> around 30-50MiB for every file
[07:50:36 CEST] <Foaly> so that's 100 gigs of input?
[07:50:55 CEST] <Whiskey> no
[07:50:59 CEST] <Foaly> dunno, possible that it takes that long
[07:51:00 CEST] <Whiskey> 50 :=)
[07:51:09 CEST] <Whiskey> ffmpeg -y -v quiet -i $SourceFullName -b:a 320k -map_metadata 0 -id3v2_version 3 -write_id3v1 1 $DestinationFullName is my settings
[07:51:26 CEST] <Foaly> why do you encode with 320k?
[07:51:32 CEST] <Whiskey> why not :D
[07:51:56 CEST] <Foaly> most people are not able to hear past 160k
[07:52:07 CEST] <Whiskey> yeh i know thats why spotify has it :=)
[07:53:00 CEST] <Foaly> does it tell you the encoding speed?
[07:56:03 CEST] <Whiskey> no
[07:56:26 CEST] <Whiskey> but the fun fact whit that is
[07:56:40 CEST] <Whiskey> if you compare FLAC YouTube Vevo and Spotify
[07:56:43 CEST] <Foaly> idk, but if it does not use all your cpu cores, you might want to run multiple encdings at once
[07:56:47 CEST] <Whiskey> ti gows in this order
[07:56:58 CEST] <Whiskey> spotify -> youtube -> flac
[07:57:03 CEST] <Whiskey> hear the diffrent is not hard at all
[07:57:24 CEST] <Whiskey> Foaly: how do you run multiply ?
[07:57:40 CEST] <Foaly> idk, how do you run it right now?
[07:57:50 CEST] <Foaly> and what operating system are you using?
[07:57:51 CEST] <Whiskey> from a powershell script
[07:57:55 CEST] <Whiskey> windows 10
[07:58:16 CEST] <Foaly> never used that, just split up your files and run multiple commandshells ^^
[07:58:28 CEST] <Whiskey> yeh thats one way :D
[07:58:45 CEST] <Whiskey> this is done soon anywayys :)
[07:59:07 CEST] <Whiskey> 250 to go :D
[07:59:32 CEST] <furq> if you wanted to save space you should have used opus or aac
[08:05:49 CEST] <Whiskey> but i still think its far to slow
[10:41:00 CEST] <alonex> durandal_1707,uhere?
[10:52:14 CEST] <durandal_1707> alonex: ?
[10:55:39 CEST] <alonex> can i ask about ScreenPressor?
[10:55:59 CEST] <alonex> how much this consuming CPU for encoding?
[10:59:39 CEST] <durandal_1707> alonex: dunno, try it, ver 2.0, not 3.0
[11:01:41 CEST] <furq> http://vpaste.net/Lrgb7
[11:01:44 CEST] <alonex> do u will really built encoder ScreenPressor for ffmpeg?
[11:01:44 CEST] <furq> this isn't confusing at all
[11:02:28 CEST] <kanacht> greets
[11:02:31 CEST] <kanacht> I wonder if it's possible to process 80's and 90's videos for better quality.
[11:02:34 CEST] <kanacht> applying modern image techniques, HDR, etc.
[11:03:06 CEST] <furq> that's a very vague question
[11:03:11 CEST] <kanacht> there are some common VHS aberrations
[11:03:34 CEST] <kanacht> furq: ok, it's possible or not?
[11:03:46 CEST] <furq> probably
[11:03:50 CEST] <kanacht> fix contrast, color, VHS aberrations, etc.
[11:03:54 CEST] <furq> it depends what you're actually trying to do as to whether it's actually going to be worth it
[11:04:50 CEST] <kanacht> furq: some aviation documentaries are nice, but horrible quality.
[11:05:08 CEST] <kanacht> the little engineering details can't be appreciated.
[11:05:42 CEST] <kanacht> compare those to today's 4k airplane landing for instance.
[11:08:03 CEST] <kanacht> furq: do you know what I am saying?
[11:16:58 CEST] <ChocolateArmpits> kanacht, you should probably look for better sources than VHS if you want to want to "enhance" anything.
[11:17:25 CEST] <ChocolateArmpits> even capturing VHS via firewire will give better quality than s-video capture
[11:18:27 CEST] <kanacht> hey, if they've managed to put color on old 1800 photos and ww1 and ww2 footage, enhancing vids from 80 and 90 should be perfectly doable.
[11:19:05 CEST] <ChocolateArmpits> well they had enough detail to work with.
[11:19:07 CEST] <kanacht> ChocolateArmpits: I didn't mean straight out of a physical vhs tape, but that type of quality on youtube vids.
[11:19:18 CEST] <kanacht> download the youtube vid, then process it.
[11:19:28 CEST] <ChocolateArmpits> vids from the 80s and 90s not delivered on master tapes don't
[11:19:43 CEST] <ChocolateArmpits> well that's even worse
[11:19:48 CEST] <ChocolateArmpits> like considerably worse
[11:22:30 CEST] <kanacht> why? nowadays there are  ryzens, powerful graphics card, clever enhancing techniques etc.
[11:23:00 CEST] <ChocolateArmpits> the premise is that there is nothing to enhance if there isn't anything. at best they'll put in fake detail
[11:23:18 CEST] <alonex> durandal_1707,screenpressor author is MSU author...
[11:23:38 CEST] <kanacht> ChocolateArmpits: upscale is a thing.
[11:23:43 CEST] <kanacht> 'upscaling'
[11:24:25 CEST] <ChocolateArmpits> if the player filter sucks then yes, you could upscale it for better reproduction, but if it doesn't, then why bother
[11:26:29 CEST] <alonex> kanacht,http://www.infognition.com/
[11:31:11 CEST] <durandal_1707> alonex: not for free
[11:32:20 CEST] <alonex> durandal_1707, what - not for free
[11:32:22 CEST] <alonex> ?
[11:33:19 CEST] <durandal_1707> alonex: i'm not charity
[11:33:30 CEST] <alonex> ah! i c.
[11:34:02 CEST] <alonex> well, interesting to hearing your price of course
[11:41:57 CEST] <durandal_1707> alonex: 1k eur
[11:42:43 CEST] <alonex> good honest price.
[11:57:57 CEST] <marcurling> Hello, (under Win10) I would like to have a log level as -v fatal but keeping the 'working' line which gives you frame, time... while compressing. Is this possible and how, please?
[11:58:48 CEST] <ChocolateArmpits> marcurling, add -stats
[11:59:36 CEST] Action: marcurling deeply thanks ChocolateArmpits 
[13:14:17 CEST] <Whiskey> Why does ffmpeg tells is Xing codac when i use liblame?
[13:16:00 CEST] <Whiskey> it happens on VRB on CBR its says its Lame
[13:30:42 CEST] <furq> ffmpeg doesn't write the xing vbr header
[13:31:00 CEST] <furq> or it does but it doesn't write the lame part of it
[13:55:16 CEST] <faxmodem> Hi! is it possible to read a mp4 while it is being written?
[13:55:40 CEST] <faxmodem> I'm dumping a VHS and would like to know if it's finished ;)
[13:55:55 CEST] <c_14> no
[13:56:13 CEST] <c_14> (unless you use fragmented moov)
[13:56:42 CEST] <furq> don't use mp4 for capturing in general
[13:56:43 CEST] <faxmodem> not even if I have a similar file (same options)?
[13:56:46 CEST] <faxmodem> ack
[13:56:51 CEST] <furq> if the input drops out then you'll be left with an unplayable file
[13:56:53 CEST] <faxmodem> what should I use?
[13:57:02 CEST] <furq> probably just mkv
[13:57:36 CEST] <furq> you can also play mkv while you're still recording to it
[13:57:49 CEST] <faxmodem> I've seen references of qt-faststart being able to "repair" files
[13:58:04 CEST] <faxmodem> furq: I'll definitely do that for the next ape :)
[13:58:08 CEST] <faxmodem> tape
[13:58:15 CEST] <furq> yeah it's not worth cancelling it if you're halfway through
[13:58:49 CEST] <faxmodem> I'll just wait for the 2 hours and then cut out the trailing white noise
[13:59:15 CEST] <faxmodem> hmmm wait I have a sound meter display on the VCR lol
[13:59:17 CEST] <furq> i also doubt qt-faststart can repair broken files, but then i've never used it
[13:59:31 CEST] <furq> afaik all it does is move the moov atom, but the broken file won't have a moov atom at all
[14:00:15 CEST] <ritsuka> faxmodem: if you make a fragmented mp4, you can read it while begin written
[14:00:24 CEST] <faxmodem> is there a good techincal reason that "moov atom" is at he end of the file?
[14:00:24 CEST] <ritsuka> I don't remember how to do it with ffmpeg
[14:00:41 CEST] <ritsuka> because you don't know how much space it will take
[14:00:53 CEST] <furq> you also don't know what you're putting in it until you're done encoding
[14:01:20 CEST] <faxmodem> so mkv puts that info periodically into the stream?
[14:02:06 CEST] <furq> i never bothered to find out what mkv actually does because it works
[14:02:17 CEST] <faxmodem> ^^
[14:02:23 CEST] <faxmodem> acpi
[14:02:26 CEST] <faxmodem> sorry :)
[14:02:32 CEST] <furq> something like mpegts will definitely periodically write those headers
[14:02:43 CEST] <faxmodem> yeah I'm grabbing VCR on battery so what? ;)
[14:03:22 CEST] <faxmodem> Discharging, 68%, 03:05:09 remaining, VHS is 120min so I'm fine ;)
[14:03:43 CEST] <faxmodem> furq: ritsuka c_14 thanks for your useful input!
[14:05:05 CEST] <c_14> recording to mp4 while on battery, playing with fire I see
[14:06:37 CEST] <faxmodem> ^^
[14:43:35 CEST] <aregidaniva> Hello, I'm having a very hard time figuring out how to blend a video and an image. I want to use a PNG as background and use the blend screen filter to put a video on top and remove black/dark colors. The two input have different color encoding (RGB/YUV), so I've used the `setsar` option to fix that, but the resulted output have messed up colors. My command line and ffmpeg output is available
[14:43:36 CEST] <aregidaniva> here: https://pastebin.com/8mNvsqkv
[14:52:01 CEST] <durandal_1707> aregidaniva: you need to force gbrp pixel format to both inputs
[14:53:51 CEST] <aregidaniva> durandal_1707: That what I can't manage to figure out. How should I do that?
[14:54:32 CEST] <furq> blend works here with one rgb and one yuv source
[14:54:45 CEST] <furq> i'm not really sure what the deal is with a png with a sar of 0
[14:55:46 CEST] <furq> oh nvm your paste doesn't match the problem
[14:56:25 CEST] <furq> aregidaniva: [0:v]format=gbrp[tmp];[1:v][tmp]blend
[14:57:31 CEST] <aregidaniva> Trying that furq
[15:02:45 CEST] <aregidaniva> Sorry furq, it doesn't work.
[15:03:24 CEST] <furq> are you sure you have the inputs to blend in the right order
[15:05:06 CEST] <aregidaniva> I think I am, but since I can't manage to have an output, I can't be sure. Still, when I had my messed up colors output it looked correct.
[15:06:38 CEST] <furq> oh right
[15:06:41 CEST] <furq> yeah add setsar back in
[15:06:48 CEST] <furq> or just save the png again because that sounds broken
[15:07:30 CEST] <aregidaniva> Really? It's coming straight out of GIMP!
[15:07:38 CEST] <aregidaniva> I'll try that.
[15:09:29 CEST] <furq> i didn't even know png had an aspect ratio
[15:09:38 CEST] <furq> apparently it does but if that's 0 then something is busted
[15:09:44 CEST] <furq> i can't imagine ffmpeg would be getting that wrong
[15:10:02 CEST] <aregidaniva> Alright, it works now furq. Thanks for the help, that's really appreciated! :)
[15:10:20 CEST] <aregidaniva> Ah, maybe. I'm not blaming ffmpeg.
[15:11:53 CEST] <furq> sure
[15:12:00 CEST] <furq> i'd be equally surprised about either really
[15:12:57 CEST] <furq> out of interest did you use setsar or just save the image again
[15:14:48 CEST] <aregidaniva> furq: I haven't saved the image again, I've used setsar.
[15:37:53 CEST] <blap> most old film source material is so blurry it doesn't seem to be very helpful to do fancier scaling than bilinear
[16:48:35 CEST] <kiroma> If av_register_all() is deprecated, what should I use?
[16:48:59 CEST] <JEEB> nothing, IIRC
[16:49:22 CEST] <JEEB> the whole registration thing was quite weird and you couldn't add external modules without *internal* symbols to begin with
[16:49:34 CEST] <JEEB> so instead of making everyone call a magical registration thing
[16:49:45 CEST] <JEEB> everything is already decided on the build level
[16:49:54 CEST] <kiroma> Oh alright
[16:49:55 CEST] <kiroma> Thanks
[16:50:06 CEST] <JEEB> I think the APIchanges document documents that
[16:50:34 CEST] <JEEB> http://git.videolan.org/?p=ffmpeg.git;a=blob;f=doc/APIchanges;hb=HEAD
[16:51:26 CEST] <JEEB> right, it doesn't note that there's no need to replace it with anything
[16:59:19 CEST] <blap> is there some linux video player that lets me display two different encodes of a film side by side, and fastforward/pause both simultaneously to compare results?
[17:00:23 CEST] <DHE> blap: you could use ffplay using complex filters to build the side-by-side. I don't think seeking will work though.
[17:00:47 CEST] <blap> this might be a useful thing to make
[17:01:13 CEST] <blap> seeking works in ffplay
[17:01:37 CEST] <DHE> but if the result isn't a file on disk but instead an algorithm building a sequence of frames, that might not work
[17:01:57 CEST] <DHE> you can build a lossless video, save it to disk, and then play that back. but that's waiting on the conversion and a possibly huge amount of disk space used
[17:03:49 CEST] <blap> i want to compare two encoded films of equal duration which already exist on disk
[17:05:20 CEST] <DHE> yes I know
[17:11:51 CEST] <blap> this should be a solved problem
[17:11:55 CEST] <blap> $current year
[17:13:20 CEST] <furq> blap: use hstack with mpv
[17:13:28 CEST] <furq> or pipe ffmpeg hstack into whatever player you want
[17:17:03 CEST] <blap> ty furq
[17:17:07 CEST] <blap> this works ffmpeg -i 1.m4v  -i 2.m4v -filter_complex vstack test.m4v
[17:17:19 CEST] <blap> but can i see the output live without outputting to test.m4v?
[17:19:39 CEST] <durandal_1707> blap: use mpv
[17:19:55 CEST] <furq> blap: -f yuv4mpegpipe - | mpv -
[17:20:10 CEST] <furq> or just use lavfi from mpv
[17:20:21 CEST] <blap> FANTASTIC!  TY
[17:20:55 CEST] <furq> someone really needs to change the default codec for nut to rawvideo so i can stop having to type yuv4mpegpipe
[17:21:02 CEST] <furq> tia
[17:21:05 CEST] <blap> this is SO helpful
[17:21:10 CEST] <blap> o.m.g
[17:21:12 CEST] <blap> ty
[17:31:56 CEST] <blap> https://files.catbox.moe/opmBsr.png  it helps SO much.
[17:34:15 CEST] Action: blap frolics
[17:42:59 CEST] <blap> https://files.catbox.moe/mqstkw.png  All hail the sharpen filter!
[21:10:05 CEST] <debianuser> Hello, I was reading docs about libfdk_aac, and noticed aac_he's "Spectral band replication" feature... If ffmpeg supports it when decoding aac_he... Is there some ffmpeg filter applying "spectral band replication" effect on its own? I'd love to test out how it sounds on some "dull" bad quality audio recordings.
[21:11:19 CEST] <JEEB> FFmpeg's decoder has HE-AAC support, no need for fdk-aac. And spectral band replication emulation would require the original signal
[21:11:45 CEST] <JEEB> at least if I understand what it was correctly
[21:14:16 CEST] <kiroma> Is there a 3bit pixel format?
[21:14:35 CEST] <furq> no
[21:14:44 CEST] <furq> ffmpeg -pix_fmts | grep 3$
[21:22:15 CEST] <debianuser> JEEB: Yeah, there's probably some additional coefficients saved to tell how exactly to "replicate" high-frequency spectral bands. What I was thinking about is a filter that does something like that to a regular sound, e.g. amplifies high-frequency harmonics based on the low-frequency ones by a coefficient passed in a command-line. Is there a filter like that in ffmpeg?
[21:22:35 CEST] <debianuser> That could make "dull" recordings sound nicer, I hope.
[22:05:17 CEST] <drathir> hi all...
[22:07:54 CEST] <drathir> guys there is a way using -map 0 exclude one stream which is Subtitle: dvb_teletext ?
[22:08:48 CEST] <drathir> or kind of -c:s no/skip/null ?>
[22:09:31 CEST] <ChocolateArmpits> drathir, use -sn
[22:10:08 CEST] <drathir> ChocolateArmpits: i see and is maybe a way per subs stream chosed it?
[22:12:10 CEST] <drathir> ChocolateArmpits: can confirm -sn working and drop all subs streams...
[22:12:24 CEST] <ChocolateArmpits> you want to choose which one to drop ?
[22:19:38 CEST] <drathir> ChocolateArmpits: yep if its even possible?
[22:20:28 CEST] <ChocolateArmpits> drathir, well by using the -map command of course !
[22:21:05 CEST] <ChocolateArmpits> map all input video and audio streams -map 0:v -map 0:a then use -map 0:s:X so select which subtitle streams you want to keep
[22:21:24 CEST] <drathir> oh that was aware sadly...
[22:21:37 CEST] <drathir> oh thats works too O.o
[22:22:44 CEST] <drathir> ChocolateArmpits: thanks a lot that much easier than every stream separatelly...
[22:22:52 CEST] <ChocolateArmpits> np
[22:29:07 CEST] <vivid> having trouble rebuilding ffmpeg from debian sources
[22:29:21 CEST] <vivid> i have enabled nonfree and nvenc and it seems to fail in the tests
[22:31:57 CEST] <vivid> just confused because i compiled it successfully on my last system, reinstalled the same release with kde instead of gnome and now i cant get it to build
[22:33:06 CEST] <Johnjay> vivid: i stopped being confused at things not building a long time ago
[22:33:32 CEST] <vivid> yeah thats good, want to point me in the direction of how to find why its failing
[22:34:14 CEST] <Johnjay> There's a compilation guide specifically for debian/ubuntu: https://trac.ffmpeg.org/wiki/CompilationGuide/Ubuntu
[22:34:30 CEST] <Johnjay> you might try checking to see if you have all of the required libs first. and try a basic compile without any codecs
[22:35:12 CEST] <vivid> i should have all the libs since i ran apt build-dep ffmpeg and apt source ffmpeg
[22:35:20 CEST] <Johnjay> i.e. something like PATH="$HOME/bin:$PATH" PKG_CONFIG_PATH="$HOME/ffmpeg_build/lib/pkgconfig" ./configure \   --prefix="$HOME/ffmpeg_build" --pkg-config-flags="--static" --extra-cflags="-I$HOME/ffmpeg_build/include" --extra-ldflags="-L$HOME/ffmpeg_build/lib" --extra-libs="-lpthread -lm"  --bindir="$HOME/bin"  --enable-gpl
[22:35:31 CEST] <vivid> should just compile, maybe i revert my tiny change and see if that changes anything
[22:35:36 CEST] <vivid> would be much more helpful to know why
[22:35:39 CEST] <Johnjay> the key word being "should" in that sentence
[22:35:41 CEST] <vivid> rather than just tyrying things
[22:36:07 CEST] <Johnjay> maybe apt failed to find the deps
[22:36:09 CEST] <Johnjay> i have no idea
[22:36:29 CEST] <Johnjay> This is linux, trying things to figure out why things don't work is the bread and butter of it.
[22:36:36 CEST] <vivid> no dpkg wouldnt start the build if they were not available, the only way that could happen is if the one header for nvenc isnt found
[22:36:55 CEST] <vivid> no reading the logs is a much better solution but i dont seem to find any of those
[22:37:15 CEST] <Johnjay> well that's all i can do, maybe ask someone in #debian on irc.oftc.net? they might know more about why dpkg failed
[22:37:43 CEST] <vivid> so first thigns first, would there be any reason for a sane install to fail tests
[22:37:49 CEST] <vivid> how about that, how do i disable the tests
[22:37:55 CEST] <vivid> completely has nothing to do with debian
[22:39:56 CEST] <Johnjay> i mean
[22:40:00 CEST] <Johnjay> do you know how to run the configure script?
[22:40:14 CEST] <Johnjay> when I did the other day it failed with missing libs a few times and i was able to fix it
[22:41:00 CEST] <vivid> do you know how to disable the tests?
[22:56:16 CEST] <c_14> vivid: pastebin your config.log
[22:56:28 CEST] <c_14> the error's usually at the very end of that
[22:56:53 CEST] <c_14> should be in ffbuild/config.log or so
[22:58:54 CEST] <vivid> no its diverted into the dpkg build log give me a sec
[22:59:06 CEST] <vivid> i see an obvious error but i just dont know why its happening
[23:00:40 CEST] <vivid> its too large to pastebin without "PRO"
[23:01:01 CEST] <c_14> just upload the last 50 or so lines
[23:01:09 CEST] <c_14> the error should be in there
[23:01:26 CEST] <vivid> https://pastebin.com/EEgmHb7U
[23:01:31 CEST] <vivid> you see the write error stdout..
[23:02:48 CEST] <c_14> wait, does the debian packager run fate?
[23:02:55 CEST] <vivid> fate?
[23:03:07 CEST] <c_14> ffmpeg automated testing environment or something
[23:03:15 CEST] <c_14> check debian/rules or whatever for a mention of `make fate'
[23:03:21 CEST] <c_14> or something with "fate" in it
[23:03:44 CEST] <c_14> if you don't care about those tests just remove or comment out the line executing it
[23:03:46 CEST] <vivid> no matches
[23:03:53 CEST] <vivid> i tried that and it still failed
[23:04:03 CEST] <vivid> overrided to nothing still same failure
[23:04:23 CEST] <vivid> what i dont get it i just compiled the same version, same package last week on the same release and it went fine
[23:04:33 CEST] <vivid> only difference is now im using kde instead of gnome3
[23:05:15 CEST] <vivid> write error, check disk all seems fine..
[23:05:16 CEST] <c_14> can you just comment out the override_dh_auto_test-arch section of debian/rules?
[23:05:31 CEST] <vivid> i can try but im pretty sure it will go back to the standard dh_auto_test
[23:06:04 CEST] <c_14> you can also just create an "empty one" that just executes /bin/true
[23:23:09 CEST] <vivid> oh nice nvEncodeAPI.h is included in ffmpeg source :p
[23:23:41 CEST] <JEEB> I'm pretty sure the reverse engineered/modified nvidia headers were removed
[23:24:01 CEST] <JEEB> because it undermined efforts of making other vendors play ball well
[23:24:21 CEST] <vivid> bam finally built thank you for the /bin/true help on that
[23:24:38 CEST] <JEEB> I think the nvidia headers are now in a separate FFmpeg repo
[23:24:39 CEST] <DHE> but it does mean you'll need to keep up with nvidia updates since only a specific version of nvenc will be supported by ffmpeg
[23:25:02 CEST] <JEEB> https://github.com/FFmpeg/nv-codec-headers
[23:25:07 CEST] <JEEB> https://git.videolan.org/?p=ffmpeg/nv-codec-headers.git
[23:25:14 CEST] <vivid> its still in the source package, compat/nvenc/nvEncodeAPI.h
[23:25:37 CEST] <JEEB> yea, no idea what debian is doing
[23:25:42 CEST] <vivid> no big deal it doesnt need to be there i have been providing it myself, just noticed it while i was going through
[23:25:58 CEST] <vivid> uh, that should be identical to the ffmpeg source tarball
[23:26:05 CEST] <vivid> for 3.4.2
[23:26:11 CEST] <JEEB> ok, then that was before it was removed
[23:26:16 CEST] <vivid> ahh
[23:26:35 CEST] <JEEB> basically other vendors just started going "just put our headers in, too"
[23:26:41 CEST] <vivid> good to know because after seeing that i wouldnt have retrieved and placed it in /usr/local/include
[23:26:44 CEST] <JEEB> and this being one of those vendors that's not an asshole about the SDK
[23:27:00 CEST] <JEEB> while nvidia is
[23:27:16 CEST] <JEEB> which is the only fucking reason someone reverse engineered and modified the header to begin with
[23:27:45 CEST] <vivid> yeah i would like to see that available in a more open sense so that nvenc can be built by default for debian
[23:28:27 CEST] <JEEB> but yea, now it's separate and no longer gets preferred treatment (intel was already doing the right thing with vaapi, and AMD is going to be doing the nice thing as well [going through properly packaged packages to depend on])
[23:28:59 CEST] <vivid> ok, thanks for the help c_14 it is now time for a fucking cerveza
[00:00:00 CEST] --- Sun Apr  8 2018