[Ffmpeg-devel-irc] ffmpeg.log.20180915

[burek]

[00:16:52 CEST] <A1den> Hi every one
[00:17:21 CEST] <A1den> I had asked earlier how to use the --headers parameter via the ffmpeg api
[00:18:16 CEST] <A1den> does anyone know how to set headers/cookies via code
[00:18:34 CEST] <A1den> I could use some help ?
[01:01:13 CEST] <poutine> I'm looking over code changes which appear to preserve SEI data in x264 when encoding, it calls them "a53 captions" in the code, is that some standard nomenclature that I'm just not finding anywhere?
[01:11:10 CEST] <transhumanist> anyone know how to fix this error Unknown input format 'dshow' cant find any solution https://pastebin.com/1ux5yJ1U
[01:11:14 CEST] <transhumanist> thanks in advance
[01:18:14 CEST] <c_14> transhumanist: there's no dshow on linux
[01:18:41 CEST] <c_14> you probably want v4l2 or x11grab or eeeh
[01:18:43 CEST] <c_14> that other one
[01:18:45 CEST] <c_14> kmsgrab
[02:00:57 CEST] <transhumanist> thanks c_14
[02:01:12 CEST] <transhumanist> I seem to have fixed the problem by updating ffmpeg
[02:02:15 CEST] <transhumanist> anyone know where i might find video of rooms in a house (Like a house interior scan so I can use it as an example?)
[02:02:47 CEST] <transhumanist> trying out software that works with opencv ffmpeg and a few other packages for rending 3d model of a home
[02:03:08 CEST] <transhumanist> its a sort of side question but thought  someone might have a quick answer
[02:03:27 CEST] <transhumanist> my camera really sucks so I cant use it to render the rooms unfortunately
[02:05:28 CEST] <transhumanist> actually I will ask in machine-learning group probably more appropriate
[07:41:02 CEST] <johnnny22> is there a way to make the demuxer be a bit faster than the muxer ? As in, currently, if i provide a 60fps source and 30fps output, 30 frames per seconds will get dropped and not fed to the muxer (if not mistaken). Is there a way to make the muxer drop frames and not the middle man?
[07:42:58 CEST] <johnnny22> ideally, i'd like the full 60 frames to reach the muxer and have it do the job of dropping frames or not based on it's personal state.
[08:18:51 CEST] <fling> johnnny22: no
[08:46:56 CEST] <johnnny22> fling: the issue i'm having is that the 'device' seems to consume the frames & samples I provide it a bit faster than the 30fps that the device is set to. I'm thinking maybe that in this case, the device will need to duplicate a frame or some samples once in a while (not sure about the approach to take on the audio samples though)
[08:48:48 CEST] <fling> johnnny22: there is an fps filter you could use to set the exact frame rate you need
[08:49:00 CEST] <fling> johnnny22: -vf fps=30
[08:50:49 CEST] <johnnny22> i tried the setpts one, but not sure it's really working :o/
[08:52:04 CEST] <johnnny22> in fact, i'm still testing it
[08:55:01 CEST] <johnnny22> fling: how is that different than setting the source at 30fps using -r 30 or 30000/1000 ?
[08:59:17 CEST] <johnnny22> also, i don't really want to speedup or slowdown the content itself.
[09:00:21 CEST] <johnnny22> clearly i'll have to stretch it a bit (but the factor is way small) it's like 0.5 seconds per 12h
[09:38:03 CEST] <johnnny22> setpts be of some help after all (guessing fps would also do the same).. to be continued.
[09:52:53 CEST] <A1den>  I had earlier asked if it's possible to set ffmpeg cookies and useragent via the api but there was no response. I couldn't find any information about this so I suppose nobody is aware of this and now I should create an issue on the github repo ?
[10:37:04 CEST] <johnnny22> A1den: something similar to "-headers " ?
[10:37:27 CEST] <A1den> yeah
[10:37:47 CEST] <A1den> I've been trying to look into the cmdline options but I can't see that option in the source code anywhere
[11:32:48 CEST] <johnnny22> A1den_: sorry, got no clue. maybe check in the hls demuxer !?
[11:33:40 CEST] <johnnny22> A1den_: maybe this might help : https://ffmpeg.org/ffmpeg-protocols.html#http
[11:33:57 CEST] <johnnny22> just for inspiration
[12:27:21 CEST] <A1den> johnnny22:Thanks let me check it out.
[12:29:30 CEST] <A1den> I think creating an issue is relevant to this since it's a lack of documentation
[14:49:39 CEST] <johnny56> try to build ffmpeg with libkvazaar fails with pkg-config saying it cannot locate kvazaar >= 0.8.1, doing pkg-config --modversion kvazaar returns 1.2.0. looking at the config.log I can see errors: https://hastebin.com/zikicosumo
[14:51:02 CEST] <c_14> is libcryptopp.so.6 in your ldpath?
[14:51:33 CEST] <c_14> and what's the output of pkg-config --libs --cflags kvazaar
[14:52:02 CEST] <johnny56> my kvazaar was built with libcryptopp.so.7
[14:52:12 CEST] <johnny56> so i don't know why it keeps looking for libcryptopp.so.6
[14:52:45 CEST] <c_14> did you build kvazaar yourself?
[14:53:01 CEST] <johnny56> yes, the output for your cmd was: -lkvazaar
[14:53:15 CEST] <c_14> Because it looks like it's trying to use a version from your package system
[14:53:48 CEST] <johnny56> there is no version 6 present on my box though
[14:54:16 CEST] <johnny56> and my kvazaar was built against version6..
[14:54:22 CEST] <johnny56> and my kvazaar was built against version7*..
[14:54:39 CEST] <c_14> Did you install kvazaar into /usr?
[14:55:14 CEST] <johnny56> it's in /opt/kvazaar
[14:55:35 CEST] <c_14> apparently you also have libkvazaar in /usr/lib and pkg-config is picking up that one
[14:55:47 CEST] <c_14> did you set PKG_CONFIG_PATH=/opt/kvazaar/lib/pkgconfig ?
[14:56:24 CEST] <c_14> (and if you're linking dynamically and not setting rpath, this'll cause issues if /opt/kvazaar/lib isn't before /usr in your ld.so.conf)
[14:56:54 CEST] <johnny56> I have /opt/kvazaar/lib in a custom /etc/ld.so.conf.d/opt.conf file
[14:57:02 CEST] <johnny56> and just updated /usr/lib/pkgconfig/kvazaar.pc
[14:57:08 CEST] <johnny56> no change
[14:57:29 CEST] <c_14> did you set PKG_CONFIG_PATH?
[14:57:40 CEST] <johnny56> nope
[14:57:43 CEST] <c_14> you have to
[14:57:49 CEST] <johnny56> too ?
[14:57:57 CEST] <JEEB> it's for pkg-config, not runtime loading
[14:58:12 CEST] <JEEB> /opt/kvazaar probably isn't in your pkg-config search path by default :P
[14:58:13 CEST] <johnny56> I changed the default though
[14:58:23 CEST] <johnny56> /usr/lib/pkgconfig/kvazaar.pc
[14:58:31 CEST] <johnny56> still ned do set PKG_CONFIG_PATH
[14:58:32 CEST] <johnny56> ?
[14:58:45 CEST] <JEEB> check pkg-config --libs kvazaar
[14:59:04 CEST] <johnny56> -L/opt/lib -lkvazaar
[14:59:34 CEST] <johnny56> ops
[14:59:59 CEST] <JEEB> ok, then try clearing out your FFmpeg build dir and re-configuring it
[15:00:14 CEST] <johnny56> it's suppose to be /opt/kvazaar/lib not /opt/lib
[15:00:21 CEST] <johnny56> it's fixed now tyty
[15:01:16 CEST] <JEEB> ah, yea that makes sense if you've manually changed the stuff
[16:55:29 CEST] <Hello71> how do I ask ffprobe to give me the bitrate even if it requires demuxing?
[16:55:40 CEST] <Hello71> per stream
[16:56:14 CEST] <JEEB> you'd have to calculate it based on the avpackets and duration
[17:09:25 CEST] <Hello71> :q
[17:10:45 CEST] <JEEB> also it all depends on what level you want your bit rate
[17:11:22 CEST] <JEEB> do you want to calculate vbv/hrd compliance, per packet bit rate or just an average over the whole runtimd
[17:27:36 CEST] <Hello71> average bitrate
[17:28:31 CEST] <JEEB> then parse the size of all avpackets, then look up the final length of that track
[17:42:34 CEST] <Hello71> ffprobe -show_packets -of compact | while read ...?
[17:56:58 CEST] <Bombo> Hello71: like this? ffprobe -select_streams v:0 -show_entries format=bit_rate -v error
[18:10:37 CEST] <Bombo> whats the equivalent in -filter_complex for '-vf reverse -af areverse' ? i got -filter_complex "[0:v]reverse" to reverse the video but "[0:v]reverse [0:a]areverse" gives me Unable to parse graph description substring: "areverse"
[18:11:03 CEST] <furq> you need ; between filters
[18:13:07 CEST] <Hello71> Bombo: one, I want per stream info, and two, that doesn't work if the file is dumb
[18:13:08 CEST] <Bombo> aaah
[18:13:47 CEST] <Hello71> s/dumb/bad/
[18:14:11 CEST] <Bombo> Hello71: per stream? like video then audio?
[18:14:49 CEST] <Hello71> no, but that's fine too
[18:15:11 CEST] <Bombo> or pre packet
[18:15:54 CEST] <Bombo> you got more than one video stream in one container?
[18:16:14 CEST] <Hello71> it is common to have more than one audio stream
[18:16:24 CEST] <Hello71> and very common to have more than one subtitle stream
[18:18:20 CEST] <Hello71> I was just hoping there was a switch for "ignore the stupid metadata and figure it out yourself"
[18:20:37 CEST] <Bombo> Hello71: for audio streams: -select_streams a -show_entries stream=bit_rate -v error
[18:22:22 CEST] <kepstin> Hello71: ffprobe does not contain any code to calculate per-stream bitrate if the information is not available from the format. There may be other tools which can do that.
[18:29:00 CEST] <Bombo> but per packet? -show_packets has size= isn't that the bits of a packet?
[18:29:08 CEST] <Bombo> or byte
[18:29:56 CEST] <kepstin> sure, but if you want the average stream bitrate like Hello71 does, then you have to add them up yourself.
[18:30:05 CEST] <kepstin> there's no option to do that.
[18:30:21 CEST] <Hello71> patches accepted?
[18:31:32 CEST] <kepstin> ... maybe? There already is a -count_frames option which does something kinda similar.
[21:51:11 CEST] <transhumanist> Hi I have an AVI file I am trying to extract all frames without loss. Found several commands for doing it but output frames are pixilated anyone able to help?
[21:51:34 CEST] <transhumanist> would prefer a .jpg output
[21:52:21 CEST] <EvanR> isn't .jpg necessarily lossy
[21:52:25 CEST] <kurufu> ^
[21:52:36 CEST] <transhumanist> hmm maybe thats the  problem then
[21:53:06 CEST] <transhumanist> should i just try png images?
[21:53:28 CEST] <nicolas17> not only jpeg is lossy
[21:53:37 CEST] <nicolas17> I think ffmpeg defaults to somewhat low JPEG quality
[21:53:58 CEST] <transhumanist> I see, so how do I specify the highest quality
[21:54:06 CEST] <transhumanist> its really really bad on the ouput
[21:54:09 CEST] <transhumanist> output
[21:54:22 CEST] <nicolas17> even highest quality will have loss, JPEG *is* a lossy format, and you said "without loss" :)
[21:54:49 CEST] <transhumanist> ok how about the best possible jpeg quality (forget lossy)
[21:54:50 CEST] <A1den_> I mean e.g set playback_complete=bool at the end
[21:55:20 CEST] <A1den_> i'm considering using the sources directly https://github.com/FFmpeg/FFmpeg/blob/master/fftools/ffplay.c
[21:55:25 CEST] <kurufu> If anyone happens to know, does ffmpeg impose h264 level limits outside of what the encoder imposes via this setting? e.g. inside vaapi_encode.c it appears level is entirely ignored, but i dont know the codebase well enough.
[21:55:25 CEST] <A1den_> without sdl
[21:56:15 CEST] <transhumanist> nm figured it out
[21:56:24 CEST] <nicolas17> "Use -qscale:v (or the alias -q:v) as an output option. Effective range for JPEG is 2-31 with 31 being the worst quality. I recommend trying values of 2-5."
[21:56:45 CEST] <nicolas17> transhumanist: what video codec does your AVI file have?
[21:57:43 CEST] <transhumanist> ffmpeg -i Qtcam-18_09_15:15_17_50.avi -qscale:v 2 LivingRoom-1_%03d.jpg
[21:57:47 CEST] <transhumanist> works great
[21:57:52 CEST] <A1den_> after getting rid of sdl, the viddeo endsup iterating on the last seconds without ending
[21:57:56 CEST] <transhumanist> -qscale 2 is the part
[21:58:02 CEST] <transhumanist> thanks guys
[21:58:27 CEST] <nicolas17> transhumanist: what video codec is in the AVI? if it happens to be MJPEG then you can do actual lossless
[21:58:28 CEST] <kurufu> People are still recording to avi in 2018. What a time to be alive.
[21:59:02 CEST] <transhumanist> oh really I will have to look up how to extract that info from the file
[21:59:12 CEST] <nicolas17> you should see it in the ffmpeg output
[21:59:50 CEST] <transhumanist> it is MJPEG
[22:00:27 CEST] <transhumanist> but even so its still slightly blurry with -qscale:v 2
[22:00:29 CEST] <nicolas17> if your input is MJPEG, you can extract the frames unmodified, instead of decoding them and re-encoding them as JPEG
[22:00:40 CEST] <nicolas17> ffmpeg -i input.avi -codec:v copy -bsf:v mjpeg2jpeg output_%03d.jpg
[22:00:45 CEST] <transhumanist> thats great thanks!
[22:00:55 CEST] <A1den_> here is my code https://gist.github.com/kenkit/1667810d121a33b71b7e4f56c8dd3f71
[22:00:55 CEST] <transhumanist> wow thats one to write down for sure !
[22:00:57 CEST] <nicolas17> because the fra,es are *already* independent JPEGs
[22:01:42 CEST] <nicolas17> transhumanist: bookmark https://stackoverflow.com/questions/10225403/ :)
[22:01:53 CEST] <A1den_> I just want to figure out when to stop and delete the thead, any help will be greately appreciated
[22:02:05 CEST] <A1den_> *where
[22:02:36 CEST] <nicolas17> A1den_: perhaps #ffmpeg-devel is more appropriate, or perhaps this is just a bad day/timezone to find people who know that :)
[22:03:33 CEST] <A1den_> if anyone knows where I can put the variable please htmu
[22:03:59 CEST] <GuiToris> hey, daala hasn't been released yet, has it?
[22:04:33 CEST] <transhumanist> thanks a lot guys
[22:04:53 CEST] <furq> A1den_: i think most of their devs are working on av1 now
[22:06:20 CEST] <A1den_> ooh, It's a simple problem
[22:06:36 CEST] <A1den_> I only got rid of sdl as the main event manager
[22:06:41 CEST] <furq> er
[22:06:42 CEST] <furq> GuiToris:
[22:06:47 CEST] <transhumanist> its actually amazing how good the quality is for a .99 cent camera
[22:06:56 CEST] <A1den_> I modified  refresh_loop_wait_event() and event_loop()
[22:07:33 CEST] <A1den_> I can't figure how to determine if playback completed so I can terminate the thread
[22:07:34 CEST] <transhumanist> doesn't deal well with motion though blurs up real quick, I assume this has more to do with the number of frames per second but maybe the photo sensitivity?
[22:07:52 CEST] <GuiToris> furq, hmm I haven't even heard this codec :O
[22:08:31 CEST] <GuiToris> I used to use theora because I wanted to be fully free, but ehh, it isn't worth it
[22:08:46 CEST] <furq> vp9 is free enough
[22:09:32 CEST] <furq> if you want to actually do stuff today, obviously
[22:09:32 CEST] <nicolas17> .99 cent means slightly less than 1 cent ;)
[22:09:34 CEST] <GuiToris> furq, I know, I know but something isn't okay with it ... would you save your family videos with a vp8-9 codec?
[22:09:49 CEST] <furq> well no i'd just use x264 because i'm in the eu
[22:10:26 CEST] <GuiToris> yes, I'm using x264 and x265 nowadays too
[22:10:39 CEST] <GuiToris> the quality is much better and the file size is smaller
[22:10:43 CEST] <furq> i assume you're doing video/film transfers
[22:11:14 CEST] <GuiToris> I edit home videos
[22:11:19 CEST] <GuiToris> mainly family stuff
[22:11:29 CEST] <furq> oh right
[22:11:40 CEST] <furq> well yeah i'd just keep an archival copy if possible so i don't have to worry about what codec i use
[22:12:57 CEST] <GuiToris> vp8 and 9 are streaming codecs, aren't they?
[22:13:09 CEST] <A1den_> I guess I still have a long way to go.....
[22:13:50 CEST] <nicolas17> define 'streaming codec'
[22:15:25 CEST] <GuiToris> broadcasting codecs, if you go online you use them, or they may be suitable for video calling too
[22:15:43 CEST] <GuiToris> I shouldn't store webm files on my computer, should I?
[22:15:48 CEST] <nicolas17> why not
[22:16:28 CEST] <GuiToris> people are obsessed with x264, I don't know why
[22:16:37 CEST] <GuiToris> I should probably go with the flow
[22:16:44 CEST] <GuiToris> theora was a bad decision
[22:16:50 CEST] <furq> h264 is decodable everywhere and x264 is very mature
[22:16:50 CEST] <GuiToris> daala may be promising
[22:17:03 CEST] <GuiToris> what's wrong with h265?
[22:17:07 CEST] <furq> hevc and vp9 are technically better than h264
[22:17:10 CEST] <furq> but the encoders aren't as mature
[22:17:19 CEST] <furq> libvpx in particular is really slow
[22:17:32 CEST] <furq> i guess people just don't really trust them yet
[22:17:44 CEST] <furq> although x265 is becoming pretty standard for >1080p
[22:18:33 CEST] <furq> the other issue with hevc is the insane licensing situation
[22:18:47 CEST] <GuiToris> furq, do you recommend x264+vorbis in an mp4 container?
[22:18:57 CEST] <kurufu> because x265 isnt even that much better than x264 at 1080p and below. especially when you consider how much more expensive it is.
[22:19:04 CEST] <furq> you can't put vorbis in mp4
[22:19:19 CEST] <furq> x264 and aac is the normal combination
[22:19:31 CEST] <furq> but i generally wouldn't reencode the audio
[22:19:50 CEST] <furq> vorbis is superseded by opus these days anyway
[22:19:57 CEST] <furq> which iirc you can mux into mp4 but it's not well supported yet
[22:20:07 CEST] <GuiToris> you must be right, I use mkv, it's free. mkv allows me everything
[22:20:48 CEST] <GuiToris> I'll give it a try, I think you can use vorbis ...
[22:20:50 CEST] <johnnny22> durandal_1707: your alsa set channel made it in  :D thanks again
[22:21:07 CEST] <furq> you can use vorbis in mkv yeah
[22:21:11 CEST] <furq> but like i said you should use opus
[22:21:18 CEST] <GuiToris> I meant in mp4
[22:21:30 CEST] <GuiToris> I think I've done this, but I'm not sure at all
[22:22:52 CEST] <GuiToris> ffmpeg has just created one
[22:23:03 CEST] <GuiToris> -c:a libvorbis output.mp4
[22:23:09 CEST] <johnnny22> durandal_1707: oups, i mean pulse fix
[22:24:23 CEST] <furq> GuiToris: weird
[22:24:25 CEST] <furq> http://mp4ra.org/#/codecs
[22:24:27 CEST] <furq> it's not listed here
[22:24:41 CEST] <furq> i wouldn't be confident that anything will play that file back
[22:26:46 CEST] <GuiToris> vlc doesn't complain
[22:28:25 CEST] <GuiToris> if the content is very important, should I stay with x264+acc and mp4
[22:28:37 CEST] <GuiToris> I remember using opus once, and it was quite good
[22:32:41 CEST] <GuiToris> I don't really know what's the downside of these codecs and containers! Will the encoders disappear sometime? Will my videos be corrupted over time?
[22:33:32 CEST] <tdr> corrupted, like the bitrot prefers the tastier formats?
[22:33:34 CEST] <durandal_1707> store everything uncompressed, and watch how bits change :)
[22:34:26 CEST] <GuiToris> durandal_1707, I'm using crf 18 now, and trying to figure out what to do
[22:35:02 CEST] <GuiToris> according to the descriptions, 18 is almost lossless
[22:35:57 CEST] <durandal_1707> without backups of backups of backups of backups your videos are cursed
[22:36:36 CEST] <GuiToris> I should have backups!
[22:37:32 CEST] <tdr> store them on punchcards, they decompose pretty slow
[22:46:09 CEST] <johnnny22> lmao
[22:52:28 CEST] <johnnny22> durandal_1707: i managed to add support for AV_PIX_FMT_BGR0, though maybe it should be actually set to BGRA to be more precise for the 'decklink' output device. I haven't really tested the resuling video on a screen, but I'm hopeful it works out nicely. I did this since sometimes the source might be RGB based and it seems to reduce the CPU usage by 50%.
[22:54:55 CEST] <johnnny22> Once confirmed that the output is right, I'll submit a patch for it.
[22:55:57 CEST] <Bombo> can anyone explain what this means? -filter_complex "[0:v]reverse,fifo[r];[0:v][0:a][r] [0:a]concat=n=2:v=1:a=1 [v] [a]" -map "[v]" -map "[a]"
[22:56:10 CEST] <Bombo> what is [r] read?
[22:56:56 CEST] <ahoo> rate
[22:57:00 CEST] <ahoo> (frame)rate
[22:57:10 CEST] <Bombo> hmm
[22:57:19 CEST] <ahoo> >usually<
[22:57:32 CEST] <nicolas17> errr no
[22:57:36 CEST] <nicolas17> that's just a stream name
[22:58:25 CEST] <durandal_1707> johnnny22: ok
[22:58:46 CEST] <Bombo> sooo the vid stream goes to the fifo named r, that is reverse
[22:59:06 CEST] <Bombo> then after the ; [0:v][0:a][r]
[22:59:52 CEST] <johnnny22> durandal_1707: should I create a ticket or just post on the ffmpeg-devel mailing list or create a ticket and point it out on the ffmpeg-devel mailing list ?
[23:00:25 CEST] <Bombo> 'there is a vid then aud then the reverse vid from fifo stream'?
[23:00:49 CEST] <durandal_1707> johnnny22: post it to list
[23:00:49 CEST] <Bombo> its too complex for me ;)
[23:00:52 CEST] <Bombo> ;/
[23:01:21 CEST] <ahoo> 🏻
[23:01:30 CEST] <ahoo> it's too complex for me, too.
[23:01:36 CEST] <ahoo> don't feel ashamed.
[23:01:48 CEST] <Bombo> ok ;)
[23:01:54 CEST] <ahoo> ;)
[23:01:56 CEST] <johnnny22> Bombo: i'm not sure you need that fifo
[23:01:57 CEST] <Bombo> what i want to do is -filter_complex "[0:v]reverse;[0:a]areverse"
[23:02:04 CEST] <johnnny22> Bombo: but I'm not sure sure.
[23:02:08 CEST] <Bombo> this reverses a + v
[23:02:32 CEST] <durandal_1707> why you want reverse?
[23:02:34 CEST] <Bombo> but i want to concat the reversed av after the orig forward av
[23:02:56 CEST] <Bombo> because its hip ;)
[23:02:57 CEST] <johnnny22> simple, framerate = -30 :P (just kidding)
[23:03:19 CEST] <durandal_1707> Bombo: how long it is?
[23:03:21 CEST] <Bombo> i always see those vid loops one fwd then bwd
[23:03:44 CEST] <nicolas17> I'm trying to pass that to graph2dot but I can't figure out the nullsrc's to make it work
[23:03:48 CEST] <nicolas17> Input pad "in0:v0" with type video of the filter instance "Parsed_concat_4" of concat not connected to any source
[23:03:57 CEST] <Bombo> durandal_1707: a second?
[23:05:07 CEST] <nicolas17> I think your problem is that you're using the same stream multiple times
[23:05:34 CEST] <nicolas17> you probably need the 'split' filter
[23:06:01 CEST] <Bombo> that complex complex i got first, does work, just not for the audio btw
[23:06:30 CEST] <Bombo> audio is just forward when the vid is bw
[23:07:34 CEST] <nicolas17> that doesn't even parse for me until I add several splits
[23:08:21 CEST] <Bombo> split what a and v?
[23:08:38 CEST] <nicolas17> no, split v into two streams so it can be consumed by both concat and reverse
[23:09:05 CEST] <nicolas17> it's not letting me connect [0:v] in two places
[23:09:29 CEST] <Bombo> so split like copy
[23:09:32 CEST] <Bombo> dup
[23:09:33 CEST] <Bombo> ;)
[23:09:39 CEST] <nicolas17> yes
[23:09:43 CEST] <Bombo> hm maybe the fifo does that
[23:09:54 CEST] <durandal_1707> fifo is not needed
[23:10:10 CEST] <durandal_1707> -lavfi "[0:v]split[v1][v2];[1:a]asplit[a1][a2],[a2]areverse[ar],[v2]reverse[r],[v1][a1][r][ar]concat=n=2:v=1:a=1"
[23:10:50 CEST] <johnnny22> looks about right
[23:11:17 CEST] <durandal_1707> if audio is in same file as video use [0:a]
[23:11:19 CEST] <nicolas17> fifo does nothing there
[23:13:50 CEST] <Bombo> durandal_1707: yes! thx :)
[23:14:42 CEST] <Bombo> so [ar] is just a name
[23:15:36 CEST] <durandal_1707> it is pad name, it was  shorter to type then [audio_reversed]
[23:16:25 CEST] <Bombo> ok i almost understand ;)
[23:16:35 CEST] <durandal_1707> make sure clips are very short and that you have enough RAM
[23:16:46 CEST] <durandal_1707> when using (a)reverse
[23:17:03 CEST] <Bombo> why are there , not ;
[23:17:18 CEST] <Bombo> after the asplit
[23:18:12 CEST] <Bombo> do this filter; do that filter, what now
[23:18:27 CEST] <nicolas17> foo,bar is the same as foo[pad]; [pad]bar
[23:18:33 CEST] <nicolas17> it connects the output of foo into the input of bar
[23:19:02 CEST] <Bombo> yes this 'effect' is just for small fun clips, i just saw such things on instagram :)
[23:19:49 CEST] <Bombo> 1 sec or some, cat jumping down the table, then backwards up ;)
[23:19:56 CEST] <durandal_1707> http://ffmpeg.org/ffmpeg-filters.html#toc-Filtergraph-description
[23:20:15 CEST] <Bombo> i just wondered if its possible with ffmpeg too
[23:21:40 CEST] <Bombo> hmmkay chain,chain
[23:24:32 CEST] <Bombo> i think i got it now thx again guys! :)
[00:00:00 CEST] --- Sun Sep 16 2018