[Ffmpeg-devel-irc] ffmpeg.log.20161220

burek burek021 at gmail.com
Wed Dec 21 03:05:01 EET 2016


[02:29:45 CET] <Kam> Hello, I'm playing a video file (tested with vp8, vp9, libvpx and libvpx-vp9) that is 2 FPS (500ms each frame) in ffplay with 'loop 0' parameter. The videos play fine, until the last frame, which only appears for a fraction of the time it should. The video has 8 frames overall, ffprobe reports its duration correctly being 00:00:04.00. The time-display in ffplay goes until about 3.50 and then resets too early back to 0.00. What can I do?
[02:52:54 CET] <furq> does it play correctly in mpv or some other actual video player
[02:59:32 CET] <Kam> furq, it looks like the same in mpv. The last frame is shown only for a fraction of time.
[04:58:06 CET] <kepstin> Kam: that's normal; in most containers, frames don't have duration info. If you have an audio track, it'll usually show the last frame until the audio ends.
[05:18:46 CET] <fling> How to use v4l output?
[06:05:59 CET] <ferdna> i need to live stream and save the live stream at the same time... how could i do this?
[06:33:00 CET] <brian_> I am building code based off the filter_audio.c file and for some reason av_frame_get_buffer always returns -22. Is it possible that I did not enable something in the configure that would cause these errors?
[07:45:37 CET] <nonex86> brian_: -22 as far as i remember is invalid parameter, you pass something wrong to function
[07:45:59 CET] <nonex86> brian_: and it not related to build configuration
[07:46:32 CET] <nonex86> brian_: enable verbose debug output in ffmpeg and check the error messages
[07:46:46 CET] <nonex86> brian_: twice check your input parameters
[08:14:17 CET] <ferdna> how do you check the headers of a m3u8 file on a webserver?
[08:39:22 CET] <Pandela> Morning
[08:40:53 CET] <Pandela> Is it possible to start synchronous ffmpeg udp streams with a single batch file?
[12:16:33 CET] <Eiken> so i am overlaying an endcards to a bunch of videos, if the aspect ratio is portrait or landscape it works
[12:16:46 CET] <Eiken> but if the video res and endcard res is square the endcard is incorrectly positioned
[12:20:17 CET] <Eiken> $FFMPEG_BIN -y -i input.mov -loop 1 -i endcard.png -filter_complex "[1:v]fade=in:0:10[v1]; [0:v]setpts=PTS-STARTPTS[v0]; [v1]setpts=PTS-STARTPTS+425/(TB*25)[v3];[v0][v3]overlay=eof_action=pass,scale=600:600[out1]" -map [out1] -map 0:a? -codec:a copy -pix_fmt yuv420p -c:v libx264 -preset veryslow -crf 12 output.mov
[12:20:27 CET] <Eiken> anyone see something that might explain it
[12:20:41 CET] <Eiken> i tried removing scale but same result
[12:25:18 CET] <Eiken> if i scale both the inputs first in the filter it works
[12:25:23 CET] <Eiken> but then it screws up the alpha
[12:42:46 CET] <Eiken> figured it out
[12:42:47 CET] <Eiken> :)
[12:42:55 CET] <Eiken> had to add scale after fade
[13:16:59 CET] <Kam> kepstin, but the duration is determined correctly by ffprobe. I want to use this in games, and adding an empty audio track to every video would unnecessarily add more load. Do you think this is a bug, or on purpose?
[13:28:53 CET] <Nacht> I'm doing some studying on cutting up TS files grabbed from a HLS stream (h.264). Of some of the things I read, people mention the Random Acces Indicator. Where other mention the IDR frame. Do they always 'align' when you're using a TS H.264 file ?
[13:29:22 CET] <Nacht> Cause from what I understood, the RAI is a property of the MPEG2-TS, and IDR of H.264 ?
[13:51:05 CET] <nonex86> in addition to idr h264 has sei recovery point
[14:17:57 CET] <DHE> Nacht: RAI is a bit in the MPEG-TS header to indicate that this frame is a good place to start decoding from. but it's a bit codec-dependent. h264 may make use of it, but mpeg2 doesn't
[14:19:14 CET] <Nacht> But if I wish to cut a TS-file which uses H.264, it's best to search for IDR and SEI Recovery points in order to determine where to cut ?
[14:20:45 CET] <DHE> if the video uses IDR frames, it should set the RAI bit. in that case the video should ideally contain 1 PAT, then 1 PMT, optionally an SDT, and then start at said RAI chunk
[14:22:20 CET] <SouLShocK> I stumbled across this old article a few days ago http://www.streamingmedia.com/Articles/Editorial/Featured-Articles/Streaming-Media-East-Netflix-Making-the-Move-to-HEVC-but-Efficiency-Gains-Lag-96981.aspx where Netflix all but confirms that they are using FFmpeg to encode
[14:22:23 CET] <SouLShocK> When we reject an upload, and send it back to a content provider, we have to give a detailed reason, he said. We cant just give an FFMPEG error code.
[14:22:36 CET] <SouLShocK> thought that might interest you guys
[14:25:24 CET] <Nacht> I see. Thank you for the information DHE. Ill have something to read upon then.
[14:25:45 CET] <DHE> wow, I'm entertained by their content providers providing provably bad source material :)
[14:47:39 CET] <popara> How i can add a Logo and use th yadif filter at the same time? Because i can do both seperate, but not at the same time. I'm using filter_complex for logo, and -vf yadif for deinterlace
[14:47:45 CET] <popara> And ffmpeg says, i can't use both at the same time
[14:50:01 CET] <popara> ah actually -filter_complex "overlay,yadif" that worked fine
[16:21:34 CET] <nirvanko> I've downloaded a video and ffprobe tells that its resolution is 1920x1040 [SAR 1:1 DAR 24:13, should it be 1920x1080? I mean there is the 1080p TV standard, why 1040?.
[16:24:33 CET] <kerio> nirvanko: why not 1920x1040?
[16:25:05 CET] <DHE> that may be standard, but videos may not conform for whatever reason
[16:25:20 CET] <nirvanko> Because Full HD supposed to be 1920x1080, isn't it?
[16:26:45 CET] <kerio> i mean the alternative is that the video somehow has a SAR of 104:108
[16:27:00 CET] <kerio> do you really want to live in that world
[16:30:15 CET] <DHE> SAR is 1:1 which means square pixels. there will be a bit of black bars on the top and bottom of the video. so it still fits in a 1080p display
[16:30:20 CET] <nonex86> to be really precise its 1920x1088
[16:30:48 CET] <nonex86> because of nature of encoders used
[16:31:19 CET] <nonex86> than it can be downscaled to real fullhd 1920x1080
[16:31:46 CET] <nonex86> so your strange 1040 can be encoder related value
[16:33:35 CET] <nonex86> but for h264 i guess 1040 is really strange
[16:34:35 CET] <nonex86> no, its ok
[16:34:39 CET] <nonex86> my mistake
[16:34:49 CET] <nonex86> 1040 divisible by 16, so its ok
[16:37:22 CET] <nonex86> 120x65 macroblocks
[16:39:24 CET] <sardok> Hi i want to extend given sound duration by adding silence. I tried 'ffmpeg -i sound.mp3 -complex_filter "apad" -t 20 out.aac' . But out file has the same duration as input.
[17:06:10 CET] <fling> sardok: aac counts silence at the end differently
[17:06:31 CET] <fling> D'oh.
[17:14:06 CET] <enderby> hi - I'm trying to modify metadata/ratings on files. Is there a better way to use ffmpeg via CLI than to create an entirely new file? or is there another tool I should be considering?
[17:25:39 CET] <furq> ffmpeg can't modify inplace, so it's probably not the best tool for metadata editing
[17:25:59 CET] <furq> if these are audio files then there are a ton of better tools
[17:26:17 CET] <furq> most of those should also work with video in mp4
[17:26:41 CET] <enderby> furq: thnx, can you name a few to get me started? i would like a cli interface
[17:31:03 CET] <furq> i've only ever really used mutagen and beets
[17:32:05 CET] <enderby> cool, will check them. thank you
[17:39:48 CET] <badasafish> Hi all, I'm trying to switch a 3D video from over under to side by side. I've researched the commands I need, I've swapped the images to be stacked left and right which is fine, but I need to resize and stretch the resulting image. The resize works fine on the test screenshot but doesn't have any stretch affect on the mkv. Thoughts?
[17:51:32 CET] <durandal_1707> badasafish: stereo3d filter
[17:53:24 CET] <nirvanko> Is it a good idea to resize video with a black padding at its bottom for subtitles? I would like to have white on black subtitles.
[17:53:45 CET] <nirvanko> Usually it's black on white but not in my case, not in my case.
[18:12:07 CET] <badasafish_> durandal_1707: I used "ffmpeg -i top-and-bottom.mov -vf stereo3d=abl:sbsl -c:a copy side-by-side.mov" which changed it from OU to SBS but the size/aspect ratio was not correct (i ended up with a 3840x540 video), therefore I resized it to 1920x1020 which worked as expected on a test screenshot, but when running that resize on the mkv although the video is 1920x1080 the video is not stretched as expected vertically
[18:15:30 CET] <durandal_1707> badasafish: sbs2l? note distinction
[18:17:09 CET] <badasafish> durandal_1707: I see, let me try that out.  So that should achieve the required output without needing a separate resize?
[18:18:29 CET] <rakshith> Hello guys, I am new to this community. Wanted to contribute. Help me get started :)
[18:20:46 CET] <durandal_1707> badasafish: it sets aspect ratio
[18:21:15 CET] <durandal_1707> rakshith: contribute what?
[18:22:16 CET] <rakshith> code
[18:24:04 CET] <badasafish> durandal_1707: I want to take a clip from my original movie file so I can test the conversion quickly without having to convert the whole file, whenever I create a clip of the original the stereo3d function won't work on the clip, is there a specific way of clipping I should use to preserve everything from the original?
[18:28:45 CET] <durandal_1707> rakshith: write qrcode source filter that use external library for example
[18:30:14 CET] <durandal_1707> badasafish: what doesn't exactly work?
[18:31:16 CET] <badasafish> 8durandal_1707: Thanks for your help with this btw.  Running stereo3d sbs2l has given the output that I wanted visually, but the file is still 3840 * 540, can I incorporate a resize filter into the command?
[18:32:34 CET] <rakshith> durandal_1707: something easier?
[18:33:03 CET] <durandal_1707> badasafish: you could resize but I do not know command to not keep aspect ratio
[18:33:29 CET] <durandal_1707> rakshith: spell check documentation
[18:36:19 CET] <badasafish> durandal_1707: I get this error: [h264 @ 000000000215b6c0] Warning: not compiled with thread support, using thread emulation [buffer @ 000000000080ada0] Unable to parse option value "-1" as pixel format     Last message repeated 1 times [buffer @ 000000000080ada0] Error setting option pix_fmt to value -1. [graph 0 input from stream 0:0 @ 0000000001f4a6c0] Error applying options to the filter. Error opening filters!
[18:52:49 CET] <badasafish> durandal_1707: I was creating the clip with the copy command, but now if I create the clip without the copy command (although it has to reencode the video) I can run the stereo3d filter on the clip
[19:09:55 CET] <blue_misfit> hey folks, I'm seeing something odd - a mkv that is 23.976 fps cfr is vfr after I remux to MP4 with ffmpeg... any idea why?
[19:11:57 CET] <JEEB> blue_misfit: I'm not sure you can save 24000/1001 as CFR in matroska
[19:12:26 CET] <blue_misfit> uhh... really? why's that??
[19:12:33 CET] <JEEB> timebase is limited
[19:12:50 CET] <JEEB> I think stuff like ffms2 etc have heuristics to try and work around oscillating duration
[19:12:59 CET] <JEEB> mpv's matroska demuxer as well
[19:13:19 CET] <JEEB> also you should check what  something actually uses to denote if a file is "CFR" or "VFR"
[19:13:35 CET] <FishPencil> Does VP8 or VP9 provide better compression than x264?
[19:13:44 CET] <__jack__> yes
[19:13:47 CET] <blue_misfit> well I'm just looking at mediainfo metadata because the people I'm delivering the file to are seeing that and whining that it says VFR
[19:13:50 CET] <JEEB> FishPencil: you are comparing formats to encoders
[19:14:05 CET] <FishPencil> JEEB: libvpx then
[19:14:15 CET] <blue_misfit> @FishPencil, vp9 absolutely can provide better compression than avc, and libvpx can provide better compression than x264 ;)
[19:14:58 CET] <FishPencil> So does anything top libvpx then in compression?
[19:15:18 CET] <blue_misfit> yes, x265 is definitely better IMO
[19:15:28 CET] <JEEB> libvpx is not really optimized in the rate control or psychovisual optimizations
[19:15:32 CET] <blue_misfit> correct
[19:15:37 CET] <blue_misfit> its rate control is horrnedous
[19:15:48 CET] <blue_misfit> as is its multithreading since it's designed for cloud scale single thread encoding
[19:15:52 CET] <furq> it's also slower than x265 in my experience
[19:15:58 CET] <furq> mostly because of the multithreading
[19:16:00 CET] <blue_misfit> well... it's better than it was, but it's still not great :D
[19:16:20 CET] <blue_misfit> vp9 decoding is surprisingly fast in software tho, that's one major advantage to the format in general IMO
[19:16:31 CET] <furq> hevc hardware decoders are more widespread though
[19:16:35 CET] <blue_misfit> hence why YouTube streams it by default almost everywhere
[19:16:35 CET] <blue_misfit> sure
[19:16:43 CET] <furq> although still nowhere close to avc
[19:16:45 CET] <blue_misfit> s/everywhere/most browsers
[19:16:46 CET] <JEEB> yeah, BBB who was making his own encoder made a Very Fast decoder
[19:17:01 CET] <JEEB> which was then included in Firefox at least
[19:17:15 CET] <JEEB> (so yes, Firefox has a separate minimal part of libavcodec in there)
[19:17:15 CET] <furq> yeah i noticed firefox finally started playing vp9 by default fairly recently
[19:17:21 CET] <furq> i expect chrome has been doing it for much longer
[19:17:25 CET] <blue_misfit> correct
[19:17:26 CET] <JEEB> (and they call it ffvp9 or so)
[19:17:30 CET] <blue_misfit> and Edge!
[19:17:36 CET] <JEEB> chromium and edge are using libvpx
[19:17:38 CET] <blue_misfit> Safari is the only holdout :D
[19:17:56 CET] <furq> obviously as soon as microsoft starts supporting new standards, apple decide they don't want to
[19:17:57 CET] <JEEB> which are somewhat slower than lavc's vp9 decoder (at least on x86)
[19:18:26 CET] <efface> Im using ffprobe to determine the resolution and aspect ratio of programs in various mpeg-ts muxes.  On some programs the information is returned and on other programs the width and height are 0 and the sample ration is 0:1 and display ration is 0:0.  Anyone know what would cause this
[19:18:31 CET] <JEEB> anyways, blue_misfit - I'd guess the timestamps in the matroska file are not actually CFR, and if mediainfo is saying it's CFR there's something funky
[19:18:43 CET] <furq> efface: maybe -probesize and -analyzeduration
[19:18:50 CET] <blue_misfit> interesting
[19:19:11 CET] <furq> does mkv really not support 24000/1001 properly
[19:19:13 CET] <JEEB> blue_misfit: in mp4 it can be CFR but I have no idea how that is actually "checked" in mediainfo
[19:19:20 CET] <blue_misfit> JEEB, what I'm doing is using mkvmerge to concat a bunch of MP4s (since I'm doing split 'n' stitch style parallel encoding)
[19:19:23 CET] <furq> does that mean it also doesn't support 30000/1001
[19:19:27 CET] <JEEB> furq: ayup
[19:19:33 CET] <JEEB> oscillating durations
[19:19:38 CET] <furq> nice
[19:19:39 CET] <FishPencil> https://en.wikipedia.org/wiki/Libvpx "Decoding performance is relatively slow, partially in order to keep the code base easier to maintain."
[19:19:41 CET] <JEEB> because you don't get exact timestamps
[19:19:44 CET] <efface> Odd part is i tried analyzeduration and it says its an unknown option
[19:19:49 CET] <furq> i guess it's close enough that you wouldn't notice
[19:20:08 CET] <furq> efface: what ffprobe version
[19:20:23 CET] <efface> Nm the duration issue...im an idiot lol
[19:20:23 CET] <JEEB> a lot of things thus have stuff in their parser that if your timestamps are close enough it will utilize what it thinks the timestamps are
[19:20:32 CET] <furq> FishPencil: that's the libvpx decoder
[19:20:34 CET] <JEEB> like ffms2 or mpv's matroska demuxer
[19:20:38 CET] <furq> ffmpeg doesn't use that
[19:21:02 CET] <blue_misfit> what would you suggest if I needed to unify chunks of video all produced from the same source
[19:21:27 CET] <JEEB> https://github.com/FFMS/ffms2/blob/master/src/core/videoutils.cpp#L71
[19:21:38 CET] <JEEB> this is how ffms2 deals with matroska
[19:21:39 CET] <JEEB> :D
[19:22:11 CET] <furq> nice
[19:22:37 CET] <efface> Analyzeduration 5000000 did not help
[19:23:00 CET] <furq> that's the default isn't it
[19:23:04 CET] <furq> it's in microseconds or something daft
[19:23:28 CET] <efface> It is ms....so 5 secs, let me try 10
[19:24:30 CET] <furq> you might as well just use $((2**31-1))
[19:24:34 CET] <furq> likewise for -probesize
[19:24:41 CET] <furq> if that doesn't work then the problem is elsewhere
[19:26:05 CET] <furq> obviously don't do that if this is streaming because that'll probably never finish
[19:27:31 CET] <efface> It is streaming
[19:28:12 CET] <furq> well yeah you probably don't want to download 2GB before anything happens
[19:28:50 CET] <efface> Lol
[19:29:06 CET] <efface> Ok i set probesize to 150M and i think that worked
[19:29:18 CET] <efface> The mux bitrate is 38mbps
[19:31:24 CET] <efface> That did work, thanks
[19:31:42 CET] <FishPencil> I'd say the most popular video encoders today are x264, x265, and libvpx. For audio encoders I would say opus, and probably acc. Am I forgetting anything in these lists? Obviously libschroedinger and the likes don't belong.
[19:32:08 CET] <furq> opus is not that popular
[19:32:33 CET] <furq> maybe it's in second place if you include youtube, but that's a huge skew of the average
[19:33:02 CET] <furq> aac is almost certainly the most popular, and mp3 is still up there
[19:33:14 CET] <FishPencil> s/popular/modern
[19:33:55 CET] <FishPencil> modern=great compression, actively developed
[19:34:06 CET] <furq> lame is still actively developed
[19:34:19 CET] <FishPencil> what....
[19:34:28 CET] <furq> mp3 is easily the most popular if you're just talking about music
[19:34:53 CET] <furq> itunes, google play music, soundcloud, bandcamp etc etc
[19:35:28 CET] <furq> actually i guess lame hasn't had an update since 2012
[19:35:45 CET] <FishPencil> In almost all other respects I'd say Opus is better, or even AAC
[19:36:03 CET] <furq> those are both definitely better codecs than mp3
[19:36:30 CET] <furq> along with vorbis and probably some other lossy codecs that nobody uses
[19:37:48 CET] <blue_misfit> yeah mp3 is everywhere lol
[19:38:08 CET] <furq> mp3 will probably be the most popular lossy codec for music forever
[19:38:15 CET] <furq> by the time it falls off, everyone will be using lossless codecs
[19:38:33 CET] <blue_misfit> yep, "good enough" for basically everything
[19:38:50 CET] <FishPencil> If it isn't broke don't fix it
[19:39:04 CET] <furq> the compression gains from using opus aren't worth it for most people, especially with the compatibility tradeoff
[19:39:17 CET] <blue_misfit> exactly
[19:39:35 CET] <blue_misfit> but for next gen streaming (vp9 + opus in webm in dash) it's a great fit :)
[19:39:53 CET] <furq> yeah i don't mean to be down on opus, it's a nice codec
[19:39:54 CET] <FishPencil> furq: I'm not seeing active development on lame
[19:40:08 CET] <furq> 18:35:28 ( furq) actually i guess lame hasn't had an update since 2012
[19:41:49 CET] <blue_misfit> JEEB, MP4 does properly support CFR for 24000/1001 right?
[19:42:05 CET] <JEEB> yes
[19:42:23 CET] <JEEB> timebase 24000 and then each sample has a duration of 1001
[19:42:26 CET] <furq> FishPencil: it's worth remembering that some people are still using xvid
[19:42:28 CET] <blue_misfit> how can I be certain I'm producing a CFR output when encoding H.264 in MP4 using libx264 and ffmpeg
[19:42:34 CET] <furq> although in this case, those people are just severely wrong
[19:42:51 CET] <JEEB> blue_misfit: you can utilize l-smash's boxdumper and look at the sample durations
[19:43:03 CET] <blue_misfit> furq, :D yeah tons of people still rock xvid and mp3 in avi :D
[19:43:08 CET] <FishPencil> I'm not sure what you're talking about, xvid all day every day
[19:43:11 CET] <furq> and mpeg2 is still widely used in broadcasting
[19:43:13 CET] <blue_misfit> JEEB, I'll look into that, thanks
[19:43:26 CET] <blue_misfit> OMG we used MPEG-2 everywhere at my last job
[19:43:41 CET] <blue_misfit> _all_ media that plays on air plays from MPEG-2 files, usually 1080i59.94 at 20, 35, or 50 Mbps
[19:43:44 CET] <JEEB> blue_misfit: it can just dump the timestamps, or you can get the whole box dump
[19:44:13 CET] <furq> some stations broadcast avc now don't they
[19:44:21 CET] <blue_misfit> yeah some of them do AVC Intra
[19:44:24 CET] <furq> i know a lot of IPTV stuff does
[19:44:27 CET] <blue_misfit> that's the only AVC format that's gotten any traction
[19:44:34 CET] <JEEB> furq: he's talking pre-broadcast
[19:44:40 CET] <JEEB> so the stuff that goes through in the broadcast house
[19:44:41 CET] <blue_misfit> yeah
[19:44:44 CET] <furq> oh right
[19:44:58 CET] <furq> well a lot of stations still broadcast mpeg2 don't they
[19:45:01 CET] <blue_misfit> most broadcast is AVC (some legacy MPEG-2, and TONS Of it in cable)
[19:45:22 CET] <blue_misfit> but yeah the files that the linear playout servers use are almost all MPEG-2
[19:45:27 CET] <JEEB> blue_misfit: but it (boxdumper) should produce pretty tangible evidence that it's CFR, because with mp4 it's deffo possible
[19:45:37 CET] <JEEB> unlike matroska
[19:45:55 CET] <blue_misfit> damn
[19:46:00 CET] <blue_misfit> what kind of problems could this cause in the real world?
[19:46:07 CET] <blue_misfit> because everyone uses MKV for 23.98 lol
[19:46:21 CET] <JEEB> and almost everything has heuristics for this in playback
[19:46:51 CET] <JEEB> and the timestamps are not too far apart from the correct values, they're just not CFR
[19:47:07 CET] <JEEB> so you switch between two durations for your video samples
[20:02:21 CET] <blue_misfit> @JEEB, where can I get windows binaries of l-smash stuff? I'm having a hell of a time finding any!
[20:06:48 CET] <blue_misfit> nm.. just made my own :D
[20:06:54 CET] <JEEB> yeh, it isn't too hard
[20:07:05 CET] <JEEB> even less so with the MSVS projects now
[20:07:22 CET] <blue_misfit> yeah lol thankfully I have a decent msys environment thanks to ffmpeg-autobuild
[20:07:56 CET] <FishPencil> blue_misfit: What are you looking for? There's Zeranoe'
[20:07:58 CET] <FishPencil> s builds
[20:08:15 CET] <JEEB> https://github.com/l-smash/l-smash
[20:09:21 CET] <FishPencil> Doesn't look like that's included in FFmpeg
[20:09:31 CET] <JEEB> nope, but it's a very useful toolkit for ISOBMFF
[20:09:43 CET] <JEEB> and it's really simple to build
[20:14:10 CET] <FishPencil> Is anyone here familiar with EME or Widevine?
[20:14:19 CET] <blue_misfit> FishPencil, yeah a bit
[20:14:38 CET] <FishPencil> blue_misfit: This isn't FFmpeg related... PM?
[20:15:00 CET] <blue_misfit> sure
[20:32:27 CET] <FishPencil> I'm doing some benchmarking and I'm looking for some native encoders that FFmpeg does a lot of threading work with (assembly optimizations as well). I'm thinking mpeg4, but I'm not sure about audio.
[20:34:19 CET] <JEEB> mpeg4 isn't really optimized
[20:34:27 CET] <JEEB> I mean, compared to something like libx264
[20:34:39 CET] <JEEB> libx264 actually had some use and thus is wroom-wroom
[20:35:08 CET] <FishPencil> Anything internal that is well optimized?
[20:35:21 CET] <FishPencil> Doesn't FFmpeg impliment its own vpx
[20:35:38 CET] <blue_misfit> JEEB, regarding my timestamp issue, any way to regenerate the timestamps when muxing to MP4?
[20:36:06 CET] <blue_misfit> I verified that an encode directly to mp4 has perfectly consistent timestamps but that they oscillate between 672 and 656 for MKV -> MP4
[20:36:26 CET] <JEEB> FishPencil: decoder only
[20:36:30 CET] <JEEB> not encoder
[20:36:39 CET] <FishPencil> oh
[20:36:41 CET] <JEEB> encoding is usually left to the projects best aimed at it
[20:39:11 CET] <phillipk> anyone know of any super basic intros to ffmpeg?  I feel this was missing from my early times with ffmpeg.  The wiki is okay, but assumes a lot of base knowledge.
[20:39:30 CET] <FishPencil> phillipk: CLI or API
[20:40:06 CET] <FishPencil> phillipk: A "not sure" means CLI
[20:40:07 CET] <phillipk> CLI I suppose... (not api to build off of ffmpeg+)
[20:40:44 CET] <FishPencil> phillipk: #ffmpeg isn't a bad place :) Do you have a specific question?
[20:41:58 CET] <phillipk> I have had questions and worked through everything I need for now.  But, for example, the docs on filter graphs seem unclear.  And even the language used in the docs could be greatly improved.  Not trying to complain but wondering if others have written up blog posts or primers.
[20:46:40 CET] <FishPencil> The docs have always been a sore spot...
[20:52:22 CET] <c_14> phillipk: patches welcome
[20:52:36 CET] <c_14> There are "a bunch" of small guides on trac.ffmpeg.org
[21:01:02 CET] <dalmatinac> Hello from sad Berlin :( I would like to download a hevc/mpeg-2 ts video from my server and as it is being downloaded and increasing in size I would like to encode it to h.264 and play it on my local browser.
[21:02:15 CET] <c_14> Does your browser support fragmented moov?
[21:03:43 CET] <dalmatinac> I'm using Electron, so it would be Chromium. I don't if it is supported by chromium. :/
[21:04:56 CET] <dalmatinac> I'm able to conver the file, and play it in the browser. But the problem is that it is not converting the whole downloaded video. Everytime it stops at a size of couple of hundred bytes
[21:04:59 CET] <dalmatinac> My command is: Spawned Ffmpeg with command: "ffmpeg -re -i testVideo.ts -y -acodec aac -strict experimental -b:a 128k -ac 2 -b:v 1024k -vcodec libx264 -f mp4 -movflags frag_keyframe+faststart -map 0:1 output.mp4" and I get this output: http://pastebin.com/dET5uYq5
[21:07:20 CET] <c_14> Consider updating your copy of ffmpeg, add -map 0:v:0
[21:09:12 CET] <dalmatinac> Thank you c_14! I will now try to update to the latest version for ubuntu, add the argument and report again :)
[21:18:12 CET] <podman> I'm running into an issue where i need to pad the beginning of a video with silence. It looks like asyncts is deprecated. I tried using the aresample filter but it didn't seem to do anything
[21:18:46 CET] <podman> do i need to do anything special to get something aresample=aysnc=1000 to work?
[21:23:23 CET] <podman> do i need to make sure ffmpeg is compiled with --enable-avresample?
[21:29:23 CET] <c_14> No, swresample should be built by default
[21:30:42 CET] <furq> podman: wouldn't adelay be a better choice for that
[21:31:04 CET] <dalmatinac> I have now updated to ffmpeg 3.2.2 and added the argument. The downloaded file was at 20mb(6sec) when I started the ffmpeg command. But at 1.4mb(6sec) ffmpeg finished, altough the downlod was still pending and increasing in size to 60mb(18secs). This is the output: http://pastebin.com/dET5uYq5
[21:31:15 CET] <podman> c_14: interesting. so I have a file that has a delayed audio track that starts about 20 seconds in. I need to create silent audio samples for those first 20 seconds
[21:32:14 CET] <c_14> dalmatinac: that's still the output from the 2.8 ffmpeg
[21:32:48 CET] <dalmatinac> Oh sorry! Here it is: http://pastebin.com/FQf1rgcr
[21:33:31 CET] <podman> c_14: hmm, maybe i want to use first_pts=0 instead?
[21:34:23 CET] <dalmatinac> I thought with the -re argument I'm telling ffmpeg that this file will still increase. But it seems he is only processing till the point, when he started.
[21:34:45 CET] <c_14> podman: have you tried adelay like furq said? (downside is you'll have to know the input channel layout)
[21:35:11 CET] <c_14> dalmatinac: how are you downloading the file?
[21:35:43 CET] <podman> c_14: i don't want to delay the audio channel, i want to pad the audio channel with silence
[21:35:59 CET] <furq> that's what adelay does
[21:36:28 CET] <dalmatinac> @c_14 With "wget www......" in another terminal
[21:36:31 CET] <podman> furq: so i need to know the exact amount of delay there is to get it to work?
[21:36:33 CET] <furq> https://ffmpeg.org/ffmpeg-filters.html#adelay
[21:36:41 CET] <furq> well presumably you'd need to know that anyway
[21:36:44 CET] <c_14> dalmatinac: ffmpeg can take input directly from http, try that
[21:37:01 CET] <podman> furq: sounds like the these other filters make more sense
[21:39:17 CET] <podman> furq: also, will that just add extra delay, or remove the delay and insert silence in it's place?
[21:46:17 CET] <dalmatinac> c_14: It worked! Thank you very much! Only one problem... This is just a workaround. I have a DVB-T stick for my laptop and I get a signal from it over a node.js module and I save it to a .ts file. Over the holidays I don't have this stick, so I uploaded a recording from it to a server and right now I try to simulate the stick by downloading the recording to a file.
[21:47:20 CET] <dalmatinac> c_14: So I'm thinking if I get it to work with this increasing file by downloading it, it will later also work with the DVB-stick. Do you know why it's not working when I specify the file, but it works when I specify the link?
[21:48:43 CET] <c_14> Probably because the file is so small that ffmpeg reads a block into the cache and the OS signals EOF
[21:51:52 CET] <dalmatinac> Ah, okay. Can we do anything about it? Maybe tell him, to ignore the EOF or define smaller blocks?
[21:56:16 CET] <c_14> try setting -blocksize 100K before -i
[22:04:44 CET] <podman> c_14: furq so, using aresample=first_pts=0 makes the audio track and the video track have the same length, so that's an improvement
[22:07:54 CET] <dalmatinac> c_14: I tried it but unfortunately, again it will stop when it reached the length of the input file, when I started the ffmpeg command :(
[22:08:38 CET] <c_14> try something smaller maybe? 1K?
[22:14:58 CET] <dalmatinac> c_14: Also not :( I start ffmpeg, when the file has 20%, 6MB. But then it will only create an output.mp4, which has a couple of seconds video and finishes...
[22:15:29 CET] <c_14> hmm, no clue then
[22:19:41 CET] <blue_misfit> hey folks, how can I regenerate timestamps when remuxing from mkv to mp4?
[22:19:43 CET] <dalmatinac> c_14: But thank you very much for your help! :)
[22:19:49 CET] <podman> c_14: furq: so, yeah, that worked! audio packets at the start of the video
[22:44:36 CET] <fling> How to deal with pulse delay when capturing sound and video?
[22:50:01 CET] <EliteHunter> I'm using this command on my batch file --start ffmpeg -ss %stTime% -i %path% -c:a copy -c:v copy -to %endTime% -copyts "output.mp4"-- to split a video. The process is completed successfully and i have no problems with windows media player but when i open the video on VLC the timestamp is all f* ed up. Why is that?
[22:52:14 CET] <fling> EliteHunter: because of -copyts and mp4
[22:53:15 CET] <EliteHunter> what should i do to fix it?
[22:53:22 CET] <fling> EliteHunter: try omitting -copyts
[22:53:42 CET] <EliteHunter> but that results in endtime being runtime
[22:54:02 CET] <EliteHunter> is the reason mp4 not supporting something with timestamps?
[22:54:32 CET] <EliteHunter> what i mean runtime is this
[22:54:35 CET] <EliteHunter> The first command will cut from 00:01:00 to 00:03:00 (in the original), using the faster seek. The second command will cut from 00:01:00 to 00:02:00, as intended, using the slower seek. The third command will cut from 00:01:00 to 00:02:00, as intended, using the faster seek.  ffmpeg -ss 00:01:00 -i video.mp4 -to 00:02:00 -c copy cut.mp4 ffmpeg -i video.mp4 -ss 00:01:00 -to 00:02:00 -c copy cut.mp4 ffmpeg -ss 00:01:00 -i video.mp
[22:54:50 CET] <EliteHunter> https://trac.ffmpeg.org/wiki/Seeking
[22:55:41 CET] <fling> ahh
[22:56:21 CET] <EliteHunter> yeah so i don't want to omit copyts since i want to give a specific time range and not calculate how much will the video run for and then enter the time
[22:57:20 CET] <fling> But you should drop and recalculate it in the end somehow anyway as you are copying it in the middle
[22:59:04 CET] <EliteHunter> doesn't -ss and -to already does that?
[23:00:32 CET] <EliteHunter> when i run this even if the timestamps are wrong, there is nothing wrong with the video
[23:01:10 CET] <EliteHunter> if i run -ss 00:10:00 -to 00:11:00 the video runs for 1 minute, but you can't seek on VLC and the end timestamp is wrong
[23:01:34 CET] <EliteHunter> so there are no errors in the cut process but somehow timestamps are wrong in vlc
[23:01:50 CET] <EliteHunter> however windows media player correctly runs the video
[23:01:54 CET] <fling> after -copyts?
[23:02:31 CET] <EliteHunter> what do you mean by after?
[23:03:13 CET] <fling> ffmpeg -i & -copyts output.mp4 && vlc output.mp4
[23:03:55 CET] <EliteHunter> i don't have vlc in the command, i open it using the client
[23:04:14 CET] <fling> Does not matter.
[23:04:16 CET] <EliteHunter> btw my bad after i tried couple more videos windows media player seems to not work either
[23:04:36 CET] <EliteHunter> i run the command and open the output in vlc
[23:04:48 CET] <fling> after -copyts?
[23:04:50 CET] <EliteHunter> yes
[23:05:07 CET] <fling> copyts copies your ts from the input
[23:05:08 CET] <EliteHunter> it is a one line command so there is no way i would be able to open it before
[23:05:30 CET] <fling> so the shrinked output has ts from the longer input
[23:05:41 CET] <fling> your players don't like this behavior and bork
[23:06:05 CET] <fling> try not using copyts or try recalculating it in the end
[23:06:14 CET] <fling> gtg, cya later, good luck!
[23:06:40 CET] <EliteHunter> ok thanks, i'll calculate it in the batch file since it's not that hard. Thanks for your help
[00:00:00 CET] --- Wed Dec 21 2016


More information about the Ffmpeg-devel-irc mailing list