[Ffmpeg-devel-irc] ffmpeg.log.20150805

burek burek021 at gmail.com
Thu Aug 6 02:05:01 CEST 2015

[00:00:21 CEST] <kyleogrg> other words, 33% of *640* is 211.
[00:01:41 CEST] <kyleogrg> and to do this, I can do scale=iw*sar:ih, then scale=iw*0.33?
[00:12:31 CEST] <PlanC> I'm trying to experiment with a new thing
[00:12:40 CEST] <kyleogrg> it works if i do 50%, but not 33%.  probably a mod 2 issue or something.
[00:12:49 CEST] <PlanC> We have a website which has a bunch of demo videos that get a decent amount of downloads
[00:13:00 CEST] <PlanC> Users can download these in different formats with the help of FFMPEG
[00:13:43 CEST] <PlanC> But the downside is that they have to wait for the video to be converted to 100% before they can start downloading
[00:14:15 CEST] <PlanC> I got the idea that I perhaps could convert chunks at a time and serve it to the user
[00:14:50 CEST] <PlanC> Does FFMPEG have a feature for this which would allow me to convert one chunk at a time and serve it to the user (I'll serve it with PHP but that's kind of irrelevant)
[00:15:21 CEST] <c_14> use the segment muxer?
[00:15:22 CEST] <PlanC> another option would be to use concatenate and have FFMPEG convert let's say 1 second at a time
[00:15:41 CEST] <PlanC> c_14: what does the segment muxer do?
[00:16:03 CEST] <c_14> cuts the video up into n x length segments
[00:16:22 CEST] <c_14> https://ffmpeg.org/ffmpeg-formats.html#segment_002c-stream_005fsegment_002c-ssegment
[00:16:46 CEST] <PlanC> was reading that just now
[00:16:56 CEST] <PlanC> so I can basically stream the conversion
[00:17:32 CEST] <PlanC> but does this update a file or does it output the chunks right in the shell?
[00:18:09 CEST] <c_14> This outputs n x-length files.
[00:18:16 CEST] <c_14> It sounds to me an awful lot like you want HLS though.
[00:18:45 CEST] <llogan> kyleogrg: use -2 in w
[00:19:08 CEST] <PlanC> HLS?
[00:19:11 CEST] <PlanC> what's that?
[00:19:39 CEST] <c_14> https://en.wikipedia.org/wiki/HTTP_Live_Streaming
[00:20:47 CEST] <PlanC> no I'll be sending the file mime type headers with PHP to have the browser force-download the file instead of playing it
[00:21:08 CEST] <PlanC> from there the chunks will be sent to the user
[00:21:42 CEST] <PlanC> now I'm trying to figure out how I can progressively get each chunk from FFMPEG using the segment muxer
[00:23:57 CEST] <c_14> You could output to a format that's streamable.
[00:23:59 CEST] <kyleogrg> llogan: hmm, can you give me an example?
[00:24:04 CEST] <c_14> So you could send it out while it's encoding.
[00:25:41 CEST] <llogan> kyleogrg: scale=-2:ih/3
[00:25:46 CEST] <PlanC> almost all formats are streamable though, right?
[00:25:59 CEST] <c_14> mp4 isn't
[00:26:12 CEST] <PlanC> I'll be streaming mp4, avi, mp3 (sound effects), etc
[00:26:37 CEST] <PlanC> actually they don't need to be streamable
[00:26:46 CEST] <PlanC> because the user will play it after it's downloaded
[00:27:11 CEST] <PlanC> I'm just trying to get the download to start sooner so that the user doesn't have to wait until the conversion reaches 100%
[00:27:40 CEST] <PlanC> my goal is to have the user start the download by getting the output from 1% to 100%
[00:27:57 CEST] <kyleogrg> llogan: what I have now is "scale=iw*sar:ih,scale=iw*0.5".  (0.5 because 0.33 didn't work.)  i'm not sure how to use your suggestion
[00:28:44 CEST] <llogan> scale=iw*sar:ih,scale=-2:ih/3
[00:29:34 CEST] <c_14> PlanC: And the user is supposed to use a media player to view the stream or download it to file with something like curl/wget
[00:29:37 CEST] <c_14> ?
[00:31:31 CEST] <PlanC> c_14: the user will download it with their regular browser, save it to their PC and play it
[00:32:00 CEST] <kyleogrg> llogan:  that works.  thanks a lot.
[00:33:04 CEST] <c_14> PlanC: Because the thing is, you can't just encode (most formats) to individual files and then concatenate them willy-nilly so that it looks like only one file to the client.
[00:35:35 CEST] <kyleogrg> since I am encoding with super-low bitrates, would it make sense to also shrink video resolution?
[00:36:57 CEST] <KarlFran1> I think it does.
[00:37:28 CEST] <KarlFran1> If you have a 1080 video with a very low bitrate you are defeating the point of high resolution.
[00:37:55 CEST] <KarlFran1> And also getting a worse bitrate per pixel than using a low resolution for the same video bitrate :-)
[00:41:31 CEST] <kyleogrg> right.  i guess i will estimate what the res should be given the bitrate
[00:42:30 CEST] <KarlFran1> You can always calculate the bitrate per pixel when you are stimating. I don't know if it is technichaly correct, but results don't look bad.
[00:43:14 CEST] <KarlFran1> Say that if a video with X bitrate on 720 looks crappy, a 1080 video with resolution X x 2.25 will look also crappy.
[00:45:31 CEST] <kyleogrg> yeah, i was wondering about that.
[00:46:22 CEST] <kyleogrg> what if it's 92k bitrate for a 640x480 video at 14.985p?
[00:47:50 CEST] <KarlFran1> wow, that is low fps.
[00:48:21 CEST] <kyleogrg> yup.  29.97/2
[00:49:00 CEST] <KarlFran1> You are likely to see compression artifacts in that.
[00:49:11 CEST] <KarlFran1> Just try it and tell us the results.
[00:49:12 CEST] <PlanC> c_14: I'll have to read more about this before I can even get a basic understanding of what I'm trying to do
[00:49:15 CEST] <PlanC> thanks for your help
[00:51:47 CEST] <kyleogrg> KarlFran1: i've tried it already, and there are artifacts.  i'm just experimenting with super-low bitrates, seeing how much video proxies i can cram on a thumb drive.
[00:52:27 CEST] <KarlFran1> I think that h.264 starts being not-so-good at 500k for 720.
[00:52:47 CEST] <KarlFran1> You can do the math for the resolutions you want.
[00:53:30 CEST] Action: KarlFran1 usually uses VP8.
[00:54:33 CEST] <kyleogrg> i'm using x265
[04:48:45 CEST] <k_sze> Does anybody know of an easy-to-use ffmpeg binding for .NET that supports seeking? I think so far I haven't seen any binding that is both easy to use and supports seeking.
[05:47:06 CEST] <vee> if i am making a slideshow from a series of pictures using -i "format string" - is there a method of verbosity that outputs the current picture it is processing (which causes a swscaler warning/error)?
[10:11:05 CEST] <hassoon> what's with this red notification i get whenever i open a media file with mpv, in debian '[ffmpeg/audio] aac: element type mismatch 1 != 0'
[10:11:23 CEST] <hassoon> https://i.imgur.com/d0Mf7wR.png
[10:42:41 CEST] <lilibox_> hi
[10:43:38 CEST] <lilibox_> does ffmpeg's development follow the latest cameras format? ARRI/Red/Sony?
[11:08:39 CEST] <JEEB> lilibox_: they should use standard stuff, and if they don't someone gives samples and support might be added depending on how disgusting the hacks can be
[11:19:57 CEST] <santhoshsd> how do I embed a closed caption file (scc) into a video file ?
[11:33:38 CEST] <whald> hi! i'm currently fighting with an IP camera producing a pretty crappy MJPEG stream. i'm using a command like "ffmpeg -f mjpeg -i "<URL>" -c:v libx264 ... -f mpegts udp://" to convert it to an multicast stream.
[11:34:38 CEST] <whald> this kind-of works, but the camera does only produce ~2fps, which, because of buffering on playback side, introduces huge latencies
[11:35:48 CEST] <whald> i've played around with the "-r" and "-vf 'fps'" options to get a 25fps stream (duplicating frames as necessary), but it seems to have no effect. any clue what i'm missing?
[11:36:54 CEST] <whald> i.e. inserting a "-vf 'fps=fps=25'" right after the "-i" has no effect, but from the documentation i assume it should?
[12:53:44 CEST] <sco_> greetings, I'm compiling ffmpeg on nvidiaa tegra tk1. After satisfying all of the .configure eroors, I'm now getting a make error: "bad instruction 'emm'. can anyone help me?
[12:53:56 CEST] <sco_> emms
[12:58:13 CEST] <spaam> maybe if you show us the configure line you used.
[12:58:34 CEST] <sco_> 1 sec ....
[12:58:56 CEST] <spaam> it will be easier to help you if you show us that :) then someone might be able to help you with it :D
[12:59:18 CEST] <sco_> have to reboot my system. Will send the config in a minue ....
[12:59:38 CEST] <iive> sco_: "emms" is mmx instruction, aka x86
[12:59:59 CEST] <sco_> I'm not use 32 bit
[13:00:04 CEST] <sco_> using
[13:01:17 CEST] <sco_> still booting zzzzzzzzzzzzzzzz
[13:03:21 CEST] <sco_> ./configure --prefix=/usr --bindir=/usr/bin --datadir=/usr/share/ffmpeg --incdir=/usr/include/ffmpeg --libdir=/usr/lib64 --mandir=/usr/share/man --arch=x86_64 --optflags='-O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector --param=ssp-buffer-size=4 -m64 -mtune=generic' --enable-bzlib --disable-crystalhd --enable-frei0r --enable-libass --enable-libcdio --enable-libcelt --enable-libdc1394 --disable-indev=jack --enable-libfreetype --enable-
[13:03:22 CEST] <sco_> libmp3lame --enable-openal --enable-libopencv --enable-libopenjpeg --enable-libopus --enable-libpulse --enable-libschroedinger --enable-libspeex --enable-libtheora --enable-libvorbis --enable-libv4l2 --enable-libvpx --enable-libx264 --enable-libxvid --enable-x11grab --enable-avfilter --enable-postproc --enable-pthreads --disable-static --enable-shared --enable-gpl --disable-debug --disable-stripping --shlibdir=/usr/lib64 --enable-runtime-cpudetect --enable-
[13:03:23 CEST] <sco_> nonfree --enable-gpl --enable-version3 --enable-libbluray --enable-libmp3lame --enable-libopenjpeg --enable-libfaac --enable-libfdk-aac --enable-libwebp --enable-opencl --enable-x11grab --enable-opengl --cpu=native --enable-nvenc
[13:03:31 CEST] <sco_> sorry for the long post
[13:03:47 CEST] <sco_> didnt know it would be more than 3 lines
[13:06:43 CEST] <spaam> sco_: are you using that line to compile ffmpeg on your nvidia tegra tk1?
[13:07:42 CEST] <spaam> sco_: it looks like you trying to use the same configure line that you have on your x86 64bit computer
[13:08:18 CEST] <spaam> sco_: last time i checked nvidia tegra tk1 have an ARM cpu  and not x86 one
[13:09:08 CEST] <sco_> yes
[13:09:29 CEST] <sco_> you are right. My mistake. What should I change
[13:09:32 CEST] <sco_> ?
[13:09:42 CEST] <sco_> arch
[13:09:47 CEST] <sco_> i see.
[13:09:52 CEST] <sco_> What else?
[13:10:07 CEST] <spaam> libdir maybe or just remove it
[13:10:37 CEST] <chungy> you shouldn't need to pass most of those options to configure on either arch
[13:10:52 CEST] <sco_> I c. I didnt take note.
[13:10:54 CEST] <spaam> that too :)
[13:11:56 CEST] <sco_> ok I removed all that nonsense. I'll recomile and see what else I messed up. :-)
[13:24:13 CEST] <sco_> http://paste.opensuse.org/39568055
[13:24:20 CEST] <sco_> more errors
[13:31:40 CEST] <spaam> configure line ?
[13:34:34 CEST] <sco_> 1 sec ...
[13:36:55 CEST] <sco_> http://paste.opensuse.org/37500319
[13:56:45 CEST] <yaron> hey guys how do i create overlay between 2 videos and not keeping them in the background ?
[13:57:21 CEST] <yaron> something like overlay=shortest=1:eof_action=pass .. but i want to pass to next vid and not for the main blank
[16:01:47 CEST] <whald> just in case anyone cares about my "-r" and "-vf fps" having no effect problems I asked a few hours ago: it was caused by using a what appears to be broken git checkout from about a month ago. i just re-compiled a current checkout and everything works as expected.
[16:02:12 CEST] <whald> *sigh*
[16:06:18 CEST] <Anaphaxeton> hello!
[16:07:14 CEST] <Anaphaxeton> i wanted to ask, can ffmpeg add srt subtitles in a file having converted them to vobsub first?
[16:08:12 CEST] <KarlFranz> I use mkvmerge for that.
[16:08:27 CEST] <KarlFranz> I think ffmpeg can, but I would have to look at it.
[16:08:41 CEST] <Anaphaxeton> i use that tool but i ve never seen that feature :-/
[16:08:53 CEST] <KarlFranz> Oh, you mean you want to do the vosub -> srt conversion with ffmpeg... I think you are out of luck.
[16:09:09 CEST] <KarlFranz> You will need tools with OCR support which, as far as I know, don't work great.
[16:09:11 CEST] <Anaphaxeton> yes
[16:09:24 CEST] <Anaphaxeton> bur mkvmerge will still do it?
[16:09:31 CEST] <Anaphaxeton> but*
[16:09:46 CEST] <KarlFranz> mkvmerge will merge any subtitle stream, but won't do the conversion.
[16:09:58 CEST] <Anaphaxeton> oh
[16:09:58 CEST] <KarlFranz> Unless you have a very good reason, I would not convert the subtitle stream.
[16:10:06 CEST] <KarlFranz> It is too much hassle for saving 3 Mb.
[16:10:12 CEST] <Anaphaxeton> i do. a crappy TV set
[16:10:26 CEST] <KarlFranz> Your TV set cannot display vosubs?
[16:10:37 CEST] <Anaphaxeton> i dont know yet
[16:10:39 CEST] <KarlFranz> vosubs are very standard.
[16:10:48 CEST] <KarlFranz> It is what DVDs have inside.
[16:10:49 CEST] <Anaphaxeton> as i said i fave srts
[16:10:57 CEST] <Anaphaxeton> yes yes
[16:12:15 CEST] <Anaphaxeton> for example i watch a movie with subs on my pc. the glyphs are so beautifully rendered by vlc... and i want that on  my tv too :(
[16:12:51 CEST] <Anaphaxeton> oops no i meant subrip -> vobsub
[16:13:08 CEST] <Anaphaxeton> no OCR needed just proper rendering
[16:14:21 CEST] <KarlFranz> Then I suppose it will be easier, but that is nothing I have ever done.
[16:14:32 CEST] <KarlFranz> ffmpeg can hardburn if you really need it.
[16:14:39 CEST] <KarlFranz> I don't like hardburning myself.
[16:15:08 CEST] <klaxa|work> nobody should :P
[16:15:45 CEST] <Anaphaxeton> indeed
[16:17:03 CEST] <klaxa|work> my (slightly outdated) ffmpeg on my server says it supports encoding dvd_subtitle (see ffmpeg -codecs to check with your version)
[16:17:13 CEST] <klaxa|work> that might be what you want
[16:17:17 CEST] <Anaphaxeton> mine too
[16:17:34 CEST] <Anaphaxeton> i am expploring it now
[16:19:49 CEST] <whald> ehm, I have to correct myself: a current git checkout does *not* fix the problem, but with version 2.5.3 it works as expected. should I file a bug report?
[16:46:06 CEST] <chungy> whald: absolutely. that's a regression
[16:46:13 CEST] <chungy> Should also see how the latest 2.7 release behaves
[16:54:12 CEST] <nyuszika7h> I'm having trouble trying to convert hdmv_pgs_subtitle to ASS / SRT
[16:55:25 CEST] <nyuszika7h> http://dpaste.com/09NWD2Y
[16:57:05 CEST] <nyuszika7h> and if I try to convert to SRT, "Error while opening encoder for output stream #0:0 - maybe incorrect parameters such as bit_rate, rate, width or height"
[16:58:18 CEST] <nyuszika7h> oh they are image subs. nvm
[17:05:56 CEST] <chungy> yea, you'll likely want some other OCR program to convert them to text.
[17:06:23 CEST] <chungy> depending on the sub, it can either be fairly quick and easy, or slow and difficult :P
[17:07:36 CEST] <andrew1> Hello! I am interested in using the audio filter 'dynaudnorm' which I see is listed in the FFmpeg Filters Documentation page at ffmpeg.org. I cannot seem to invoke the filter with any success, however. Running a simple test on the command line, e.g. 'ffmpeg -i in.wav dynaudnorm out_test.wav' returns an error indicating that FFmpeg cannot find a filter by that name. I just recently built the most recent stable
[17:07:39 CEST] <andrew1> version of FFmpeg on an Ubuntu LTS 14.40 system (2.7.2) - no dice. Any help would be appreciated!
[17:08:04 CEST] <c_14> AFAIK that filter is only in recent git head
[17:08:12 CEST] <c_14> try it with a static build
[17:08:13 CEST] <c_14> http://johnvansickle.com/ffmpeg/
[17:33:19 CEST] <sco_> that's good then
[17:38:59 CEST] <andrew1> c_14 Thanks so much, that is a great resource! Will report back with my results.
[17:56:40 CEST] <edoceo> ffmpeg really fails on this timing, shows the video is 200+ hours long
[17:56:42 CEST] <edoceo> http://edoceo.io/paste?p=WWBa9V
[18:11:44 CEST] <jdp> hey all. i'm using ffmpeg to extract sections of large mp4 files stored remotely (say on s3), using the seek flags. so on a multiple-hour movie i'm only grabbing a couple minutes toward the middle
[18:12:51 CEST] <jdp> invoked like `ffmpeg -i https://domain/folder/video -ss ... -d ... output.mp4` or somesuch
[18:13:30 CEST] <jdp> it seems like ffmpeg downloads the whole movie up until the seek time, so there is the network overhead before transcoding starts
[18:13:44 CEST] <klaxa> move -ss before -i
[18:13:47 CEST] <klaxa> and see if that helps
[18:14:00 CEST] <klaxa> also see: https://trac.ffmpeg.org/wiki/Seeking
[18:14:43 CEST] <jdp> right on i'll check it out. thanks
[18:14:56 CEST] <klaxa> if you use -ss after -i it is doing accurate seeking and needs to decode every frame until the time specified by -ss
[18:23:09 CEST] <jdp> klaxa: that did it. thanks!
[18:23:22 CEST] <kyleogrg> hi
[18:23:23 CEST] <kyleogrg> In ffmpeg, is there a difference between actual resolution and display resolution?  So let's say I encode a video a 320x240 but want it to be resized (in VLC for example) to 640x480 when it is played.
[18:23:50 CEST] <kyleogrg> Is this possible, and is there even any reason to do this?
[18:28:44 CEST] <jdp> a follow up question: is there a way to re-scale the progress reporting? i see progress climb from 0% to ~0.7% when doing the input seek and taking a couple minutes from the movie, and then the command completes
[18:29:03 CEST] <jdp> is it possible to rescale so it goes from 0 to 100%?
[19:08:38 CEST] <JEEB> kyleogrg: usually stuff like display res is just used as a display aspect ratio kind of thing in those formats that have such
[19:08:46 CEST] <JEEB> (like matroska, IIRC)
[19:09:11 CEST] <JEEB> how big the actual shown size is is completely up to the player
[19:35:19 CEST] <kyleogrg> JEEB: ok.  but i mean, if I make a 640x480 mkv, VLC will display at that size, right?  why would it change it?
[19:35:55 CEST] <kyleogrg> I know you can change display *aspect ratio*, so VLC may change the width, for example -- ie, change 720x480 to 640x480...
[19:35:58 CEST] <kyleogrg> brb
[19:38:21 CEST] <JEEB> kyleogrg: there's a whole lot of reasons why a player would change or not change the size of your video when played back. including user settings etc. and if the file has a display width/height that is one of the things used for changing that size.
[19:38:46 CEST] <JEEB> of course, even if a container does have a display width/height kind of thing - that is usually used for aspect ratio purposes in the player
[19:39:03 CEST] <JEEB> even if the field is "diplay width" or "display height"
[19:39:05 CEST] <kyleogrg> ok
[19:39:31 CEST] <JEEB> and what I meant is that for the longest time you set the display aspect ratio in matroska by setting the display resolution :P
[19:39:37 CEST] <kyleogrg> yesterday i was talking about encoding super low bitrate videos, so we were discussing the idea of shrinking video res.
[19:39:55 CEST] <JEEB> only recently did matroska get a way of setting a display aspect ratio
[19:40:01 CEST] <kyleogrg> now, what if the bitrate is so low that the res needs to be tiny?
[19:40:25 CEST] <JEEB> (which I think just changes the meaning of the display width/height fields)
[19:40:30 CEST] <kyleogrg> what if it needs to be encoded at 320x240, but i want it to be displayed at 640x480?
[19:40:33 CEST] <kyleogrg> ok
[19:40:46 CEST] <JEEB> pretty much "tough luck"
[19:40:50 CEST] <kyleogrg> yeah
[19:40:57 CEST] <JEEB> because while formats like matroska let you set a display res
[19:41:02 CEST] <kyleogrg> basically the conclusion i came up with
[19:41:08 CEST] <JEEB> that is usually interpreted as the aspect ratio
[19:41:18 CEST] <JEEB> not as the size the player should default to
[19:41:19 CEST] <kyleogrg> i see...
[19:41:59 CEST] <kyleogrg> all right.  thanks
[19:42:01 CEST] <kyleogrg> brb
[23:06:20 CEST] <sine0> c_14: are you here
[23:33:57 CEST] <ploopkazoo> I'm trying "ffmpeg -i file.mkv out.ogg" but it's transcoding the video too, how can I get it to ignore video and only transcode audio?
[23:34:20 CEST] <ploopkazoo> (ogg intended to be vorbis obviously)
[23:36:27 CEST] <grepper> ploopkazoo: -vn will ignore video
[23:37:15 CEST] <ploopkazoo> grepper: thanks
[00:00:00 CEST] --- Thu Aug  6 2015

More information about the Ffmpeg-devel-irc mailing list