[Ffmpeg-devel-irc] ffmpeg.log.20130903

burek burek021 at gmail.com
Wed Sep 4 02:05:01 CEST 2013


[00:59] <joelogan> 	avconv -i input.ts -q 0 -filter:v setpts=2*pts out.ts
[01:00] <joelogan> did avconv change in this regard or does that look right
[01:00] <durandal_1707> this is not avconv help channel
[01:00] <joelogan> does such a thing exist
[01:10] <joelogan> can i record internet radio with ffmpeg?
[01:14] <durandal_1707> sure....
[04:17] <jangle_> greetings all.  it seems that if I want to use libswscale, I neeed to also include libavutil/pxfmt.h in order to use the pixelformat enums required in the sws_getCachedContext call
[04:18] <jangle_> but I'm finding that after including the libswscale library, I'm getting linker errors
[04:18] <jangle_> that suggest that libswscale depends on other libraries in the ffmpeg project.
[04:26] <jangle_> its looking for av_free av_freep av_get_bits_per_pixel etc
[10:18] <kode101> i'm trying to record audio from an internet radio station
[10:18] <kode101> 		ffmpeg -i http://64.79.109.199:8010 -y -y 50 rec.mp3
[10:18] <kode101> but then i get an error such as
[10:18] <kode101>  http://64.79.109.199:8010: Invalid data found when processing input
[10:19] <viric> double y ?
[10:19] <kode101> my mistake just yping here, the second y is a -t
[10:21] <kode101> this works in avconv but i'd much rather use ffmpeg
[10:22] <viric> I can't help.
[12:56] <[bjoern]> Using the same settings, every now and then ffmpeg produces mp4 files that my Android phone cannot seek properly (say, 42min video, seeking between 00:00:00 and 00:13:00 works fine, but if you try to jump to 00:20:00 it instead plays the video from the beginning).
[12:56] <[bjoern]> What can I do to debug this problem?
[12:59] <[bjoern]> (The videos do play back fine on Android)
[13:37] <[bjoern]> It also sometimes happens that in such videos you ask the player to seek to 00:30:00 and it jumps to 00:06:00 instead.
[13:37] <cbsrobot_> [bjoern]: does it happen with ffplay too ?
[13:38] <[bjoern]> Not with vlc on Windows, going to try ffplay now
[13:39] <cbsrobot_> try with: ffplay -show_streams
[13:40] <[bjoern]> With default options ffplay gives an "error while seeking" error message when seeking to certain points with frozen video and broken audio until you seek elsewhere.
[13:43] <[bjoern]> cbsrobot_, -show_streams seems to be a ffprobe, not a ffplay parameter, my copy of ffplay does not recognize it at least.
[13:44] <[bjoern]> http://dpaste.com/1364593/ is the output of ffprobe -show_streams for a sample file with seeking problems
[13:49] <[bjoern]> ffplay does not fail to seek to the same positions as the android player though, with a couple of ~42min videos it's fairly consistently at 12min where it starts failing
[13:54] <wawrek> How can I loop a video for a given amount of frames? -loop seems to work only with images
[14:17] <cbsrobot_> [bjoern]: sorry - I meant can you test with ffprobe -show_packets ?
[14:17] <cbsrobot_> maybe ffplay -show_packets works aswell
[14:18] <[bjoern]> That generates a lot of output, is there anything to grep for or would you like me to upload it somewhere?
[14:19] <cbsrobot_> well try to seek to the problematic part and see the dts pts values
[14:22] <cbsrobot_> [bjoern]: btw what version of ffmpeg do you use ?
[14:27] <[bjoern]> "ffmpeg version N-54643-g15cee5e built on Jul 11 2013 03:29:01 with gcc 4.7.3 (GCC)"
[14:43] <[bjoern]> I've recreated the video with -an no audio to see if the problem persists and shrinken the packet dump output, and the problem does indeed persist
[14:47] <[bjoern]> cbsrobot_, with the problematic videos I can seek to 11:54 but not to 12:xx on the android phone. Going through the -show_packets dts= and pts= values in the video stream around that point, I see nothing irregular.
[15:02] <[bjoern]> Encoding only the first 14 minutes does not change anything either, I guess next I'll try skipping the beginning
[15:16] <[bjoern]> Okay, I skip the first 10 minutes and then encode the next 14. Previously 12min+ was the unseekable part, so in the new video it should not be possible to seek past 2min+ if there was something in the input video at that position that causes the problem.
[15:17] <[bjoern]> But seeking there works fine
[15:17] <[bjoern]> But again, seeking past 12min in the new video (that position corresponds to roughly 22min in the input video) is not possible in the android player
[15:42] <[bjoern]> http://forum.doom9.org/showthread.php?t=158930 is fairly similar to my problem (he says 14min, I say 12min; he says crash/freeze, I say starts-over-playback-from-beginning)
[15:47] <Guest61994> Hello every body
[15:48] <Guest61994> is somebody here who can help me with ffmpeg. I have problem with audio/video sync when converting flv from red5 to mp4. I described problem here: http://stackoverflow.com/questions/18585607/red5-1-0-2-recorded-flv-convert-to-mobile-html5-format-with-ffmpeg-av-out-of-syn
[16:02] <[bjoern]> Fairly strong correlation between my problem and SAR/DAR values http://dpaste.com/1364721/
[16:02] <[bjoern]> What is the meaning of "[SAR 1:1 DAR 39:22], SAR 180224:180219 DAR 8192:4621" for the good ones?
[16:04] <[bjoern]> (I imagine the values in [] are rounded while the trailing ratios are the exact numbers?)
[16:04] <[bjoern]> Then the question would be why the "bad" inputs do not have the trailing values
[16:05] Action: [bjoern] reads http://www.mir.com/DMG/aspect.html
[16:22] <BallsDeep> Hey guys, I have a simple question: I convert a lot DTS streams to AC-3 using "ffmpeg".  When this transcode is taking place, what format does the DTS decoder use"? Does it decode the DTS stream to PCM (RAW)  or does it take the PCM (RAW) and place it into a WAV or AIF container?
[16:29] <BallsDeep> Hey guys, I have a simple question: I convert a lot DTS streams to AC-3 using "ffmpeg".  When this transcode is taking place, what lossless format does the DTS decoder use"? Does it decode the DTS stream to PCM (RAW)  or does it take the PCM (RAW) and place it into a WAV or AIF container before it is encoded to AC-3?
[16:32] <BallsDeep> Is there a switch I can use to see what "ffmpeg" is doing while it transcodes my DTS stream to an AC-3 stream? The -v switch does not give enough detailed information on what the DTS decoder is actually doing.
[16:36] <BoR0> hi. when building ffmpeg dylib's on Mac, everything is great. however when I try to build DLL's on Windows, I get make[1]: *** No rule to make target `libavdevice/avdevice.dll', needed by `all-yes`.  Stop.
[16:37] <BoR0> static linking doesn't cause this issue. any thoughts?
[16:37] <BoR0> I get .a files but no .dll
[16:42] <BallsDeep> BoRo, I think we have 210 bots in this channel. Everyone is still sleeping.
[16:42] <spaam> BoR0: what? you need to show us some lins
[16:42] <spaam> BoR0: what? you need to show us some lines abouve that error message.
[16:42] <spaam> BoR0: use a pastebin for that
[16:43] <BallsDeep> I convert a lot DTS streams to AC-3 using "ffmpeg".  When this transcode is taking place, what lossless format does the DTS decoder use"? Does it decode the DTS stream to PCM (RAW)  or does it take the PCM (RAW) and place it into a WAV or AIF container before it is encoded to AC-3?
[16:44] <[bjoern]> BallsDeep, we've read your question a couple of times now.
[16:44] <BallsDeep> alright, thanks
[16:44] <spaam> np
[16:45] <BoR0> spaam, okay, here it is: http://pastebin.com/jTZtjAjL
[16:45] <BoR0> and this is the configure: ./configure --disable-doc --disable-ffmpeg --disable-ffprobe --disable-ffserver --disable-dxva2 --enable-runtime-cpudetect --disable-muxers --disable-bsfs --enable-openssl --disable-static --enable-shared. after that my script runs make
[17:34] <rurtle> @BoR0: I'm completely new to FFmpeg. However, from the log, it seems that you don't have pkg-config on your machine or not added to the path for some reason.
[17:35] <BoR0> unfortunately I tried with pkg-config as well, but no result
[17:36] <rurtle> Hmmm, weird
[17:36] <[bjoern]> Interestingly 2^14 / (23.98/second) is almost 12 minutes; that might be indicative of where the problem lies...
[17:52] <rurtle> Don't know if you have checked these already -  http://ffmpeg-users.933282.n4.nabble.com/Hi-ffmpeg-gurus-does-the-latest-revision-build-in-the-mingw-td935943.html & http://www.gooli.org/blog/building-ffmpeg-for-windows-with-msys-and-mingw/
[18:07] <BoR0> rurtle, thanks but yeah I already went through those
[18:23] <Leoneof> hello, what to do with this error? http://devstation.org/978
[18:24] <saste> Leoneof, you need ancient ffmpeg libraries
[18:24] <saste> some of the required symbols were dropped since a few years
[18:25] <Leoneof> saste: good, where i can find this ancient  ffpmeg lib?
[18:26] <Leoneof> sourceforge? :)
[18:26] <saste> Leoneof, http://ffmpeg.org/olddownload.html
[18:26] <saste> 0.5 or 0.6 may be good enough
[18:27] <saste> but note that we discourage the use of ancient libraries, for security reasons
[18:28] <Leoneof> :/
[18:28] <Leoneof> i need it for openwrt
[18:28] <Leoneof> thanks
[18:44] <[bjoern]> cbsrobot_, my problem, not being able to seek certain videos on android video player past ~12 minutes, appears to be related to the problematic videos having "2997003.00 tbn" while the good ones have values like "24k tbn" and "11988 tbn". Setting -r 23.98 (even though that's what they would come out with anyway) seems to fix the problem.
[19:15] <[bjoern]> And ... log(12 minutes times (2997003 per second))/log(2) is pretty much exactly 31
[20:13] <CentRookie> hi, i know it is old, but do you guys know if ogm is supported with ffmpeg 2?
[20:13] <CentRookie> i want to convert ogm files
[20:15] <klaxa> should be?
[20:15] <JEEB> as long as you don't seek you should be fine
[20:16] <JEEB> there's a bug on seeking on the trac
[20:16] <JEEB> https://trac.ffmpeg.org/ticket/2739
[20:27] <CentRookie> ok
[20:27] <BallsDeep> I convert a lot of DTS streams to AC-3 using "ffmpeg".  When this transcode is taking place, what format does the DTS decoder use"? Does it decode the DTS stream to PCM (RAW)  or does it take the PCM (RAW) and place it into a WAV or AIF container before it is encoded to AC-3?
[20:28] <CentRookie> when there an encoded video has a slow framerate and the audio seems to be moving faster than the image, but the original raw is fine, is it a problem with variable framerate?
[20:28] <CentRookie> *is
[20:29] <JEEB> unless you are trying in some weird way to put it into some CFR-only container and failing at it, it's not that
[20:29] <JEEB> VFR by itself should be OK
[20:30] <JEEB> BallsDeep, most lossy decoders output 32bit floating point PCM which is then pushed to the encoder
[20:31] <CentRookie> well i encoded a lot of videos, all to mp4 and just fine. but there are some outliers and i would like to know what makes them different
[20:31] <JEEB> can't say anything without knowing more specifics. No-one's an esper here
[20:33] <BallsDeep> Is there a switch I can use to see what "ffmpeg" is doing while it transcodes my DTS stream to an AC-3 stream? The -v switch does not give enough detailed information on what the DTS decoder is actually doing.
[20:33] <BallsDeep> [
[20:33] <BallsDeep> 2:36pm
[20:33] <BallsDeep> ]
[20:34] <CentRookie> http://pastebin.com/NW0JXPQY
[20:44] <CentRookie> found anything?
[20:44] <CentRookie> i think it might have to do with the gpac iso audio handler
[20:47] <CentRookie> ah it is the time base
[20:47] <CentRookie> is there anway to ignore the timebase from the container?
[20:50] <CentRookie> some genius set the tbn to 20000k
[20:50] <CentRookie> so 20m
[22:58] <pyBlob> I've got an instance of ffmpeg writing an mpegts-stream to stdout, this can be piped into ffplay, which works fine
[22:58] <pyBlob> how would I go for sending that stream over network and displaying it on another machine (with ffplay)
[23:14] <pyBlob> ... no one there?
[23:23] <Mavrik> pyBlob, pass a udp address as output? rtp? rtmp?
[23:23] <Mavrik> depends on which protocol you want to use and what is your use case.
[23:26] <pyBlob> I want a live stream that can be accessed from another machine ... preferably using rtp
[23:28] <Mavrik> RTP is really a PITA protocol
[23:28] <Mavrik> but you can set it as output
[23:28] <Mavrik> you'll get SDP on stdout
[23:33] <pyBlob> ... ok, now I get that ouptut: pastebin.com/n1yKS7mu
[23:36] <pyBlob> forgot to specify a port -.-
[23:44] <pyBlob> oh ... and you're right, rtp doesn't work correctly, using just udp is easier ._.
[23:48] <joske> hello. i have a question about packet loss. how do i let the audio decoder know i have a lost packet? i can't feed anything to avcodec_decode_audio4...
[23:49] <Mavrik> joske, you don't, you just feed it the next one
[23:49] <Mavrik> PTS will jump and that's i
[23:49] <Mavrik> t
[23:50] <joske> then the next audio frame will be LONGER?
[23:51] <joske> i'm using the Opus codec, which should recover that lost frame partlialy with data from the previous frame
[23:52] <joske> it's for a VOIP application, so waiting for the next packet means i don't have any audio to play at all
[23:55] <Plorkyeran_> passing a frame with null data/zero length may work
[23:55] <Plorkyeran_> unless it gets handled by something in lavc rather than making it to libopus
[23:55] <Mavrik> joske, why would be longer?
[23:56] <Mavrik> you TELL the decoder what the timestamp of packet is
[23:56] <Plorkyeran_> if that's the case you'd probably have to use libopus directly to use its error concealment
[23:56] <Mavrik> why would you get longer output?
[23:56] <Mavrik> (unless the decoder does something really dumb)
[23:56] <joske> because something needs to take up the time, or else it gets out of sync
[23:56] <Mavrik> and that something isn't the decoder
[23:57] <Mavrik> you'll get back decoded frame with proper timestamp
[23:57] <Mavrik> if you lost a frame, that's your problem
[23:57] <joske> yes, but if i lost the frame i have no audio to play back
[23:57] <Mavrik> decoder doesn't care as long as it can do its job ;)
[23:57] <Mavrik> tough.
[23:57] <Plorkyeran_> opus has overlapped frames with error recovery
[23:57] <joske> instead of a reconstructed frame
[23:57] <Plorkyeran_> to handle dropped packets
[23:57] <Mavrik> Plorkyeran_, that'll be handled internally by decoder then.
[23:58] <Plorkyeran_> you have to explicitly pass null packets to libopus
[23:58] <Plorkyeran_> in place of missing packets
[23:58] <joske> according to Opus documentation: "Lost packets can be replaced with loss concealment by calling the decoder with a null pointer and zero length for the missing packet."
[23:58] <joske> which i can't do via ffmpeg, as it will segfault
[23:59] <Plorkyeran_> lavc's wrapper for libopus is dreadfully simple
[23:59] <Plorkyeran_> if you know that the audio will always be opus then just using libopus directly is probably not a big deal
[00:00] --- Wed Sep  4 2013


More information about the Ffmpeg-devel-irc mailing list