[Ffmpeg-devel-irc] ffmpeg.log.20120806

[burek]

[00:49] <nilsge> hello. I am on Archlinux, ffmpeg 1:0.11.1 and why does this command, with a plain png and 44100/16bit wave gives me the following error:
[00:49] <nilsge> ffmpeg -shortest -y -loop 1 -i test.png -i out.wav -acodec copy -vcodec libx264 result.mp4
[00:49] <nilsge> [mp4 @ 0x139fd00] Tag [1][0][0][0]/0x00000001 incompatible with output codec id '65536' ([0][0][0][0])
[00:55] <nilsge> yes
[00:55] <nilsge> in the meantime I switched to webm and ogg, not at least it is doing something
[00:56] <nilsge> not -> now
[00:58] <nilsge> I can't scroll up in my terminal :|
[00:59] <llogan> welcome to 1981.
[01:01] <nilsge> now, http://pastebin.com/30nNJKcK
[01:02] <llogan> nilsge: i guess wav in mp4 is verboten.
[01:03] <nilsge> flac as well
[01:04] <nilsge> webms message was very clear:  "[webm @ 0x733d00] Only VP8 video and Vorbis audio are supported for WebM.
[01:07] <llogan> nilsge: why now use .mkv?
[01:07] <llogan> *not
[01:08] <nilsge> it was intended as youtube upload. mp3 was the wrong choice anyway since google now favours webm and there is no conversion time involved after upload
[01:08] <llogan> news to me
[01:09] <nilsge> as in "that is wrong" or really?
[01:09] <llogan> i'm skeptical that it would not alter user submitted content
[01:09] <nilsge> it checks and converts, yes. But it is not wild conversion. If you upload the wrong format it can take hours.
[01:09] <nilsge> now it was seconds.
[01:10] <llogan> where is the documentation that it favors webm?
[01:11] <nilsge> no docs, that is the trend google does at the moment. Since it is their own format and they want to push it
[01:11] <nilsge> source: several news articles over the last few years since they bought it and made it free
[01:11] <nilsge> VP8 that is
[01:11] <llogan> that does not mean that webm inputs are "better"
[01:12] <nilsge> look, I don't know what google does. This is all guessing. What I know for sure is that the conversion time depends on the format you send there
[01:12] <llogan> general recommendation here is to give it the highest quality that is practical to upload since it's going to be re-encoded anyway. that usually means that people end up using libx264.
[01:30] <nox-> moin, anyone know the new syntax for streams like applehttp+http://www.nasa.gov/multimedia/nasatv/NTV-Public-IPS.m3u8 ?
[02:34] <TBK> hi, where can i find an ffmpeg version that contain the pkg-config files ?
[02:34] <TBK> cuz it is not there in the latest release
[02:35] <llogan> nox-: http://pastebin.com/WFbyqRwC
[02:36] <llogan> the bare url didn't work for me with ffplay, but s/applehttp/hls/ did
[02:36] <nox-> ah cool
[02:36] <nox-> thx :)
[02:38] <nox-> <hypnocat> nox-: "hls+http://www.nasa.gov/multimedia/nasatv/NTV-Public-IPS.m3u8: No such file or directory"
[02:39] <hypnocat> i'm here, btw
[02:39] <llogan> hls+http://www.nasa.gov/multimedia/nasatv/NTV-Public-IPS.m3u8
[02:40] <llogan> hmm. it looks the same. worksforme.
[02:40] <hypnocat> what version of ffmpeg are you using?
[02:41] <llogan> 0.11.1 (i was testing something from arch repo). 99% of the time i use Git head.
[02:49] <llogan> works with Git head too
[02:56] <hypnocat> llogan: http://pastie.org/4397004
[03:05] <nox-> maybe your ffmpeg version is too new to know applehttp but too old to know hls?
[03:05] <nox-> is that possible?
[03:07] <llogan> hypnocat: i don't know. try more recent ffmpeg. what's your distro?
[03:07] <hypnocat> gentoo
[03:07] <hypnocat> i'm running 0.10.3 and the newest on gentoo is 0.10.4
[03:08] <hypnocat> i could also try compiling by hand from git..
[03:08] <llogan> does gentoo not have a "ffmpeg-git" build script?
[03:08] <llogan> doesn't matter. you can easily compile and test without installing
[03:09] <llogan> git clone git://source.ffmpeg.org/ffmpeg.git ffmpeg && cd ffmpeg && ./configure && make
[03:09] <hypnocat> thanks
[03:09] <llogan> ./ffmpeg -i hls+http://www.nasa.gov/multimedia/nasatv/NTV-Public-IPS.m3u8
[03:10] <llogan> git clone --depth 1 git://source.ff...
[03:10] <llogan> since you don't need full Git history
[04:01] <hypnocat> llogan: i just finished compiling ffmpeg from git, ran it as you suggested, and got this error:  http://pastie.org/4397292
[04:01] <hypnocat> ahh.. looks like i had to use ffplay, not ffmpeg
[04:02] <hypnocat> that works
[04:02] <hypnocat> thanks!
[04:02] <nox-> :)
[06:11] <Ruxton> I'm using Jon Severinsson's PPA for ffmpeg on 10.04 is there any way to get AAC encoding? seems like that and extras don't include libfaac? and using just 'aac' returns '[NULL @ 0x1f4ccc0] Codec is experimental but experimental codecs are not enabled, see -strict -2'
[06:25] <llogan> Ruxton: add "-strict experimental"
[06:26] <Ruxton> ahh thank you :) i thought that was a compiler setting ;P
[06:27] <Ruxton> how experimental is this aac encoder? I'm using it in a batch processing queue
[06:28] <llogan> it's probably not as bad as people assume at higher bitrates (=>128k)
[06:28] <llogan> you may also want to monkey with -cutoff with this encoder
[06:28] <llogan> http://d.hatena.ne.jp/kamedo2/20120729
[06:29] <Ruxton> thanks llogan, appreciate the assistance
[06:29] <llogan> otherwise you can compile ffmpeg and use fdk-aac or libfaac.
[06:30] <llogan> or pipe to an external encoder and mux the video and audio together.
[06:31] <llogan> ffmpeg -i input -f wav - | faac -q 100 -o output.mp4 -
[06:31] <llogan> see https://ffmpeg.org/trac/ffmpeg/wiki/UbuntuCompilationGuide for compile directions
[06:32] <llogan> lots of options.
[06:32] Action: llogan drones on and on...
[12:41] <Kay-Z> Anyone awake this morning ?
[12:42] <Kay-Z> I'm trying to figure out what the internal name of the "copy" codec is. ie: if I specify -vcodec copy what is the internal api name of the codec used ? since copy doesn't appear in the enum of codecs ...
[12:43] <burek> did you check the source code?
[12:46] <Kay-Z> burek: yeah, I guess I'm not looking in the right place. I've been reading through libavcodec/ but I havne't found anything. any pointers ?
[12:51] <Kay-Z> specifically I looked through avcodec.h pretty closely
[12:51] <Kay-Z> where it Enum's CodecID
[12:52] <Kay-Z> but copy isn't one of the options. Yet it's valid for the command line
[12:52] <burek> # grep -nRH '"copy"'
[12:53] <Kay-Z> lol, did that. Lots of noise wrt copyright
[12:53] <burek> no you didn't
[12:53] <burek> if you did what I wrote, you wouldn't get that noise
[12:54] <Kay-Z> your right, I didn't do exactly that grep but I did grep.
[12:54] <burek> ffmpeg.c, line 4823
[12:54] <Kay-Z> yours was much more precise
[12:54] <Kay-Z> mind enlightening me what H and n do ?
[12:54] <burek> man grep? :)
[12:57] <Kay-Z> so i'm not normally a C guy
[12:57] <Kay-Z> but as I read that
[12:57] <Kay-Z> if the codec is "copy" it uses "stream_copy"
[12:59] <burek> that's what it says :)
[13:00] <Kay-Z> but stream_copy isn't a codec either...
[13:00] <burek> it seems you lack the basic understanding of what 'copy' actually means
[13:00] <burek> it means "don't touch my streams, they are ok, I just want to remux"
[13:00] <Kay-Z> ok.
[13:00] <burek> so, why would you need encoder?
[13:00] <burek> or decoder for that matter?
[13:01] <Kay-Z> I'm trying to replicate the command line options using the C api so I have a frame grabber and a recorder and I'm trying to set the videoCodec and the AudioCodec
[13:02] <Kay-Z> just trying to find the right options
[13:02] <burek> that's not the proper way to do it
[13:02] <Kay-Z> ah.
[13:02] <Kay-Z> what is ?
[13:02] <burek> read the ffmpeg.c
[13:02] <burek> and see
[13:02] <Kay-Z> care to narrow that 5000+ line file down any ?
[13:03] <burek> I'm not a developer :)
[13:03] <burek> so, you might try asking in #ffmpeg-devel, but be patient
[13:04] <Kay-Z> :(
[13:48] <Pyriel0> burek are you here?
[14:08] <Pyriel0> anyone know how to flush the rtbuffer after the picture is sent to a pipe?
[14:15] <MMlosh> Hmmm....   Can anyone help me with joining files?  I tried the mpeg-ts way, but all I am getting is 'Application provided invalid, non monotonically increasing dts to muxer in stream 0' for each frame of the 2nd part
[15:20] <MMlosh> I take my help request back...  Even though the joined file cannot be played, ffmpeg takes it as input just fine
[15:25] <GA-Flix> Hi everyone, could anyone explain shortly what "--enable-static" and "--enable-shared" means? (which one has which results)
[15:44] <iive> GA-Flix: static would build .a libraries for libavcodec/format/util etc. the resulting ffmpeg binary would contain these libraries code in full. It would probably incorporate other .a libraries if they are installed (system ones etc).
[15:46] <iive> the shared build would produce .so libraries. The resulting ffmpeg binary would be seeking the .so files in the filesystem to be loaded before execution starts (linux kernel call ld.so loader to do that). You will need all files to use it.
[15:47] <iive> shared libraries are preferred when more than one program uses them. Given what most foss players use them it is good to have them.
[15:49] <GA-Flix> iive, thank you a lot =D
[15:51] <GA-Flix> i did not see the difference between static and shared yet :/ ffmpeg is a bit too complex for me ^^
[17:23] <Pyriel0> burek i have a quick question for you iwhen you have time
[17:23] <burek> yes, sure, go ahead :)
[17:24] <Pyriel0> i was wondering if there is a command to flush the buffer after a frame has been encoded and sent to the pipe
[17:24] <burek> did you ask in #ffmpeg-devel ?
[17:24] <burek> that's kinda advanced thing
[17:24] <Pyriel0> thought maybe something like that could work around the problem since i have no clue how to fix it
[17:24] <burek> not really covered with cmd line options
[17:24] <Pyriel0> ah =/
[17:25] <Pyriel0> i only use cmd line, anything more advanced would be useless for me
[17:25] <Pyriel0> i guess ill just wait and hopefully someone works on / figures it out from the ticket
[17:26] <emerica> Is there something I could change setting wise in ffmpeg when creating mp4's to have the size of the moov atom to become smaller? Running into a problem with some devices having issues when a moov atom is larger than 6mb. When doing long movies 1.5-2 hours we start to break that 6mb. The mdia and trak atoms get rather large. Any ideas ?
[17:27] <burek> Pyriel0, a lot of people are using -f flv rtmp://... without problems
[17:28] <burek> so you might spend your time
[17:28] <burek> on google, searching for a lot of cmd line examples
[17:28] <burek> and try it one by one
[17:28] <burek> one must work
[17:28] <burek> :)
[17:28] <Pyriel0> i have already
[17:28] <Pyriel0> its most likely a windows thing somehow
[17:29] <Pyriel0> i think most people on windows just use programs so they don't run into this
[17:29] <burek> well, install oracle's virtualbox
[17:30] <Pyriel0> looking it up
[17:30] <burek> and just install debian, without gui or anything, just the bare bone shell
[17:30] <burek> and try from within linux to see if the bug will go away
[17:30] <burek> if it does
[17:30] <ubitux> emerica: you could skip iods but i don't think that's really going to help; maybe you could feature request the support for cmov (zlib compressed moov)
[17:30] <burek> then you can let know the people behind Zeranoe's build (on the forum) that it's not an ffmpeg's issue (probably)
[17:31] <ubitux> also, some tables allow compressed values somehow, not sure if this can't be reduced as well; it will need some code most likely
[17:33] <emerica> running the same file through mp4box take the moov size in about half, I'll have to see if that plays and then try to figure out whats different
[17:41] <Pyriel0_> got virtual box and downloading edbian now
[17:41] <Pyriel0_> can't believe its 8 dvd's
[17:42] <brontosaurusrex> JEEB, when starting mplayer from bash, is there a way to catch if exit was user pressing esc or usual EOF ?
[17:42] <brontosaurusrex> i know its not ffmpeg ...
[17:51] <Pyriel0_> burek could i just use ubuntu in virtualbox instead of debian?
[17:55] <burek> yes
[17:55] <burek> just avoid installing desktop/gui
[17:55] <burek> to shorten the time
[18:35] <Pyriel0_> i have ubuntu installed and i can use the internet through it now
[18:36] <Pyriel0_> i feel lost using this lol =)
[18:51] <Pyriel0_> ok i have ubuntu running and your lastest build dl'd / extracted now
[18:51] <Pyriel0_> do i have the windows ffmpeg stream to udp and have the ubuntu one pick it up to pass through rtmp?
[20:23] <VooDooNOFX_> Is there a difference in the way DNxHD is created using the Linux and OSX versions of ffmpeg? I'm seeing an odd banding pattern in a mov created on linux, but not when created on a similarly built ffmpeg on osx.
[20:24] <Mavrik_> hmm
[20:24] <Mavrik_> VooDooNOFX_, some people that build ffmpeg disable MMX for some reason on ffmpeg
[20:24] <Mavrik_> that could cause difference in resampling algorithms
[20:26] <VooDooNOFX_> In this case, i've tested 0.10 release and git head from yesterday. Both do the same thing
[20:28] <VooDooNOFX_> Also, the banding issues disappear when I generate a DNxHD 115 or 175. They seem to only appear in dnxhd 36, and only on linux render nodes
[20:29] <Mavrik_> VooDooNOFX_, that's pretty much ready for a bug ticket I'd say
[20:29] <VooDooNOFX_> Yeah, me too :(
[20:30] <VooDooNOFX_> I'll have to do some more tests on non-copyrighted images so I can post em
[20:32] <llogan> burek: thanks!
[20:33] <burek> :beer: :)
[20:56] <MMlosh> ubitux, I tried  "-vf select='gte(n,\2)',yadif" as you suggested, but it resulted in reported video length 26h51m38s most of the video was fine and with a correct time information, but the end suddenly jumped over the 26h timestamp and lost audio...     Hmm... tried swapping them.. The resulting reported length is still wrong :(
[20:59] <ubitux> MMlosh: can you pastebin the full command line and output?
[21:02] <MMlosh> the output file?  hmmm...  not the input one?
[21:02] <sacarasc> The text output ffmpeg gives.
[21:03] <MMlosh> ah...  ooops
[21:03] <MMlosh> the command line was     -i "$fn" -vcodec libx264 -preset medium -crf 26 -vf yadif,select='gte(n\,2)' -acodec copy -sn "$to"
[21:04] <MMlosh> I tried swapping yadif and select, but it didn't change anything
[21:07] <MMlosh> hmm... It seems I used avconv as ubuntu told me to...  I can retry with ffmpeg if you think that will help...   btw: the commandline worked on the rest of my files....  but wait, they were mpeg-ts made by dvbcut, not mpeg-ps as this one is
[21:09] <llogan> MMlosh: many of us here are not familiar with avconv or its associated "ffmpeg" since they are not from FFmpeg.
[21:11] <MMlosh> There are two lines like this one..  That's probably the reason... [mpeg @ 0xb649c0] Invalid timestamps stream=0, pts=61413325, dts=61416925, size=1537
[21:11] <MMlosh> if someone want to browse the output, here: http://fpaste.org/oPIj/
[21:14] <ubitux> MMlosh: yes just please try the latest FFmpeg and paste the full output
[21:14] <MMlosh> llogan, I know...    avconv does not seem different to me, I forgot it may be way more different to you...  So unless that timestamp error means something to you, thanks for you help
[21:14] <ubitux> the libavfilter really differs
[21:14] <MMlosh> I assume the same will happen if I drop the filter
[21:14] <llogan> oops.
[21:15] <llogan> https://ffmpeg.org/trac/ffmpeg/wiki/UbuntuCompilationGuide
[21:15] <llogan> there
[21:15] <ubitux> not sure that's really recent
[21:15] <ubitux> maybe burek's static build could help to do some quick testing
[21:16] <ubitux> try this MMlosh
[21:16] <MMlosh> Tried dropping the "select", the reported length still sky-rockets..  does not feel like a filter issue
[21:17] <llogan> we can't really give much help with libav products.
[21:17] <llogan> that's the first issue.
[21:17] <ubitux> and please paste the uncut output
[21:17] <ubitux> but well avconv output is different anyway
[21:18] <ubitux> please just use a recent ffmpeg, or maybe ask #libav if you insist on using avconv; we can't really support it here
[21:18] <MMlosh> ah... well... I was unable to find their channel easily  (and though it's just a new name, as ubuntu suggested)
[21:18] <JEEB> in libav
[21:19] <MMlosh> thanks for being so supportive anyway...
[21:19] <JEEB> they basically rewrote parts of ffmpeg and named it avconv
[21:19] <JEEB> and then deprecated it
[21:19] <JEEB> the version in current ubuntu still has the ffmpeg binary, which wasn't updated with the avconv rewrites
[21:19] <JEEB> ffmpeg (the project) merged many of the avconv changes into its ffmpeg
[21:19] <JEEB> anyways, libav support is over at #libav
[21:20] <ubitux> s/many/almost everything/
[21:20] <JEEB> aye
[21:21] <MMlosh> I don't wonder why they say it's a "downstream project", if you put it that way, it sounds like that
[21:22] <MMlosh> conclusion: the file was broken, I am sure it's not a converter error..
[21:22] <ubitux> MMlosh: if the file is available somewhere i can try with a recent ffmpeg
[21:22] <ubitux> if that's too much trouble for you to upgrade
[21:23] <llogan> the static builds can't be made much easier to try
[21:24] <ubitux> llogan: sometimes unwillingness is stronger ;)
[21:24] <MMlosh> the file is quite big, ubitux ..  400M  but I can put that up temporarily, no issue but upload speed
[21:25] <ubitux> not available in the internet pony world?
[21:25] <MMlosh> no, it's a recording
[21:25] <MMlosh> regular tv recording, with broken packets as it happens
[21:26] <MMlosh> Sure, i am able to get the footage via the internet, but for no reason I prefer to record them myself
[21:27] <ubitux> then you'd better try the static build ;)
[21:27] <ubitux> and pastebin the full uncut command line and output
[21:29] <MMlosh> ok, I'll try
[21:37] <MMlosh> observed so far - only one "concealing DC AC MV errors" message... and better performance due to full usage of both cores
[21:39] <MMlosh> now pretty much all errors popped up as with avconv, including those timestamp errors
[21:43] <MMlosh> Hmm... I would expect any program to automatically stop the runtime-updated state information if it's redirected into a file...  (or at least putting that to stdout so it can be separated)
[21:44] <MMlosh> ubitux, huh, the timestamps didn't skyrocket that much (still went over 2h for a 20min video), but as the ffmpeg finished, it was corrected...        summary: Okay, your ffmpeg is definitely better
[21:45] <ubitux> if you still think something can/need to be fixed, please describe it then :)
[21:46] <MMlosh> only separating messages between stdout and stderr
[21:47] <MMlosh> (or telling people who are told to provide logs to adjust loglevel)
[21:47] <ubitux> to provides log, you have the -report option; otherwise -loglevel might help, and your shell as well :p
[21:48] <MMlosh> my shell just saved every byte that passed
[21:48] <MMlosh> with the ^M after each "status report" line
[21:52] <MMlosh> ubitux, Do you want me to upload the log, even though it went "fine" ?    (the "length" was wrong when I played the result before it was finished, that's the only thing that might have been better)
[21:52] <ubitux> no it's okay
[21:52] <ubitux> ;)
[21:57] <MMlosh> Thanks for help...   I wonder what would ubuntu-shipped ffmpeg do, if they shipped it  (read - as old as shipped avconv)
[22:37] <msmithng> is there any reason why print_report uses \r rather than \n?
[22:37] <msmithng> av_log(NULL, AV_LOG_INFO, "%s    \r", buf);
[23:04] <llogan> burek: your fflogger pastebin shows "!ubuntu - provides a link to ubuntu compilation guide" but it points to the PPA
[23:06] <burek> just a sec
[23:06] <burek> hmh
[23:06] <burek> wasn't me :D
[23:09] <burek> llogan better? :)
[23:13] <llogan> burek: LGTM. thanks again.
[23:14] <burek> np :)
[23:49] <ylluminate> on recent builds there is no longer any `libx264-medium.ffpreset` file provided
[23:50] <ylluminate> it appears as though these have been eliminated.
[23:52] <ylluminate> we have been hitherto running `ffmpeg -r 6 -i watermarked.jpg -vcodec libx264 -vpre medium -vpre baseline -crf 24 -threads 0 -acodec libfaac -aq 100 -s hd1080 output.mp4`
[23:52] <ylluminate> which obviously fails now since there's no "medium" preset for libx264
[23:54] <JEEB> -profile:v baseline
[23:54] <JEEB> -preset medium (which is default)
[23:54] <JEEB> also I think -threads 0 is default by now for libx264
[23:54] <JEEB> presets are applied as libx264 would apply them internally
[23:54] <JEEB> http://mewiki.project357.com/wiki/X264_Settings#preset
[23:57] <JEEB> also you might want to look into https://github.com/mstorsjo/fdk-aac instead of faac for AAC
[23:57] <JEEB> fraunhofer's encoder, open sourced
[00:00] --- Tue Aug  7 2012