[Ffmpeg-devel-irc] ffmpeg.log.20181109

burek burek021 at gmail.com
Sat Nov 10 03:05:02 EET 2018

[00:08:35 CET] <fella> ffmpeg -i MOV.SUF1 -i SND.SUF2 -map 0:0 -map 0:1 -map 1:0 -filter_complex "[0]volume=0.5[some_name];[some_name]amix" OUTPUT.SUF1
[00:31:17 CET] <petecouture> When compiling ffmpeg from source with rtmpdump, do you need the full rtmpdump library or can you just use the included librtmp ver 2.4?
[00:33:42 CET] <JEEB> I think librtmp is what FFmpeg utilizes, but check yourself. for the record, there's an internal rtmp implementation as well in libavformat for RTMP(E)
[00:37:52 CET] <petecouture> @JEEB is that why the latest compliation build no longer include rtmp dump in them?
[00:38:14 CET] <JEEB> I have no idea, there are no official compiles
[00:38:25 CET] <JEEB> FFmpeg only provides source code of FFmpeg
[00:38:47 CET] <JEEB> I think there are some 3rd party builds linked on the site, but those are 100% under the control of those providing those
[00:39:04 CET] <petecouture> I spent the night learning rewriting the librtmp build so it would work with the ffmpeg command line compile.
[00:39:16 CET] <furq> i assume librtmp still does something that the internal rtmp stuff doesn't
[00:39:18 CET] <JEEB> also the internal support for RTMP(E) is a rather old thing.
[00:39:22 CET] <furq> otherwise someone ought to have got rid of it by now
[00:39:28 CET] <furq> but nobody has ever been able to tell me what that is
[00:39:33 CET] <petecouture> It used to be included but I wondered why it was removed in the latest compliation script ffmpeg.org provides
[00:39:46 CET] <JEEB> I don't think we provide compilation scripts?
[00:39:54 CET] <JEEB> other than possibly something on the community wiki
[00:40:13 CET] <JEEB> furq: mostly what I've seen is that the protocols' AVOptions don't match
[00:40:21 CET] <petecouture> Let me pull up the link
[00:40:36 CET] <JEEB> so people cannot just use the same command line with ffmpeg.c, for example
[00:40:42 CET] <petecouture> https://trac.ffmpeg.org/wiki/CompilationGuide/Centos
[00:40:44 CET] <furq> that's a pretty lame reason
[00:40:53 CET] <JEEB> yes, that is 100% community based
[00:40:55 CET] <petecouture> That guide used to include a librtmp build I believe
[00:40:57 CET] <furq> there have been at least two major versions since then
[00:41:36 CET] <JEEB> the only other thing is that there's those DMCA'd half-v10 supporting code dumps
[00:41:43 CET] <JEEB> but I am not sure how many actually use that
[00:42:01 CET] <petecouture> :-/
[00:42:03 CET] <furq> well that's an actual feature so that's ok
[00:42:17 CET] <JEEB> it's not in an official rtmpdump release or the repo
[00:42:25 CET] <furq> nice
[00:42:28 CET] <JEEB> that's what I mean with "how many actually use that"
[00:42:44 CET] <JEEB> since most people who build rtmpdump don't generally go searching for some shady repos
[00:43:09 CET] <petecouture> I think youll find a bit of people new to ffmpeg that might try
[00:43:49 CET] <petecouture> Are there shady repos of it?
[00:43:52 CET] <furq> yeah configure could make it clearer that rtmp is builtin anyway
[00:43:54 CET] <JEEB> usually no, those who go for that stuff tend to know more or less either what they need or that they need some specific less-official version of rtmpdump
[00:44:21 CET] <JEEB> the nr1 reason for wanting to build with librtmp that I've seen is that people know the AVOptions
[00:44:24 CET] <JEEB> and guides are made around them
[00:44:34 CET] <JEEB> compared to the internal RTMP thing
[00:44:48 CET] <furq> based on this channel, the number one reason seems to be that people didn't know about internal rtmp
[00:44:52 CET] <petecouture> So building with RTMP is more powerful then using the built in library?
[00:44:57 CET] <petecouture> lol
[00:45:31 CET] <JEEB> furq: ok, yes. that's the nr1 thing, and then those who try the "native" RTMP support attempt to use the same option names and get stuck or something
[00:45:38 CET] <petecouture> I was going to fork the rtmpdump repo and with the changes I made to the build script so it can be compiled in
[00:45:45 CET] <furq> petecouture: if you don't know that you need librtmp then you probably don't
[00:45:54 CET] <furq> but yeah you might need to rewrite your command
[00:46:00 CET] <petecouture> I know enough to be dangerous
[00:46:30 CET] <petecouture> Im used to building with it for the raspi arm
[00:47:19 CET] <petecouture> Can I pull up the difference between using the librtmp and the internal libraries?
[00:47:24 CET] <petecouture> Just to do a compair
[00:47:58 CET] <furq> https://ffmpeg.org/ffmpeg-protocols.html#rtmp
[00:48:00 CET] <furq> that's the internal one
[00:48:13 CET] <furq> i don't think the librtmp stuff is documented on site
[00:48:35 CET] <furq> oh nvm it's just not sorted properly
[00:48:38 CET] <furq> https://ffmpeg.org/ffmpeg-protocols.html#librtmp-rtmp_002c-rtmpe_002c-rtmps_002c-rtmpt_002c-rtmpte
[00:49:27 CET] <petecouture> @furq thank you!
[01:30:43 CET] <orbea> Hi, I am experiencing a crash with ffmpeg while running a single alephone scenario (Game), any ideas if this is a ffmpeg or alephone bug? Or maybe both? https://pastebin.com/YZ1qSnCC I made an alephone issue for now with more details - https://github.com/Aleph-One-Marathon/alephone/issues/121
[01:31:10 CET] <orbea> this is with the ffmpeg master (As of today)
[01:41:43 CET] <MarioMey> Hi, there. I have a movie with japanese and english audio streams. I want to add a spanish one, from another file. But I need to recode the audio stream, because by copying it, it doesn't work. Command to copy all streams is:
[01:42:23 CET] <MarioMey> ffmpeg -i movie_jap_eng.mkv -i movie_esp.avi -map 0 -map 1:1 -codec copy movie.mkv
[01:42:52 CET] <MarioMey> (taken from https://stackoverflow.com/a/44136072/3868690 )
[01:43:12 CET] <MarioMey> How should I encode only the spanish audio stream?
[01:46:51 CET] <dreamseeker> can you show us what error you get when you try to copy the Spanish audio stream?
[01:47:03 CET] <dreamseeker> like, the full command and output, in a pastebin
[01:59:13 CET] <furq> MarioMey: -c:a:0 copy -c:a:1 copy -c:a:2 aac
[01:59:15 CET] <furq> or whatever codec you need
[02:50:25 CET] <MarioMey> dreamseeker: by encoding, it shows more or less 16 errors like this:
[02:50:26 CET] <MarioMey> [matroska @ 0x56026f190960] Starting new cluster due to timestampate=9804.8kbits/s speed=26.4x
[02:50:46 CET] <MarioMey> VLC open the file, but audio is unsynch... and shows some other errors. Like:
[02:50:54 CET] <MarioMey> [00005623dca7acb0] vlcpulse audio output error: digital pass-through stream connection failure: No soportado
[02:51:03 CET] <MarioMey> [00005623dca7acb0] main audio output error: module not functional
[02:51:23 CET] <MarioMey> [00007f9200ce04a0] main decoder error: failed to create audio output
[02:51:46 CET] <MarioMey> Now, I'm encoding again with what furq said.
[02:52:15 CET] <MarioMey> I didn't know that "-c:a:0" way to do it.
[08:44:14 CET] <LigH> Hi.
[08:45:36 CET] <LigH> Is it possible to let ffmpeg cut n seconds from the end of a clip, without first discovering its length and using it explicitly as parameter? (provided the length can be determined by ffmpeg from header data)
[09:01:32 CET] <LigH> I guess --sseof is the one.
[09:02:31 CET] <skident> Hi guys, does anybody know is it ok to call function avfilter_register_all (void) from different threads and more than once?
[09:03:47 CET] <LigH> Hi ... it's quiet here. America asleep?
[09:11:37 CET] <LigH> Bye.
[12:50:13 CET] <skident> Let's ask again :)
[12:50:14 CET] <skident>  does anybody know is it ok to call function avfilter_register_all (void) from different threads and more than once?
[12:51:27 CET] <durandal_1707> yes, and note that such function have been deprecated in latest version
[12:51:30 CET] <skident> I did some tests an everything work fine, but maybe somebody knows something about it for sure?
[12:54:47 CET] <atomnuker> I think they were safe to call before they were deprecated as well given every action was locked
[16:17:09 CET] <Pinchiukas> Can I somehow suppress the printing of library information in ffprobe? I can't find an option in the manpage.
[16:18:30 CET] <TheAMM> -hide_banner
[16:19:48 CET] <TheAMM> -loglevel panic (or such) if you want to hide the rest
[16:31:31 CET] <Pinchiukas> -hide_banner doesn't hide it.
[16:31:36 CET] <Pinchiukas> It only hides the banner. :)
[18:49:27 CET] <ingodid> anyone have a good one-liner for extracting a volume table from an audio file?
[18:52:09 CET] <durandal_1707> ingodid: volume what?
[18:53:15 CET] <ingodid> a CSV (or STDOUT) table of frames and volumes
[18:53:42 CET] <ingodid> i've found something similar for video here "https://stackoverflow.com/questions/38056970/ffmpeg-txt-from-audio-levels"
[18:54:04 CET] <ingodid> just starting to wrap my head around the CLI switches
[18:55:40 CET] <ChocolateArmpits> ingodid, do you want the volume value for every sample?
[18:56:00 CET] <ingodid> wait -- that does seem to work with a WAV
[18:56:03 CET] <ingodid> indeed
[20:29:53 CET] <ingodid> now onto FFT fun :)
[21:19:17 CET] <`St0ner> i'm trying to extract OPUS audio from a YouTube WEBM file losslessly into a format playable by my iPod touch (so either ALAC or WAV). a WAV file is 42.6MB while the ALAC M4A file is 120MB. is it normal for ALAC to be triple the file size of WAV? i thought ALAC is supposed to use lossless compression
[21:21:22 CET] <Pocari> `St0ner, how did you go from the Opus embedded in the YouTube video to WAV and ALAC?
[21:22:21 CET] <Pocari> what you're saying is not possible... at most, the ALAC file's audio stream will be as big as the uncompressed WAV, only being the same in the case of uniform white noise
[21:22:30 CET] <Pocari> so maybe there's something going wrong during the conversion
[21:22:52 CET] <`St0ner> for WAV: ffmpeg.exe -i "<SourceFileName>" -acodec pcm_s16le -ar 48000 -ac 2 "<OutputPath><OutputFileName>.wav"
[21:23:11 CET] <`St0ner> for ALAC: ffmpeg.exe -i "<SourceFileName>" -acodec alac "<OutputPath><OutputFileName>.m4a"
[21:23:44 CET] <`St0ner> ffmpeg build 4.0.2 win64 static dll
[21:25:08 CET] <Pocari> `St0ner, check if both WAV and the ALAC, as well as the original Opus, have the same audio hashsum: ffmpeg -i your_file.foo -vn -f md5 -
[21:27:13 CET] <Pocari> so, since you're seemingly on Windows,     ffmpeg.exe -i your_file.foo -vn -f md5 -
[21:33:52 CET] <`St0ner> the MD5= line is different
[21:34:03 CET] <Pocari> for all three??
[21:34:41 CET] <`St0ner> sorry didnt run the original. the original and the wav match, the ALAC doesnt
[21:35:14 CET] <Pocari> right, that figures... so it's the ALAC indeed that has an issue, as the overbloated filesize indicated
[21:36:38 CET] <`St0ner> ok, and is this a problem with the ffmpeg build or should i be feeding non-default parameters into the "acodec alac" section?
[21:37:10 CET] <Pocari> `St0ner, try using the WAV then to go to the ALAC;   ffmpeg.exe -i track.wav -acodec alac track.m4a
[21:39:16 CET] <`St0ner> 31.5MB when i do it that way, which is lower than the WAV file
[21:39:23 CET] <`St0ner> so that seemed to have worked
[21:39:32 CET] <Pocari> good sign, now to make sure, md5 check that one too
[21:40:22 CET] <`St0ner> yes, md5 matches
[21:40:42 CET] <Pocari> excellent, so that's a genuine lossless copy... not too sure why the Opus one failed, tbh
[21:41:23 CET] <`St0ner> hmm. how can i avoid doing a 2step process from the youtube files?
[22:14:27 CET] <ChocolateArmpits> `St0ner, youtube-dl has some encoding options
[22:15:01 CET] <ChocolateArmpits> you can download the audio track along with it
[22:15:39 CET] <`St0ner> i ended up using AnotherGUI, converted source to WAV, then used the second pass section to convert WAV to ALAC, then used the postprocessing section to delete the WAV. seems to work fine
[22:17:01 CET] <saml> how can I write a filter like mandelbrot?
[22:17:01 CET] <Pocari> you could have also written a batch file to process the downloaded Opus file with ffmpeg, I guess
[22:17:09 CET] <saml> I want my own visualization
[22:35:56 CET] <saml> https://www.ffmpeg.org/ffmpeg-filters.html#silenceremove  how can I make it also remove video so that there won't be no a-v sync
[22:48:39 CET] <threebar> hey everyone, what's the first version of ffmpeg to support 64bit windows builds?
[23:03:29 CET] <JEEB> threebar: pretty sure it goes far enough it doesn't really matter
[23:06:02 CET] <JEEB> officially change log gained win64 related stuff @ http://git.videolan.org/?p=ffmpeg.git;a=commit;h=61b1f0f909f6adb5e1e32c8b45a692f67a5a40fb
[23:06:23 CET] <JEEB> but it clearly seemed to build post-2008?
[23:06:52 CET] <JEEB> http://git.videolan.org/?p=ffmpeg.git;a=commit;h=dcc01c06650b193c11d3021346077d663de387fb
[23:07:01 CET] <JEEB> because this is the first commit that mentions win64
[23:08:25 CET] <kepstin> win64 stuff wasn't really usable until vista or win7 time period, so around 2008 is a place in the timeline that makes sense.
[23:12:04 CET] <threebar> so, what would be the earliest usable/compilable version?
[23:13:06 CET] <JEEB> decide by yourself whatever hits your criteria http://git.videolan.org/?p=ffmpeg.git&a=search&h=HEAD&st=commit&s=win64
[23:13:17 CET] <JEEB> clearly it was compilable (?) in 2008 already
[23:14:04 CET] <JEEB> I know around 2010 people started to want to use it more
[23:14:14 CET] <JEEB> which is when the change log entry appeared
[23:16:04 CET] <kepstin> win 7 came out in 2009, so that would make sense as being when people started actually caring - upgrades from 32bit xp to 64bit win7. (vista was kinda unpopular)
[23:17:13 CET] <kepstin> if you're just going be release tarballs, i wouldn't be surprised if the 0.5 (oldest version still listed on ffmpeg.org/olddownload.html ) would compile, if you can get the right compiler versions.
[23:21:43 CET] <JEEB> a snapshot of dec, 2010 by archive.org of fate shows a gcc 4.4.4 x86_64 mingw32 test :P
[23:21:54 CET] <JEEB> https://web.archive.org/web/20101221074132/http://fate.ffmpeg.org:80/
[23:22:31 CET] <JEEB> which matches with teh win64 change log part
[23:22:45 CET] <JEEB> 727/737 tests pass
[23:23:07 CET] <durandal_1707> WOAH
[23:26:14 CET] <threebar> nice
[23:26:15 CET] <threebar> thank you!
[23:27:05 CET] <JEEB> you'd have to grep the IRC logs and the mailing list regarding whenever someone was able to test building for win64
[23:28:09 CET] <threebar> how do i match this to an ffmpeg version number>
[23:28:10 CET] <threebar> ?
[23:29:14 CET] <JEEB> https://github.com/FFmpeg/FFmpeg/commit/61b1f0f909f6adb5e1e32c8b45a692f67a5a40fb
[23:29:18 CET] <JEEB> that's the change log entry
[23:29:26 CET] <JEEB> only linking on github because it shows a list of branches
[23:29:31 CET] <JEEB> in which the commit is contained
[23:30:05 CET] <JEEB> 0.7.1 by the looks of it?
[23:30:14 CET] <JEEB> which was released like half a year later
[23:30:26 CET] <JEEB> unless those releases are not in the git repo
[23:41:17 CET] <threebar> hmm
[23:41:44 CET] <threebar> 0.7.1's ./configure says "--target-os=win64" is not a valid option for target-os
[23:42:05 CET] <JEEB> not surprising
[23:42:50 CET] <JEEB> whatever was used back then with all the configure parameters :P
[23:43:02 CET] <JEEB> I don't even know why you're attempting to build a version from 2011 in 2018
[23:43:20 CET] <threebar> getting old code to run
[23:44:11 CET] <JEEB> then you should have much better ideas regarding how old something you need is
[23:44:14 CET] <JEEB> like SONAMEs
[23:47:07 CET] <JEEB> just because it's highly unlikely what I just explained to be what you need :P
[00:00:00 CET] --- Sat Nov 10 2018

More information about the Ffmpeg-devel-irc mailing list