[Ffmpeg-devel-irc] ffmpeg.log.20130220

[burek]

[00:15] <ezekiel> saste: v4l2 capture with out-of-sync audio to MP4: http://ezekiel.endpoint.com/vid/sync-test-video1.txt and resulting video: http://ezekiel.endpoint.com/vid/sync-test-video1.mp4
[00:16] <ezekiel> I understand if you don't personally have time to troll through my logs
[00:17] <ezekiel> seems like I _may_ be a pretty good test case, though. Willing to rebuild and test.
[00:17] <saste> ezekiel, check that audio timestamps start with 0 as well
[00:17] <saste> video timestamps at least should be good
[00:17] <saste> -af ashowinfo
[00:19] <ezekiel> is it fairly significant that a 30fps capture would show: n:30 pts:1000068 pts_time:1.00007
[00:24] <ezekiel> wow - now that I'm looking at the ashowinfo output, I can see that the above is not nearly as significant as I might've imagined
[00:29] <saste> ezekiel, audio ts start from 0?
[00:30] <saste> also ffmpeg has a host of tools for dealing with timestamps
[00:30] <saste> check for example resample, setpts, asetpts filters just to name a few
[00:31] <saste> and the v4l2 ts option
[00:36] <stephanedev> i wrote some code to stream a video encoded with vp8 over rtp using ffmpeg C API and i play it using ffplay. when packets are dropped, many messages related to missed RTP packets are written to the console
[00:36] <stephanedev> i noticed those 2 in particular:
[00:36] <stephanedev> [vp8 @ 00a6e960] RTP: missed 1 packets
[00:36] <stephanedev> [NULL @ 00a6e960] RTP: missed 1 packets
[00:37] <stephanedev> i thought the first part (e.g. vp8) was extracted from AVClass and the part after '@' was an address
[00:37] <stephanedev> am i correct?
[00:46] <stephanedev> if my understanding is correct, what could explain the fact that the information extracted from the AVClass ("vp8") turn into "NULL"?
[00:47] <stephanedev> turns*
[00:47] <stephanedev> as obviously the address is the same in the 2 log messages above
[01:49] <defaultro> saste
[01:53] <ezekiel> still investigating, comparing the first 12 ashowinfo of out-of-sync git master: http://pastebin.com/ibCGHt76 with in-sync v1.0.4: http://pastebin.com/ymxh8jxK
[02:05] <ezekiel> recording audio and video using separate ffmpeg processes, then combining the videos, still totally out of sync
[02:05] <ezekiel> maybe an alsa issue, not a v4l2 issue
[02:08] <ezekiel> ffprobe seems to pretty regularly show "start: 0.032000" - which I know is mere tenths of a second...
[02:24] <stephanedev> is pthreads support required to stream over rtp?
[02:24] <stephanedev> i am asking because of this message when running ffmpeg on windows:
[02:24] <stephanedev> [udp @ 00645b80] 'circular_buffer_size' option was set but it is not supported on this build (pthread support is required)
[02:25] <stephanedev> using the zeranoe build
[06:51] <jagginess> hi
[06:51] <jagginess> there good guidelines for merging timing information from a video file to an audio one?
[13:24] <kvz> if i set bitrate, is that per channel, or do channels share that bitrate?
[13:24] <durandal_1707> share
[13:24] <kvz> durandal_1707: thank you!
[13:49] <robertbuhren> Hi there, i'm using a product which uses fmmpeg to transcode video files to a lower resolution/codec/container. I'm having the problem that audio and video is out-of sync, could this be anyhow related to running ffmpeg inside a virtual machine (qemu-kvm) I know the information is very vague, but maybe someone here has a hint for me
[13:50] <robertbuhren> the reason for asking is that the fact that it's run from a virtualized evinvironment is the only thing which differs from other setups that work well
[14:18] <creep> how do you like this? :) http://images.puszta.com/hazi_kifli/Hazi%20kifli%201000.jpg
[16:03] <Ziarkaen> I am running an ffmpeg screen capture command in a Makefile, for conveinience.  I can run the command separately with no problem, but have to use Ctrl-C to stop recording.  When this is put in the Makefile, I believe make captures the ctrl-C and exits with error.  Is there a way to pass Ctrl-C directly into FFMPEG, or another way to stop recording?
[16:22] <Ziarkaen> I have solved the problem by spawning a graphical terminal running the command, although this isn;t too nice.
[16:26] <sacarasc> What do you need the capture for?
[16:50] <teratorn> Ziarkaen: Make does not intercept SIGINT
[16:50] <teratorn> Ziarkaen: the process exits with non-zero on SIGINT, so make fails at that point
[16:51] <teratorn> Ziarkaen: you can add "| true" at the end of the command in your Makefile, probably
[16:51] <teratorn> err, || true
[17:03] <n1n0> which is the command line, first to set a certain video to be used , then to set  ffmpeg to extract frames at the original frame rate?
[17:36] <pedro_sland> should it make a difference if i switch the output order when i'm doing output to 2 files? eg ffmpeg -i file -filter:v ... out1.mp4 -filter:v ... out2.webm
[17:39] <relaxed> pedro_sland: using -map sure wouldn't hurt
[17:41] <relaxed> also -filter:v:$stream_id
[17:42] <pedro_sland> I appologise for not being more explicit. here is what i'm using http://pastebin.com/zFRtCeqE
[17:43] Action: pedro_sland try again
[17:44] <pedro_sland> i'll go to the beginning :) i am trying to convert one video into 2 suitable for the web. i'm happy with just webm and mp4.
[17:45] <pedro_sland> sure, i was going to give the story first?
[17:45] <relaxed> durandal_1707 loves stories
[17:46] <pedro_sland> i thought some context might help. @relaxed guessing that was sarcasm
[17:49] <pedro_sland> i checked my command http://pastebin.com/ThZ5WBk2
[17:50] <relaxed> not really, I just like messing with him.
[17:50] <relaxed> (or her?)
[17:51] <pedro_sland> k. my problem is that the webm freezes for a second at the beginning (at about 1s in).
[17:51] <pedro_sland> i started cutting down the parameters
[17:52] <pedro_sland> eventually i tried switching round the output files (and their associated commands)  and the problem seems to be fixed
[17:53] <pedro_sland> The difference in output is this "[Parsed_overlay_1 @ 0x2e99260] Buffer queue overflow, dropping." line only shows when it jumps (when webm is last)
[17:53] <relaxed> pedro_sland: try,  ffmpeg -i input -map 0 (first output options) $first_output -map 0 (second output options) $second_output
[17:55] <relaxed> oh yeah, remove 99% of those options and use "-t 5" for both outputs to get a 5 second sample.
[17:56] <relaxed> until you iron this out
[17:56] <pedro_sland> yep, that's what i was doing hence the first paste.
[17:56] <pedro_sland> it was fine without the filter but the filter broke it
[18:02] <durandal_1707> does it happen if you write only webm file?
[18:03] <pedro_sland> i think i tested that but i'll test again just now
[18:03] <pedro_sland> http://pastebin.com/GQYwvpBA
[18:03] <relaxed> pedro_sland: did you check out ffmpeg from an old repo? Your version is much older than mine: ffmpeg version N-49946-ga08ebf0
[18:03] <pedro_sland> oh
[18:03] <pedro_sland> hmm, i'll check
[18:03] <relaxed> git clone git://source.ffmpeg.org/ffmpeg.git ffmpeg
[18:04] <relaxed> they switched at some point
[18:04] <relaxed> once you update, look at -filter_complex in the man page.
[18:06] <pedro_sland> i've moved that repo and downloaded another. compiling now zzzz
[18:08] Action: relaxed sticks a beer under burek's nose
[18:09] <durandal_1707> relaxed: how you found out that pedro_sland is using old version?
[18:10] <relaxed> his pastebin,  http://pastebin.com/ThZ5WBk2 , shows "ffmpeg version N-38907-g39b0393"
[18:10] <pedro_sland> i downloaded a new one anyway but looking at it i don't think it is old. last commit i have is 9d9420d2 at Wed Feb 20 11:54:06 2013 +0100
[18:11] <relaxed> 11,000 commits behind
[18:11] <pedro_sland> :O
[18:12] <pedro_sland> something fishy here. .git/config says [remote "origin"] url = git://source.ffmpeg.org/ffmpeg
[18:12] <pedro_sland> nvm almost done compiling
[18:14] <durandal_1707> relaxed: how you get that number?
[18:14] <pedro_sland> latest version should say "ffmpeg version git-2013-02-20-2ccaab9" right?
[18:15] <durandal_1707> yes
[18:15] <pedro_sland> :)
[18:15] <durandal_1707> and previous build you used was just few commits older
[18:17] <pedro_sland> that's what i thought i was using - a build from earlier today but nvm
[18:17] <relaxed> durandal_1707: hmm, I thought the number in N-38907-g39b0393 was the number of commits. Am I mistaken?
[18:18] <durandal_1707> relaxed: yes you are
[18:20] <pedro_sland> i ran the broken command again and same behaviour http://pastebin.com/UqKXRBgv. checking with just webm
[18:21] <pedro_sland> that worked fine http://pastebin.com/ya30Zgvk
[18:22] <relaxed> durandal_1707: then what is it?
[18:25] <durandal_1707> relaxed: last commit hash
[18:25] <durandal_1707> (partial one)
[18:25] <relaxed> isn't this the commit hash --> g39b0393
[18:29] <durandal_1707> relaxed: without g
[18:37] <pedro_sland> relaxed: thanks for the -filter_complex hint. i imagine that this way only loads the overlay once and "movie=" loads it 2x? other than looking nicer on the command line, are there more benefits of this way?
[18:41] <relaxed> I thought it was required with 2 or more inputs, but I guess the old way still works for now.
[18:42] <pedro_sland> well, i just tested, this way works properly unlike the other way :)
[18:42] <pedro_sland> ty much for your help. i really appreciate it.
[18:43] <relaxed> unless you read the man page from time to time they'll change things right under your nose.
[18:43] <pedro_sland> :)
[18:43] <relaxed> you're welcome
[18:45] <pedro_sland> if they're going to remove that other way, does that mean that we won't be able to have different filters on different streams? not that i want that but just curious.
[18:56] <relaxed> pedro_sland: no, -filter_complex is just a better implementation for use with multiple inputs.
[18:57] <durandal_1707> s/better/red
[18:58] <relaxed> red?
[18:59] <pedro_sland> i don't get it either, unless he/she's playing with you
[19:00] <relaxed> I think she meant "rad"
[19:00] <relaxed> which it totally is
[19:00] <pedro_sland> totally. i love it already
[19:03] <pedro_sland> thanks again and bye all
[19:45] <Emmanuel_Chanel> hi
[19:45] <Emmanuel_Chanel> Someone is alive?
[19:46] <Emmanuel_Chanel> I want to encode Columbo: Friend in Deed aired on BSP. But the audio is one track with ual monaural.
[19:47] <Emmanuel_Chanel> It has only 1 Stereo sound track that has Japanese "left" and English "right". How can I split the track into two, Japanese and English?
[19:55] <relaxed> Emmanuel_Chanel: look at -map_channel in the man page
[19:57] <relaxed> Emmanuel_Chanel: there are example commands
[20:06] <someperson> anyone happen to know anything about video editing ?  particularly automated skin softening/retouching ?   just wondering what tools i need to look at
[20:07] <Emmanuel_Chanel> Yeah, I did. But I couldn't get a good result. Anyway, ok, I try. Thanks.
[20:09] <durandal_1707> Emmanuel_Chanel: use filters
[20:12] <durandal_1707> Emmanuel_Chanel: see documentation for channelsplit
[20:19] <Aaronds> Hi, I'm trying to convert a video to .mp4 using ffmpeg CLI. However, at the moment my video plays normally in Firefox but in Chrome I only get audio, no video. I'm using -pix_fmt yuv420p as I've seen on the net that this is needed for chrome, but otherwise I have no idea what is causing the problem. Can anyone help?
[20:19] <Aaronds> I'll paste command/output shortly
[20:23] <durandal_1707> Aaronds: pasted anything?
[20:23] <Aaronds> Command & output for the aforementioned question: http://pastebin.com/fE51jgVX
[20:24] <Aaronds> IRC bugged out on me briefly
[20:25] <durandal_1707> does chrome plays aac?
[20:25] <Aaronds> durandal_1707, sorry aac?
[20:25] <JEEB> yeah
[20:25] <JEEB> H.264 + AAC
[20:25] <JEEB> in mp4
[20:26] <Aaronds> I don't know... AFAIK others have done the same conversion and had everything worked, the only problem I picked up on was the yuv240p thing.
[20:26] <Aaronds> obviously H.264 is fine
[20:29] <durandal_1707> maybe it support only certain samplerates?
[20:30] <durandal_1707> it is little strange because AFAIK chrome use libavcodec
[20:30] <JEEB> but he is getting the audio :D
[20:30] <Aaronds> durandal_1707, are you able to provide a set of parameters which you'd expect to work?
[20:31] <JEEB> that should in all theory work
[20:31] <Aaronds> >_<
[20:31] <JEEB> are you testing on chromium or chrome_
[20:31] <JEEB> ?
[20:31] <Aaronds> chrome
[20:31] <JEEB> weird
[20:33] <Emmanuel_Chanel> durandal_1707: How shall I fix this command for that? ffmpeg -i ${FILENAME}.ts -t 00:03:00 -acodec libfaac -aq 100 -vcodec libx264 -preset ultrafast -crf 24 -vf hqdn3d,crop=in_w:in_h-8:0:0,scale=1280:720,unsharp -aspect 16:9 -threads 0 -map 0:0 -map 0:1 -map 0:2 -metadata:s:1 language=jpn -metadata:s:2 language=eng -async 1000 ${FILENAME}test1.mp4
[20:36] <Aaronds> Oh, the issue is with chrome... Going to try and have to use a different file format for now then -.-
[20:39] <durandal_1707> Emmanuel_Chanel: try to add "-af channelsplit" before -vf
[20:40] <Emmanuel_Chanel> ok.
[20:42] <Emmanuel_Chanel> I got "Cannot map stream #0:2 - unsupported type." with "ffmpeg -i ${FILENAME}.ts -t 00:03:00 -acodec libfaac -aq 100 -vcodec libx264 -preset ultrafast -crf 24 -af channelsplit -vf hqdn3d,crop=in_w:in_h-8:0:0,scale=1280:720,unsharp -aspect 16:9 -threads 0 -map 0:0 -map 0:1 -map 0:2 -metadata:s:1 language=jpn -metadata:s:2 language=eng -async 1000 ${FILENAME}test1.mp4"
[20:44] <durandal_1707> Emmanuel_Chanel: replace "-af" with "-filter_complex"
[20:48] <Emmanuel_Chanel> ffmpeg -i ${FILENAME}.ts -t 00:03:00 -acodec libfaac -aq 100 -vcodec libx264 -preset ultrafast -crf 24 -filter_complex channelsplit -vf hqdn3d,crop=in_w:in_h-8:0:0,scale=1280:720,unsharp -aspect 16:9 -threads 0 -map 0:0 -map 0:1 -map 0:2 -metadata:s:1 language=jpn -metadata:s:2 language=eng -async 1000 ${FILENAME}test1.mp4
[20:48] <Emmanuel_Chanel> says: Cannot map stream #0:2 - unsupported type.
[20:48] <Emmanuel_Chanel> ...
[20:51] <Emmanuel_Chanel> ok.
[21:10] <Emmanuel_Chanel> durandal_1707: thx. It seems that I can encode TS with Dual Monaural now.
[21:12] <durandal_1707> Emmanuel_Chanel: how? what happened?
[21:12] <Emmanuel_Chanel> It seems that -filter_complex channelsplit helps.
[21:15] <Emmanuel_Chanel> durandal_1707: Where on the document can I see channel_split , the option?
[21:20] <durandal_1707> Emmanuel_Chanel: http://ffmpeg.org/ffmpeg-filters.html
[21:24] <Emmanuel_Chanel> thx.
[22:53] <burek> beer! :D \o/
[22:53] Action: burek bows to relaxed :)
[23:00] <burek> wieshka, trouble in paradise? :)
[23:02] <burek> does anyone use ffplay extensively enough to answer this q: http://ffmpeg.gusari.org/viewtopic.php?f=26&t=826
[23:02] <ubitux> burek: hey long time no see.
[23:02] <burek> it's related to controlling ffplay programatically
[23:02] <burek> ubitux hello :)
[23:02] <burek> it was a busy holiday week :)))
[23:02] <Mavrik> holiday week? O.o
[23:03] <burek> couple of holidays merged together in a whole week :D
[23:03] <ubitux> burek: ffplay has no "slave" mode like MPlayer has
[23:03] <ubitux> burek: redirect him to MPlayer/mplayer2/mpv
[23:03] <burek> ok thanks
[23:04] <burek> is there any interesting news around here :)
[23:05] <beastd> ubitux: didn't mpv drop slave-mode? or has it something different now?
[23:05] <ubitux> http://git.videolan.org/?p=ffmpeg.git;a=blob;f=Changelog;hb=HEAD ?
[23:05] <ubitux> beastd: mmh good question
[23:06] <ubitux> beastd: yeah right seems they dropped it
[23:06] <ubitux> so MPlayer/mplayer2 then
[23:30] <burek> halp :) again :)
[23:30] <burek> "How can i change the "scan type" into interlaced? How can i change the "scan order" (Bottom Field First - Top Field First) fot interlaced encoding?"
[00:00] --- Thu Feb 21 2013