[Ffmpeg-devel-irc] ffmpeg.log.20130305

burek burek021 at gmail.com
Wed Mar 6 02:05:01 CET 2013


[00:54] <barque> When I convert videos with ffmpeg I get sometimes stalls in the middle of the video
[00:54] <barque> middle of the output video
[00:54] <barque> any idea how I can tell ffmpeg to minimize erros like this?
[00:56] <barque> I'm basically downscaling 1 ogg file into another ogg file
[01:03] <barque> Hello?
[01:21] <barque> Helloooooo
[01:21] <barque> anyone ?
[01:21] <barque> anyone at all ?
[01:30] <burek> barque
[01:30] <burek> let me see what's wrong
[01:30] <barque> these are the params I use -s 768x484 -qscale 10 -r 25 -vcodec libtheora -f ogg
[01:30] <barque> -qscale 10 gives me highest quality for Theora/OGG
[01:34] <burek> pastebin...
[01:34] <barque> gimme a sec, system's tied up with something
[01:45] <barque> http://pastie.org/6385365
[01:45] <barque> I get a stutter at ~6 seconds
[01:45] <barque> a stutter that does not exist in the original video
[01:50] <michaelni> barque, "built on Feb  6 2012" <-- thats a bit old
[01:50] <barque> right, right... so any idea why it's stuttering?
[01:50] <michaelni> does this problem also exist with latest ffmpeg ?
[01:51] <barque> Haven't tried
[01:51] <barque> but shouldn't the log say anything meaningful?
[01:55] <bchapman> hey all, I'm having a problem when decoding x264 audio using ffmpeg, it adds about .02ms of delay to the audio.  If I use -ss to seek the audio to about 00:00:00.02 everything mostly lines up.  Anybody know why this is happening?
[01:55] <bchapman> Here's a paste bin of the log:
[01:55] <bchapman> http://pastebin.com/8LcdfCWe
[02:20] <burek> barque, can you update your ffmpeg?
[02:21] <burek> oh, i didnt read messages above
[02:21] <burek> bchapman, what is "x264 audio" ?
[02:21] <burek> does such thing exist?
[02:22] <bchapman> h264
[02:22] <burek> ok, what is "h264 audio" then?
[02:22] <bchapman> the audio stream on an h264 quicktime
[02:23] <burek> bchapman, try typing: ffmpeg -i <yourfile>
[02:23] <burek> and use pastebin-like site to show the output
[02:23] <bchapman> http://pastebin.com/waredcRS
[02:25] <burek> bchapman, you use: Stream #0:0(eng): Audio: pcm_s16le (sowt / 0x74776F73), 48000 Hz, stereo, s16, 1536 kb/s
[02:25] <burek> pcm audio, for short
[02:25] <burek> there is no such thing as "h264 audio"
[02:25] <burek> h264 is a video codec
[02:27] <bchapman> true
[02:27] <bchapman> most of the time I have the issue on mp4 containers.  They have aac audio encodings.  Such as this one:
[02:28] <bchapman> http://pastebin.com/0pyVPG9L
[02:28] <burek> since that MOV container already contains pcm audio, why not just using -c:a copy
[02:28] <burek> ffmpeg -i test090.mov -c:a copy test090_audio.wav
[02:29] <bchapman> yeah, that works
[02:29] <bchapman> still not working on the mp4 though
[02:29] <burek> what exactly "not working" means?
[02:30] <burek> oh you want to extract audio from mp4?
[02:30] <bchapman> correct
[02:30] <burek> ffmpeg -i test100.mp4 -c:a copy output.aac
[02:34] <bchapman> Using aac, everything lines up as far as waveforms between the source and the resulting aac.  However, I do need this as pcm.  Which when encoded is 2ms off.  Here's my command:
[02:34] <bchapman> ffmpeg -loglevel debug -y -probesize 5000000 -i test070.mp4 -c:a pcm_s16le -ar 48000 -vn test070_audio.wav
[02:34] <burek> ffmpeg -i test100.mp4 output.wav
[02:35] <bchapman> yep, that produces a waveform that doesn't match the source
[02:35] <bchapman> http://pastebin.com/HNkMgFuj
[02:37] <burek> how exactly do you know it's 2ms off
[02:39] <bchapman> looking at the waveforms in after effects
[02:41] <bchapman> http://postimage.org/image/czqxvylx1/
[02:43] <burek> can you save the audio track from mp4 (within after effects) to a wav?
[02:43] <burek> and load it again
[02:44] <bchapman> yep, just tried and it lined up
[02:44] <burek> could you type ffmpeg -i that_wav_file_from_ae.wav
[02:45] <bchapman> http://pastebin.com/u0Fatyp6
[02:47] <burek> hm, this might be a bug
[02:47] <burek> do you have time to report it on our bug tracker?
[02:48] <bchapman> sure
[02:48] <burek> thanks
[02:48] <bchapman> I've had one other issue sort of in the same boat, the first frame in the mp4 gets duplicated if I encode it to another format, do have time to look at that?
[02:49] <bchapman> video
[02:49] <burek> do you have a pastebin
[02:49] <bchapman> sure, just a sec
[02:49] <burek> ok
[02:50] <bchapman> http://pastebin.com/fSQR99Sr
[02:52] <burek> there were no errors in that log, so I assume this one is also buggy
[02:53] <bchapman> k
[02:53] <burek> if you could provide some samples with those tickets
[02:53] <bchapman> sure
[02:53] <burek> our developers would be able to recreate the issue and fix it properly
[04:46] <burek> does it sound reasonable to run 2-pass h264 encoding in such a way that the 1st pass is done with -crf and the 2nd pass is done using some -b value?
[09:30] <defaultro> hey boys, thanks for all your help. I've completed the timelapse :D It's now in Youtube, here it is - http://www.youtube.com/watch?v=LLL86Y01CMY
[09:34] <cbsrobot_> defaultro: congrats
[09:47] <defaultro> :D thanks cbsrobot_
[09:47] <defaultro> I really like it
[09:48] <defaultro> now, I need to hit the bed. it's 3am here, hahaha
[09:48] <defaultro> i learned that -crf 0 is bad
[09:48] <cbsrobot_> gn
[09:48] <defaultro> i had to reencode everything with -crf 15
[09:48] <defaultro> good night too ;) thanks for your help the other day
[12:31] <dmonjo> hi
[12:31] <dmonjo> is ffmpeg better than vlc when it comes to streaming webcam?
[12:32] <JEEB> if it's a protocol that ffmpeg can do without ffserver, then they're more or less on similar standing
[12:32] <JEEB> if you need ffserver you want to run away from it as fast as you can, with all due respect to the app
[12:34] <dmonjo> i am using icecat server VORBIS THEORA
[12:34] <JEEB> <burek> does it sound reasonable to run 2-pass h264 encoding in such a way that the 1st pass is done with -crf and the 2nd pass is done using some -b value? <- shouldn't be a problem, some video services do exactly that, too. Run a crf first pass at first (without fast first pass settings), and then if the result's too big, hit it with a set bit rate for the second pass.
[12:37] <relaxed> couldn't you achieve the same thing using -crf and -maxrate?
[12:40] <JEEB> well, this is for cases when you just want to control the average size, and then possibly have a maxrate/bufsize separate of that
[12:45] <blez> hello
[12:45] <blez> is there a way to 'predict' how big will be the output after convertion?
[12:50] <blez> I can't find the formula for this
[12:57] <relaxed> blez: Look at two pass: https://ffmpeg.org/trac/ffmpeg/wiki/x264EncodingGuide
[13:05] <blez> but I want to calculate it _before_ transcoding, just to know the output size
[13:12] <relaxed> I don't think you read it. Try again.
[13:57] <DannyZB> I'm trying to transcode all video formats quickly to mp4 ( low quality is fine ... )
[13:57] <DannyZB> I keep getting "buffer underflow" errors
[13:57] <DannyZB> ffmpeg -i /var/www/transmission_out/71c5dd14c7ddc1451f06ff8ba129868e4936a697/Family.Guy.S11E09.HDTV.x264-LOL.mp4 -an -x264opts bitrate=800:vbv-maxrate=800:vbv-bufsize=34 -vcodec libx264 -f mpegts -c:a copy -preset ultrafast /var/www/media/360p/out.mp4
[16:08] <sabton> if avcodec_decode_audio4() returns error, should discard the current AVPacket and get the next one?
[17:18] Action: Fjorgynn loops pictures
[17:19] <durandal11707> sabton: almost always yes
[17:22] <defaultro> which is correct, Magic Lantern on Canon50 or Magic Lantern in Canon50
[17:24] <defaultro> btw, I shared this last night. If you guys didn't see my post, here it is. I used ffmpeg for this video :)     http://www.youtube.com/watch?v=LLL86Y01CMY
[17:24] <durandal11707> cgi?
[17:28] <durandal11707> i think it is explanation how to do it hard way
[17:28] <durandal11707> eg, its too much work while it should be fast and easy
[17:31] <defaultro> cgi?
[17:32] <durandal11707> computer-generated imagery
[17:54] <defaultro> just got back
[17:54] <defaultro> i shot the skies with my canon 50d
[18:00] <Fjorgynn> rip the skies
[18:11] <circut> morning all, quick question. I've noticed I can run: ffmpeg -i file.mp3 -acodec libfaac file.m4a, and I get a beautifully rendered auto-detected mpeg4 audio file
[18:12] <Mavrik> lucky :P
[18:13] <circut> but when i try to send the output to stdout, I get a failed to detect audio format. I'm just wondering which format I need to be passive to '-f'. I've tried mp4 which seems the closest to what I want, but I get an error stating:
[18:13] <circut> Could not write header for output file #0 (incorrect codec parameters ?): Operation not permitted
[18:14] <Mavrik> circut, mp4 should be the right one yea
[18:14] <Mavrik> circut, can you paste full output (other error messages as well)?
[18:14] <circut> sure thing let me pastebin
[18:16] <circut> http://pastebin.ca/2328464
[18:17] <saste> defaultro, you don't need to rename the input images, check -start_number in image2 (relatively recent option)
[18:17] <Mavrik> circut, uh... yeah, I see how this could be a problem
[18:17] <saste> circut, muxer does not support non seekable output
[18:17] <circut> oh? what do you see?
[18:18] <saste> so you can't use a pipe for writing an mp4
[18:18] <circut> aw man :(
[18:18] <saste> since the muxer needs to go back and write
[18:18] <circut> i see, so the only way is to send it to a flat-file, then just read that file back?
[18:19] <Mavrik> circut, hmm, depends on what do you want to do?
[18:19] <defaultro> saste, wasn't aware. Thanks for the tip
[18:19] <Mavrik> circut, (you can also use another container if your use case allows it)
[18:20] <saste> circut, the point is, why do you want to write to stdout?
[18:20] <saste> why not ffmpeg -i ... OUT.mp4?
[18:20] <circut> well i'm trying to setup ffmpeg as an xinetd service so i can just dump a file there and read the output
[18:20] <circut> ex: cat file.mp3 | nc localhost 1234 > file.m4a
[18:21] <circut> i mean i could hack around it to make the xinetd service use a temporary file
[18:21] <circut> but i was hoping to just streamline the whole thing using pipes
[18:22] <Mavrik> circut, yeah, mp4 is a format which wasn't meant for streaming and I think ffmpeg does some seeking around while writing it :\
[18:23] <circut> i got ya
[18:23] <saste> circut, exactly, the point is "use the right tool", which is not mov in this case
[18:23] <circut> well thanks guys, I can deal with that
[18:23] <circut> saste! but m4a sounds so poppin fresh!
[19:04] <synapse_tsk> is it possible to make ffmpeg re-establish connection to a RTSP stream (tcp transport) after some time of inactivity?
[19:04] <synapse_tsk> it seems there is no timeout option for RTSP, but my webcam does not have HTTP transport
[19:05] <synapse_tsk> and it gets freezed after an hour or so, only ffmeg restart helps
[19:29] <circut> synapse_tsk: maybe sending a signal to it might help? SIGINT or something?
[19:46] <synapse_tsk> circut: ffmpeg just closes. I need it to detect that a source is timed out, and restart/quit in this case
[22:59] <pr0ph3t> cant see video on my stream in twitch.tv but audio works fine http://pastie.org/6398659
[23:06] <pr0ph3t> i posted the pastie url
[23:06] <pr0ph3t> http://pastie.org/6398659
[23:08] <durandal_1707> use -pix_fmt yuv420p as most of players sucks and do not support 444
[23:09] <pr0ph3t> and how would i change this?
[23:10] <durandal_1707> adding that argument after encoder one
[23:11] <pr0ph3t> durandal_1707: i am quite the newb at this so i am sorry in advance where would i find this
[23:12] <durandal_1707> it is explained numerous time, you can search it on web
[23:20] <ubitux> pr0ph3t: your paste doesn't provide the command line used
[23:20] <ubitux> it's a function/script/alias/whatever that gives absolutely no hints about the command line
[23:21] <ubitux> which you will have to change anyway (by adding the pix_fmt thing durandal_1707 was talking about)
[23:22] <ubitux> and for this, we can't help you since we know nothing about your setup (and we don't want to help you fixing your .bashrc or whatever file you decided to edit :p)
[23:47] <LunaVorax> Hello
[23:47] <LunaVorax> Where did the option -loop 1 go?
[23:48] <LunaVorax> I'm using FFmpeg version 1.1.3 on archlinux x86-64 and the -loop option doesn't exist neither as a ffmpeg or in the manual
[23:49] <LunaVorax> -loop_input and -loop_output does exist in the manual but it says that they are depreciated and -loop should be used instead
[00:00] --- Wed Mar  6 2013


More information about the Ffmpeg-devel-irc mailing list