[Ffmpeg-devel-irc] ffmpeg.log.20141116

[burek]

[00:05] <Wader8> that's encoder limitation
[00:06] <Wader8> it won't give you uneven output
[00:06] <Wader8> probably goes deep into how computers work imo, can't you see everything is even, 64, 128, 1024, 4096 ... etc
[00:07] <Wader8> but im not the expert, might be another reason
[00:10] <Intruder_> It will give me 820 and 816px, but not 818? strange
[00:11] <c_14> Might have to be divisible by 4
[00:12] <Intruder_> hm okay, that might be true
[00:17] <milosz96> hi all, i have problem with ffmpeg compile on opensuse 13.2 x64. Here is some log: http://pastebin.com/ZfMx3Ai2
[01:06] <Wader8> gi again guys
[01:06] <Wader8> this doesn't work for me, says Unable to Connect
[01:07] <Wader8> ffmpeg -loglevel debug -re -i rtmp://xxx.xxx.xx:xxxxx/xxxxxxxxxxx/_definst_/ -c:v mpeg4 -vf scale=640:360 -r 29.97 -g 60 -pix_fmt yuv420p -minrate 368k -maxrate 432k -bufsize 576k -c:a copy -f rtsp -muxdelay 3 rtsp://192.168.1.101:8554/ajshow
[01:13] <MikeJoel> trying to convert an wma to mp3 - what does it mean "Queue input is backward in time"? I am getting skipping in the resulting mp3
[01:16] <MikeJoel> also: "Application provided invalid, non monotonically increasing dts to muxer in stream 0:"
[01:17] <Wader8> does ffmpeg actually create a rtsp server or it's just feeding it ?
[01:26] <Wader8> oh i just had a malformed URL, i had to add in the title as well that wasn't part of the url in the tracing but it had to be specified in rtmpdump separately
[01:26] <Wader8> works with VLC now too
[01:28] <Wader8> althought I have the same problem with a disruption after  around 59 seconds  , i think that might be the real player's issue on the nokia itself
[01:28] <Wader8> not 20 secs as i previously said
[02:41] <biofool> I have a problem that I think ffmpeg hangs while reading /dev/video0.  After 4 or 5 seconds of running it stops responding to keystrokes and the status screen stops updating
[02:44] <biofool> frame=   30 fps=7.5 q=0.0 size=      56kB time=00:00:01.00 bitrate= 458.8kbits/s dup=9 drop=0
[02:46] <biofool> frame=  107 fps=3.8 q=0.0 size=     128kB time=00:00:03.56 bitrate= 294.0kbits/s dup=76 drop=0
[02:46] <biofool> then no updates
[02:52] <biofool> http://pastie.org/private/u5l8umtgdfmavp0h3f0nw
[02:53] <sacarasc> What's the specs of the machine you're doing this on?
[02:54] <biofool> Core i5, 16G ram
[02:55] <biofool>  17:55:13 up  8:58,  5 users,  load average: 1.59, 1.33, 1.24
[02:55] <sacarasc> Something is weird, then, because doing under 4 frames per second on an i5 is slow... And probably causing it.
[02:56] <biofool> it slows down
[02:56] <sacarasc> After 30 frames, it's going at 7.5...
[02:57] <biofool> then it keeps dropping
[03:00] <biofool> strace shows it still running
[03:00] <biofool> in this loop
[03:00] <biofool> poll([{fd=4, events=POLLOUT}], 1, 100)  = 1 ([{fd=4, revents=POLLOUT}])
[03:00] <biofool> sendto(4, "fm\0\0\0\5\7\360/x\36`\0\314,H\350\331\263..85*\254\206\357wy\356w\271"..., 4096, MSG_NOSIGNAL, NULL, 0) = 4096
[03:00] <biofool> poll([{fd=4, events=POLLOUT}], 1, 100)  = 1 ([{fd=4, revents=POLLOUT}])
[03:00] <biofool> sendto(4, "\r\n", 2, MSG_NOSIGNAL, NULL, 0) = 2
[03:00] <biofool> but the time almost always shows 3.56 seconds
[03:01] <biofool> Yep frame 107 & 3.56 seconds
[03:01] <biofool> it's not frozen I just typed q and then about 5 seconds later it quite
[03:05] <biofool> strace output shows this line 7000 times in succession
[03:05] <biofool> ioctl(3, VIDIOC_DQBUF, 0x7ffff8a3b070)  = -1 EAGAIN (Resource temporarily unavailable)
[03:11] <biofool> red herring. That was an older strace
[03:12] <biofool> strace output shows this line 7000 times in succession
[03:12] <biofool> so now I dont get that issue
[03:15] <biofool> frozen at frame=  101 fps=4.4 q=0.0 size=     112kB time=00:00:03.36 bitrate= 272.5kbits/s dup=74 drop=0
[03:29] <jewt> hey folks. Having a slight problem concating mp4 files with the same codecs. There's a very slight delay between each file and I'd like it to be seamless
[03:29] <jewt> just using a very simple command structure of: ffmpeg -f concat -i list.txt -c copy output.mp4
[03:32] <biofool>  why does -vcodec libx264 cause "Unknown encoder"
[03:32] <sacarasc> Did you compile with x264 support?
[03:32] <biofool> ah because ffmpeg -encoders shows  D.V.LS h264                 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10
[03:34] <biofool> Not specifically I have ./configure  --enable-libvpx --enable-libvorbis
[03:34] <sacarasc> Then you don't.
[03:34] <jewt> So I'm curious it I can combine multiple mp4's to have no audible gaps between them
[03:36] <biofool> Ok before I go build something I don't need I"m just trying to stream my local webcam /dev/video0 to a local web page using ffserver
[03:36] <biofool> I found an example that The following command will stream media from your the system's webcam and microphone. Note that the High profile and level 4.0 is used. Also tested and working is Medium profile with level 3.1.
[03:36] <biofool> ffmpeg -f video4linux2 -s 320x240 -r 16 -i /dev/video0 -f oss -i /dev/dsp \
[03:36] <biofool> -g 52 -strict experimental -acodec aac -ab 56k -vcodec libx264 -vb 452k \
[03:36] <biofool> -profile:v high -level 40 -r 16 \
[03:36] <biofool> -f flv "rtmp://example.com:8081/publish/first?secret"
[03:37] <biofool> so I thought I'd try a different encoding to see if that helps
[03:40] <biofool> so 264lib comes from git clone git://git.videolan.org/x264.git
[03:49] <biofool> jewt: sounds like it should work, have you tried?
[10:22] <biofool> I've been trying all day to get ffmpeg and ffserver to stream the video from my webcam to html5 is this remotely possible?
[14:20] <xanal0verlordx> I encode FLV videos with x264 and FFmpeg built-in AAC encoder. Sometimes sound are not in the position the video in.
[14:20] <xanal0verlordx> It's because this AAC encoder is experimental?
[14:32] <xanal0verlordx> https://trac.ffmpeg.org/wiki/EncodingForStreamingSites
[14:32] <xanal0verlordx> Fix -video_size $(xwininfo -root | grep 'geometry' | awk '{print $2;}').
[14:32] <xanal0verlordx> Using grep and awk together is stupid.
[14:40] <relaxed> fixed
[16:11] <relaxed> llogan: I have a one liner to x11grab windows if you'd like to add it to the wiki:  x=($(xwininfo|awk '/Wid/||/Hei/||/Cor/{gsub(/+/,",");print $2}') );ffmpeg -video_size "${x[0]}"x"${x[1]}" -framerate 60 -f x11grab -i :0.0"${x[2]/,/+}" -t 30 -c:v rawvideo -pix_fmt yuv420p output.mkv
[16:36] <ubitux> relaxed: http://trac.ffmpeg.org/ticket/2563
[17:01] <relaxed> ubitux: is -x11select covered by xcbgrab's -x and -y ?
[17:02] <ubitux> it's not fixed yet :P
[17:02] <ubitux> just an idea..
[19:19] <skwaap> Hello.  I'm trying to concatenate two mp4s using the concat demuxer.  Things aren't quite working because I think the two mp4s are encoded a bit differently.  I'm willing to re-encode one of the pieces to make them match, but what exactly needs to be set in order for concatenation to work?  Thanks in advance.
[19:24] <c_14> afaik, frame rate, codec, (profile where applicable), sar, dar, resolution, audio codec, audio channels, audio sample rate
[19:25] <c_14> Some of those might be extraneous, but making them all match won't hurt.
[19:25] <skwaap> hrmm, okay
[19:25] <skwaap> does codec include codec_time_base?
[19:26] <c_14> No, frame rate includes codec_time_base (according to my subjective little organization)
[19:26] <skwaap> Okay, but regardless both codec_time_base and time_base should match for it to work?
[19:26] <c_14> I think so, yes.
[19:29] <skwaap> And what are sar and dar?  Sorry I'm somewhat new to all this.
[19:30] <c_14> Eh, Source Aspect Ratio and Display Aspect Ratio
[19:30] <c_14> (Not sure if it's source, it's something with an s though.)
[19:30] <c_14> Basically the DAR is 16:9 or 4:3 or whatever.
[19:30] <skwaap> ahh, looks like sample
[19:30] <c_14> And the SAR is either 1 or something fucked up.
[19:30] <skwaap> from the ffprobe output
[19:30] <skwaap> 0:1
[19:31] <skwaap> I think?   Anyway, looks as if for my two mp4s everything matches except time_base  (based on the ffprobe -show_streams output).
[19:31] <skwaap> is there a way to change just the time_base without changing the codec_time_base?
[19:33] <c_14> Hmm, try `ffmpeg -i file -copytb 0 -c copy outfile'
[19:37] <skwaap> hmm
[19:37] <skwaap> I still seem to get the same time base
[19:37] <skwaap> the first video is at 1/30000
[19:37] <skwaap> the second at 1/90000
[19:38] <c_14> with -copytb 1 ?
[19:38] <skwaap> oh, with -copytb 0
[19:38] <c_14> Nah, I meant "Can you retry with -copytb 1?"
[19:38] <c_14> But was too lazy to write the whole thing.
[19:38] <skwaap> oh, I see.  I'll try
[19:38] <skwaap> all good
[19:41] <skwaap> looks like it's the same.
[19:41] <c_14> Damn, was worth a shot though.
[19:43] <skwaap> yeah, thanks.
[19:43] <skwaap> At any rate, it seems like if I can just figure out how to change the time_base without changing the codec_time_base then I'll likely be in business.
[19:43] <c_14> Try reencoding with -r 'fps'
[19:43] <skwaap> everything else seems to match
[19:43] <skwaap> okay
[19:43] <c_14> ffmpeg might pick something suitable
[19:45] <skwaap> hrmm, still has the same time base
[19:47] <c_14> Try reencoding with -copytb 0
[19:47] <skwaap> okay
[19:50] <skwaap> still the same
[19:50] <c_14> :(
[19:50] <c_14> Do the files have audio streams?
[19:50] <skwaap> yes, they do
[19:51] <skwaap> they're both stream copied
[19:51] <skwaap> so they concatenate fine
[19:51] <skwaap> the result of the concatenation
[19:51] <skwaap> is that the first piece of the video plays fine
[19:51] <skwaap> then the second piece doesn't have video, but the audio continues just fine
[19:52] <skwaap> the first piece was re-encoded from one mp4, and it has the time_base 1/30000.  I used the libx264 codec.
[19:52] <skwaap> The second piece was just stream copied.
[19:52] <c_14> Try the reencoding with -copytb 0 -an
[19:52] <skwaap> and it has a time_base of 1/90000.
[19:52] <skwaap> okay
[19:52] <c_14> If that doesn't work, I'll need to take a look at the ffprobe outputs
[19:54] <skwaap> okay, sure.
[19:54] <skwaap> Same as before.  This time the one I used -an on doesn't have an audio stream.
[19:54] <skwaap> but time_base is the same
[19:54] <skwaap> Where should I paste ffprobe outputs?
[19:55] <c_14> Some sort of pastebin
[19:55] <c_14> sprunge.us ix.io pastie.org pastebin.com etc
[19:55] <skwaap> kk.
[19:56] <xanal0verlordx> Why does my video sometimes blink?
[19:57] <xanal0verlordx> http://ovrload.ru/t/40125_test.flv
[19:58] <xanal0verlordx> But it normalizes when there is some move on screen.
[20:02] <skwaap> cs_14: http://pastebin.com/q2HjbAU9
[20:02] <skwaap> Those are the ffprobe outputs with -show_streams for the two pieces.
[20:07] <skwaap> also looks like the bitrate on the first one is lower.
[20:08] <c_14> Bitrate shouldn't matter.
[20:08] <xanal0verlordx> It's because of zerolatency tune, I think.
[20:09] <xanal0verlordx> Changed it to film, all looks good.
[20:09] <c_14> skwaap: try muxing to mkv with no audio: ffmpeg -i mp4 -c:v copy -an out.mkv
[20:09] <c_14> Just want to overwrite that pesky value.
[20:11] <skwaap> when I do that to either one, it ends up with time_base 1/1000
[20:11] <skwaap> so that might work
[20:14] <c_14> You'll have to concat the audio and mux it back in later though.
[20:15] <c_14> Unless it also works without -an
[20:15] <skwaap> that shouldn't be too bad. I'm still having issues concatenating, however. Let me play with this a bit, though.
[20:15] <skwaap> well, it gives the same time_base without -an
[20:16] <skwaap> but when I concatenate I get the same artifacting in the video
[20:18] <skwaap> wait, this is with -an.  No audio.  I'll take a look at it a bit more, not sure why it's not concatenating right.  I'm using ffmpeg -f concat -i inputs.txt output.mp4  The inputs.txt file just has the two mkv files.  I'll take another look at the ffprobe outputs on the mkv files.
[20:20] <skwaap> The only differences I see are has_b_frames 1 versus has_b_frames 2.
[20:20] <skwaap> and start time, etc.
[20:21] <skwaap> I get this message when I concatenate:  Non-monotonous DTS in output stream 0:0; previous: 0, current: 0; changing to 1. This may result in incorrect timestamps in the output file.
[20:22] <c_14> That _should_ be fine normally.
[20:24] <skwaap> hrmmm.  weird.
[20:24] <skwaap> well, I'll play with it some more on my own.  At least now I can get time_base to match with mkv.  Might be something stupid I'm doing on my end.  Thanks a lot for the help!
[20:28] <xanal0verlordx> c_14: why does screen blink on still images using zerolatency tune?
[20:29] <xanal0verlordx> On x264, ofc.
[20:30] <c_14> I don't know the exact reason, but the zerolatency tune can reduce quality.
[20:33] <c_14> It basically kills the lookahead.
[20:33] <c_14> Why killing the lookahead causes those distortions is beyond me.
[20:53] <xanal0verlordx> c_14: what should I use for fast encoding and good quality?
[20:53] <xanal0verlordx> In games.
[20:54] <c_14> Do you notice any latency without the zerolatency tune?
[20:54] <xanal0verlordx> No.
[20:55] <xanal0verlordx> And what's the default tune in x264?
[20:55] <JEEB> none
[20:55] <JEEB> the defaults should be generally fine
[21:09] <iive> without lookahead you cannot average bitrate, so you spend too much bitrate on easy scenes and too little on hard ones
[21:16] <xanal0verlordx> It works.
[21:16] <xanal0verlordx> Thanks all.
[21:17] <xanal0verlordx> Now I need a streaming service.
[21:17] <xanal0verlordx> Twitch lags a lot.
[21:17] <xanal0verlordx> hitbox too.
[21:35] <LanDi> ffmpeg -i "$FILE" -acodec libmp3lame "$FILE".mp3 >>log.txt
[21:35] <LanDi> why this does not work ?
[21:36] <JEEB> the log part?
[21:36] <JEEB> use &> instead of >
[21:36] <LanDi> instead of the output on the screen ... I just want to save it to the log file
[21:36] <LanDi> yep
[21:36] <JEEB> so you get both stdout and stderr
[21:36] <JEEB> ffmpeg in general outputs most of its stuff into stderr so that you can use stdout for piping
[21:36] <JEEB> common thing with command line video encoding things
[21:37] <polysics> hello!
[21:37] <LanDi> JEEB: so, I can not do that ?
[21:37] <polysics> quick question: is there any way in FFmpeg to change pitch without changing speed?
[21:37] <JEEB> what?
[21:37] <JEEB> I just told you to also redirect stderr
[21:37] <polysics> ie. the pitch filter in SoX
[21:37] <polysics> I haven't found that
[21:37] <JEEB> just stderr is 2> and both stdout and stderr together are &>
[21:37] <JEEB> :P
[21:38] <LanDi> JEEB: I'm not very good at bash,.... I just started yesterday
[21:38] <LanDi> link me please for some wiki thing
[21:38] <JEEB> > redirects stdout (the default, no number or anything)
[21:38] <JEEB> 2> redirects stderr
[21:38] <JEEB> &> redirects both stdout and stderr
[21:39] <JEEB> and if you want it to append instead of clear the file, then you add another > just like you've done in your example :P
[21:39] <LanDi> hmmm let me try
[21:41] <LanDi> haha
[21:41] <LanDi> JEEB: great !
[21:41] <LanDi> thanks man =)
[21:44] <xanal0verlordx> LanDi: because ffmpeg writes all into stderr.
[21:47] <LanDi> I got it
[21:47] <xanal0verlordx> I see, yes. Came here 3 minutes ago.
[21:50] <SlimG> How do I stream the URL at http://pastebin.com/RndJCf2T to a local file _with_ subtitles? ffmpeg says there are 3 streams: audio, video and unknown. I suspect the "unknown" stream is the subtitles.
[22:04] <xanal0verlordx> >This paste has been removed!
[22:31] <SlimG> exit
[23:08] <brainopia> Does ffmpeg support interleaving for mp4? Can't find any way to achieve this except using mp4box
[23:21] <Chocola4> brainopia, you mean faststart ?
[23:25] <brainopia> faststart will only move moov atom upfront
[23:26] <brainopia> Chocola4: but my problem is related to a big size of moov atom (it slows down start of streaming)
[23:26] <brainopia> I was looking for solution to separate moov or shrink it somehow
[23:27] <brainopia> I've tried fragmentation but chrome won't play fragmented mp4
[23:27] <brainopia> I've heard about interleaving but it seems I was mistaken it has nothing to do with moov size
[23:28] <brainopia> so the last thing I can do is to shrink moov size somehow
[23:31] <brainopia> the CTTS atom in my files are pretty big, is there a trade-off to get rid of them?
[23:53] <brainopia> another question: cfr + maxrate are supported with each other?
[23:55] <JEEB> you mean crf?
[23:55] <JEEB> if so, yes. but you also need to provide bufsize
[23:55] <JEEB> because maxrate is always calculated within a bufsize
[23:56] <JEEB> match the bufsize between your client/player and the encoder, and you shall not need to buffer ever again :)
[23:56] <JEEB> (unless x264 screams at you hard and tells you it was forced to break VBV, although by now that is very, very rare)
[23:57] <JEEB> and if your player or whatever sets or lets you set the amount of buffer in seconds, then you use maxrate*<amount of seconds>. For example, "two seconds" worth of buffer with maxrate of 2000 is 4000
[00:00] --- Mon Nov 17 2014