[Ffmpeg-devel-irc] ffmpeg.log.20180619

[burek]

[06:32:54 CEST] <hotbobby> when you do a conversion there's a status bar. it is very noisy and it's clearing and redrawing doesnt always work in every terminal. how can i disable it?
[06:33:36 CEST] <hotbobby> my command is simply ffmpeg -y -i input.avi -f flv output.flv
[06:35:43 CEST] <hotbobby> im sorry for the scrolling but i wanted to specify im trying to silence the lines "frame=   41 q=7.0 size=     116kB time=1.6 bitrate= 579.7kbits/s"
[09:23:09 CEST] <cdscds> sbd here
[10:12:28 CEST] <hamidi> hi
[10:12:53 CEST] <hamidi> i need to know how to detect what av_read_frame returns is an I-Frame or not
[10:12:58 CEST] <hamidi> is there a way?
[10:17:48 CEST] <bencoh> hamidi: https://ffmpeg.org/doxygen/trunk/structAVFrame.html#af9920fc3fbfa347b8943ae461b50d18b ?
[10:18:39 CEST] <bencoh> (iirc)
[10:23:22 CEST] <hamidi> thx
[10:24:05 CEST] <hamidi> flags == AV_PKT_FLAG_KEY?
[10:26:39 CEST] <chrysalis> anyone know if its possible to preserve encoding options with -copy in either avidemux or ffmpeg
[10:29:35 CEST] <chrysalis> im also interested abot preserving timestamps
[10:35:28 CEST] <bencoh> hamidi: right, av_read_frame, not decode ... then probably yeah
[10:36:09 CEST] <hamidi> yeah i need to avoid decoding
[10:36:53 CEST] <hamidi> or it's better to use ((flags & AV_PKT_FLAG_KEY) != 0)?
[11:14:02 CEST] <mort> Is this flow correct? Each iteration, I create an AVPacket variable, then av_init_packet that packet, then set its data and size parameters, then avcodec_send_packet, then avcodec_receive_frame, then av_frame_unref the frame and av_free_packet the packet.
[11:14:23 CEST] <mort> I ask because I definitely seem to have a memory leak caused by an av_malloc happening as a result of avcodec_send_packet
[11:17:00 CEST] <Mavrik> mort: you should free the packet right after avcodec_send
[11:17:18 CEST] <Mavrik> Since the decoder will keep a reference (or copy data)
[11:17:24 CEST] <Mavrik> And there's no point keeping it around
[11:17:47 CEST] <mort> I see
[11:31:07 CEST] <mort> Mavrik: alas, moving the avcodec_free_packet doesn't do anything
[11:31:40 CEST] <mort> https://s.mort.coffee/d/img/scr-.png - the amount lost predictably goes up as the amount of decoded frames goes up
[11:32:09 CEST] <mort> (that line 344 is `int result = avcodec_send_packet(av_context_.get(), &packet)`)
[11:34:20 CEST] <Mavrik> um.
[11:34:26 CEST] <Mavrik> I can't find avcodec_free_packet
[11:34:33 CEST] <Mavrik> in the API, which version are you using?
[11:34:40 CEST] <Mavrik> (the call should be av_packet_unref)
[11:34:57 CEST] <Mavrik> (Also if you're mallocing the AVPacket struct, you need to free it yourself.)
[11:35:46 CEST] <mort> right, it's av_free_packet
[11:35:53 CEST] <mort> the packet struct is stack allocated
[11:38:17 CEST] <mort> huh, apparently av_free_packet is deprecated and av_packet_free is not
[11:38:34 CEST] <mort> odd, I'd have expected that to print warnings
[12:38:33 CEST] <roxlu> hi, I'm trying to use ffplay to playback from a RTSP stream and I see the /OPTIONS, /DESCRIBE, /SETUP and /PLAY messages from ffmpeg. After the /PLAY the RTP packets start flowing but ffplay doesn't start "rendering". Is there a way to see if ffplay actually receives and decodes the RTP packets?
[12:44:20 CEST] <mort> Do I have to free the AVCodec separately, in addition to freeing the AVCodecContext it's associated with?
[12:45:07 CEST] <mort> I can't imagine that'd be the case, but there's definitely memory that's not getting freed
[13:09:05 CEST] <Mavrik> mort: I strongly suggest you look at source in those cases
[13:09:11 CEST] <Mavrik> You'll quickly see what's being freed.
[13:10:21 CEST] <mort> I did, and didn't see the AVCodec being freed, but from what I know, that's not unreasonable, because the AVCodecs are initialized once and stored in a big linked list
[13:10:43 CEST] <ravi_> Hi, I am using webm_chunk muxer to generate live webm dash chunks. I need a way to get notified when each chunk write is completed from ffmpeg. I have configured, interrupt_callback in AVFormatContext. But i am not able to access its priv_data in order to know the current chunk index. Is there a way to access AVFormatContext's priv_data, so as to read current chunk index from there ?
[13:22:46 CEST] <mort> Mavrik: apparently, the exact same code I'm using, except with the h264 decoder instead of h264_v4l2m2m, leaks no memory
[14:32:24 CEST] <mort> Mavrik: it seems to very definitely be a memory leak in the h264_v4l2m2m decoder... Compare running ffmpeg on the CLI with the default software decoder vs with the h264_v4l2m2m decoder:
[14:32:34 CEST] <mort> https://p.iotek.org/m32, https://p.iotek.org/c80
[14:56:33 CEST] <Mavrik> mort: uggghhhh
[14:57:00 CEST] <mort> well, it could also potentially be a false positive from valgrind
[15:23:41 CEST] <Hello71> 651k? kinda *shrug*
[15:50:32 CEST] <Prodicus> Is it possible to extract a single channel from lossy stereo audio without re-encoding? (I imagine this would be impossible for mid-side/joint stereo but technically feasible for LR simple stereo)
[15:53:13 CEST] <mort> Hello71: 651k for a small example is not "kinda shrug" when it continuously leaks memory and I want to use ffmpeg to decode streaming video over an extended period of time
[16:02:09 CEST] <Hello71> oh, well you didn't specify that it increased
[16:02:59 CEST] <Hello71> Prodicus: man ffmpeg, /-map_channel
[16:33:53 CEST] <analogical> Prodicus, typo?
[16:34:17 CEST] <analogical> Hello71, typo?
[16:35:26 CEST] <Hello71> i what
[18:17:31 CEST] <kepstin> Prodicus: it'll be very rare to find stereo audio tracks encoded in pure L-R mode, good encoders automatically select and sometimes switch within tracks. ffmpeg has no support for doing anything like this, you'd have to find (or even build) codec specific tools
[18:21:13 CEST] <furq> will there be any lossy audio from the last decade or so that was pure l-r stereo
[18:21:18 CEST] <furq> s/was/is/
[18:23:06 CEST] <kepstin> I think if you tell lame to do really high constant bitrate mode, it might do l-r stereo
[18:23:32 CEST] <kepstin> other than that, it would only happen if you manually specified it :/
[18:23:57 CEST] <kepstin> basically no chance with any codec more modern than mp3, of course.
[19:24:29 CEST] <iive> there is option for full stereo
[19:37:38 CEST] <saml> let's build stuff
[19:46:40 CEST] <saml> i built
[19:46:42 CEST] <saml> mpv --external-file original.mp4 source.mp4 --lavfi-complex '[vid2]scale=800:-1[orig]; [vid1][orig]scale2ref[resized][orig]; [resized][orig]hstack[vo]'
[19:46:59 CEST] <saml> copied from furq's
[19:50:14 CEST] <saml> https://github.com/Argon-/mpv-stats  wtf mpv
[20:00:57 CEST] <CoreX> saml mpv have a channel on this network its #mpv
[20:03:28 CEST] <saml> nice thanks
[22:17:09 CEST] <xzj22> Could somebody assist me with some CLI for youtube streaming? I'm using code included in the pastbin like. I would like to mute the audio or only stream video, If Youtube will except only the video. https://pastebin.com/60GQRLpZ
[22:19:49 CEST] <kepstin> xzj22: youtube doesn't accept video-only streams. The normal recommendation is to include a silent audio track
[22:20:58 CEST] <xzj22> Ok, I didn't think it would like no audio. Could you give me some info on how to include a silent audio track?
[22:21:32 CEST] <kepstin> xzj22: probably easiest to to replace the alsa audio input with a filter that generates silence, e.g. "-f lavfi -i aevalsrc=0"
[22:23:11 CEST] <xzj22> I'll give that a try, Thanks for your response.
[00:00:00 CEST] --- Wed Jun 20 2018