[Ffmpeg-devel-irc] ffmpeg.log.20190408

[burek]

[00:01:16 CEST] <JEEB> coolhp48: it uses swresample, but it might be utilizing the non-AVFrame API
[00:01:45 CEST] <JEEB> just make sure you check the headers for AVFrame utilizing functions
[00:01:55 CEST] <coolhp48> Ok. I'll dig into the AVFrame API. Thanks again.
[02:17:13 CEST] <KombuchaKip> How robust are FLAC files to data corruption? I'm doing some testing. I flipped a bit half way in the file and ffmpeg doesn't detect any bad frames. If I corrupt more data, eventually it emits a warning about bad frames. Doesn't FLAC headers contain a checksum?
[02:19:48 CEST] <KombuchaKip> If I pass it through flac(1) via -wst it confirms data corruption.
[02:23:35 CEST] <TheAMM> https://xiph.org/flac/features.html They do
[02:23:40 CEST] <TheAMM> and https://github.com/FFmpeg/FFmpeg/blob/a0ac49e38ee1d1011c394d7be67d0f08b2281526/libavcodec/flacdec.c#L553
[02:24:11 CEST] <TheAMM> but also https://github.com/FFmpeg/FFmpeg/blob/a0ac49e38ee1d1011c394d7be67d0f08b2281526/libavcodec/flacdec.c#L611
[02:24:45 CEST] <TheAMM> '
[02:32:15 CEST] <KombuchaKip> TheAMM: Ok, so basically I need to specify AV_EF_CRCCHECK or AV_EF_COMPLIANT to something.
[02:32:28 CEST] Action: KombuchaKip tries to find where.
[03:21:55 CEST] <KombuchaKip> TheAMM: I've tried setting m_DecoderContext->err_recognition = AV_EF_EXPLODE & AV_EF_CRCCHECK after opening the codec, but before decoding any data. It still doesn't seem to detect the corrupt bit in the FLAC file.
[04:58:35 CEST] <lindylex> How do I fix these errors when compiling ffmpeg with gltransition https://pastebin.com/9WLAm901
[07:19:29 CEST] <rjp421> ffmpeg is shown on screen during the movie Glass when Samuel L Jackson is hacking a computer
[07:41:21 CEST] <circuitbone> cracking the movie to complete the mission.
[07:42:53 CEST] <circuitbone> rjp421:  Had a giggle at mr robot screens ?
[07:42:54 CEST] <circuitbone> https://www.youtube.com/watch?v=SZQz9tkEHIg
[11:32:39 CEST] <dablitz> good morning channel.
[11:33:48 CEST] <dablitz> start : --->  ffmpeg -f alsa -i hw:1,0 -acodec libopus -ar 8000 -sdp_file sdp -f rtp rtp://192.168.100.21:52000
[11:34:38 CEST] <dablitz> and when i play : ---> ffplay -nodisp rtp://192.168.100.20:52000
[11:35:41 CEST] <dablitz> it should allow the audio to stream from one computer to the next. I do get cli showing stings are happening on the sending side...but nothing on the recieving side
[11:44:15 CEST] <dablitz> ffplay just sits :  nan    :  0.000 fd=   0 aq=   0KB vq=   0KB   f=0/0
[17:42:10 CEST] <pkeroulas_> Hello, I'd like to transcode rtp raw streams to mpegts (h264 + acc + cc53). Audio and video are fine. Closed caption (EIA 608) can be decoded to output file, as scc or text but it doesn't appear in the output stream. https://pastebin.com/UvCU08uS
[17:43:11 CEST] <pkeroulas_> any idea someone, please?
[17:43:17 CEST] <DHE> that uses a distinct subtitle stream for the closed captions?
[17:46:47 CEST] <pkeroulas_> DHE, ideally yes, but it could also be inserted in SEI metadata of h264, but I know how to verify that. Vlc doesn't detect any cc/subtitles
[17:47:46 CEST] <DHE> if the subtitles are present as SEI data in the input, most h264 encoders have an option -a53cc to just copy them as SEI data to the destination
[17:49:03 CEST] <pkeroulas_> The input is not h264. It 3 "raw" rtp streams.
[17:50:59 CEST] <DHE> oops, missed the pastebin link
[17:53:19 CEST] <pkeroulas_> by the way, how to check if SEI data contains subtitles?
[17:54:21 CEST] <DHE> Stream #0:0: Video: bitpacked (UYVY / 0x59565955), yuv422p10le(top first), 1920x1080, 90k tbr, 90k tbn, 90k tbc   # this line would indicate h264, and there would be a "Closed Captions" clause listed
[18:03:16 CEST] <pkeroulas_> DEH, not sure I understand. What the input video decoder, bitpacked, which  is not aware of subtitles, has to do with h264 SEI?
[18:06:32 CEST] <pkeroulas_> But my real question is: why does ffmpeg shows a distinct output stream ("Stream #0:1: Subtitle: text") but mediainfo and ffplay don't recognize that track?
[18:10:58 CEST] <JEEB> pkeroulas_: if your input has the captions as a separate stream ffmpeg.c will show them as such. if they are in a video track the command line applications have limited support because they're not dynamic
[18:12:14 CEST] <JEEB> (any ffmpeg api using application can utilize those packets though as they're properly presented as side data within the framework
[18:16:21 CEST] <aboxer__> does FFMpeg process AFD data in mpeg user data ?
[18:16:30 CEST] <aboxer__> parse it
[18:23:00 CEST] <pkeroulas_> JEEB, understood. Could you please take a look at this ? https://pastebin.com/UvCU08uS  Everything looks good in ffmpeg, but ffplay and mediainfo fail at detecting the subtitle track.
[18:33:18 CEST] <DHE> the ffmpeg cli does very little with side data, which is what SEI-based closed captions are carried by
[20:26:40 CEST] <egrouse> Hey all. I am using a complex filter map to concatenate x amount of videos together and stream to rtmp. the first video plays fine - 30fps, 1x speed; however later files seem to get exponentially faster almost. The 2nd one is currently on 35fps at 1.17fps and keeps getting faster
[20:27:51 CEST] <egrouse> which eventually results after a few hours in huge video skips and dropping of 100s of frames at a time i guess as its trying to catch up with itself
[20:28:01 CEST] <egrouse> in fact, it seems to be related to VBV underflow messaging - when i get that message the speed/fps seems to go up
[20:28:49 CEST] <egrouse> bit of a vague question i'm sure but any suggestions or ideas how i could possibly counteract this!?
[20:33:28 CEST] <egrouse> https://pastebin.com/T3rFW9LQ is the current command
[20:48:40 CEST] <DHE> you have two [scaled] labels.. so there's that
[20:49:29 CEST] <egrouse> so there is
[20:49:46 CEST] <DHE> I don't think that's the cause of the problem, but it's the first error I found
[22:40:21 CEST] <pkeroulas_> I have a situation where closed caption can be written to a srt file or in a mp4 container. But to insert this cc in a distinct mpeg ts stream (not embedded in video)?
[22:41:54 CEST] <another> not sure ts supports subtitles
[22:42:05 CEST] <JEEB> it does
[22:42:06 CEST] <pkeroulas_> I tryed different codecs: srt, text, dvb, ass, no success. Mediainfo seems to detect an empty track
[22:42:36 CEST] <JEEB> pkeroulas_: not sure if  you can put closed captions into a separate stream in mpeg-ts
[22:42:43 CEST] <JEEB> usually dvb teletext is what's being utilized
[22:42:51 CEST] <JEEB> or dvb subtitles (which are images)
[22:43:19 CEST] <pkeroulas_> and text-to-bitmaps is complicated, right?
[22:43:44 CEST] <JEEB> I wouldn't say complicated, but not sure how easily you can do it with the command line app
[22:43:45 CEST] <another> nvm, it does
[22:44:15 CEST] <JEEB> I don't think we have a libzvbi encoder module (if it supports *encoding* dvb teletext)
[22:44:21 CEST] <JEEB> but in theory that way you could stick text into mpeg-ts :P
[22:44:29 CEST] <JEEB> in a way that receivers like
[22:45:00 CEST] <JEEB> but since your input is US style closed captions more often you stick those closed caption packets to video tracks
[22:45:10 CEST] <JEEB> not sure how easily or if at all the command line app supports that
[22:45:40 CEST] <JEEB> which is going to become a theme I think. since the API gives you possibilities to do a lot of stuff, but you can guess how many things are properly implemented in ffmpeg.c and friends because it has its limitations :P
[22:49:30 CEST] <pkeroulas_> "./configure --help [...] --enable-libzvbi         enable teletext support via libzvbi [no]"
[22:49:58 CEST] <JEEB> yea but I think that was decoder only
[22:50:14 CEST] <JEEB> yes
[22:50:15 CEST] <JEEB> libavcodec/libzvbi-teletextdec.c
[22:50:20 CEST] <JEEB> > teletextDEC
[22:50:52 CEST] <JEEB> you would have to add a module yourself and propose it to be added to FFmpeg if you want to create teletext subtitles :)
[22:52:43 CEST] <pkeroulas_> interesting
[22:54:06 CEST] <pkeroulas_> what about inserting the CC in video frames? It doesn't work either in my case.  -a53cc doesn't seem to help.
[22:55:18 CEST] <JEEB> pkeroulas_: you can do that with the API just fine. the command line tools work with this when the input is in the incoming video packets (because the tools don't have to do anything in that case)
[22:55:30 CEST] <JEEB> also they get passed by default in most encoders by now
[22:56:03 CEST] <JEEB> so in case of the input having the captions in a separate track, you would have to make the command line application take the packets from that stream and stick them together with decoded video frames being pushed into the encoder
[22:56:13 CEST] <JEEB> which it currently does not do
[23:02:53 CEST] <pkeroulas_> yes, JEEB I have been trying to stick the cc track to the video track, using the command line. Developing a new app on top of API sounds more complicated to me than adding libzvbi :)
[23:03:09 CEST] <pkeroulas_> thanks for your help
[23:43:51 CEST] <faLUCE> Hello. Which could be the best way to force an h264  keyframe with lavc?  InputLibAVFrame->pict_type = AV_PICTURE_TYPE_I  before send_frame() ?
[23:44:53 CEST] <DHE> basically yes
[23:59:05 CEST] <faLUCE> thanks DHE
[00:00:00 CEST] --- Tue Apr  9 2019