Recherche avancée

Sur d’autres sites (9066)

• Gstreamer RTSP Server over UDP artifacts on video when there is a packet loss

  28 août 2023, par Usama

  I want to have a RTSP server which should work on top of UDP , but when I tried to use Gstreamer RTSP server its giving video artifact/glitches when there is a packet loss.

  


  I have tried to follow this example for RTSP re-transmission.

  


  gst_rtsp_media_factory_set_profiles (factory, GST_RTSP_PROFILE_AVPF);

  


  When the profile is set like above , ffmpeg / mplayer / VLC is not able to play and says that it is an unsupported protocol. I don't know which player supports this protocol, only the gstreamer player is able to play the video.

  


  When the gst_rtsp_media_factory_set_profiles line is removed then all the players are working but there is a lot of artifact on the video if there is packet loss.

  


  everything is working fine on local env but on cloud there are artifact because of packet loss

  


  I also tried making retransission_time to 10 sec but it did not work.

  


  gst_rtsp_media_factory_set_retransmission_time (factory, 10000 * GST_MSECOND);

  


  Tried enabling RTCP as required for NACK (did not work i can see RTCP messages in wireshark )

  


  gst_rtsp_media_factory_set_enable_rtcp(factory , 1);

  


  Tried setting retransmission true (Did not worked )

  


  gst_rtsp_media_factory_set_do_retransmission(factory,1);

  


  I still get the artifacts / glitches on the video some time video just stuck on the client side

  


  I can confirm that its due to packet loss because when i switch the RTSP to use TCP instead of UDP everything works fine.

  


• Online RTMP radio with Python + FFmpeg + Nginx : unable to have a continuous output stream

  2 décembre 2023, par Igor Longo

  Thinking about an application similar to an online radio, I'm trying to transmit a sequence of audio files chosen by the user, where he chooses which file will be played in the sequence, sufficiently in advance for the reproduction not to be muted. I'm using FFmpeg in a Python script and an RTMP server running Nginx.
The problem is that interrupting the transmission between one file and another appears to be breaking the transmission on output. Sometimes files are skipped and other times it simply crashes for the client.

  


  I have already tried transmitting the sequence of files in real time, using '-re', and I have also tried not using '-re', creating a custom queue manager in Python with the intention of trying to keep only 'n' files in the server buffer .
Below is an excerpt of the code :

  


  ffmpeg_command = [
'ffmpeg',
'-loglevel', 'warning',
# '-re', # read input at native frame rate
'-i', file_path,
'-c:a', 'aac',
'-ar', '44100',
'-bufsize', '1024k',
'-af', 'atempo=1.0',
'-b:a', '128k',
'-f', 'flv',
'-flvflags', 'no_duration_filesize',
rtmp_url # server url
]
subprocess.run(ffmpeg_command, check=True)

  


  Any tips on how I can resolve this ?

  


• How the ffmpeg is fixing Handle page termination with same page number ...? [closed]

  11 avril 2024, par adarsh

  Im working on the teletext subtitle feature and faced with subs not clearing issue.
the subtitle lines remain displayed until being replaced by a new version of the specific line. So the subtitles never disappear and often an old line is still displayed, after another, more recent line is displayed/updated.
take a look at the issue in this sample video file : https://code.videolan.org/videolan/vlc/uploads/58c2e1660e94e2df5e8cdb75edf531d9/GODFATHER__03_.ts
Issue reported : https://sourceforge.net/p/zapping/bugs/203/

  


  The player is using ffmpeg and libzvbi library.

  


  At first I suspected that the Erase Page Flag (C4) in the Page Header was not set, but the flag was indeed set (at least once between two LOPs with subtitles). After further debugging it turned out that the Teletext service in the two issues consists of just a single Teletext page - which is the reason here :

  


  The Teletext spec defines the Magazine Serial Flag (C11) in Table 2 (ETSI EN 300 706 v1.2.1) as (bold font by me) :

  


  When set to '1' the service is designated to be in Serial mode and the transmission of a page is terminated by the next page header with a different page number.
When set to '0' the service is designated to be in Parallel mode and the transmission of a page is terminated by the next page header with a different page number but the same magazine number.
The same setting shall be used for all page headers in the service.
(BTW it doesn't really matter, but in these two cases this flag is always 0)

  


  In vbi_decode_teletext in /src/packet.c, storing the page is aborted, if the condition "page terminated" in the C11 definition is not met. But as the service here consists of only one page, this condition is never met !

  


  Luckily, this issue was fixed in ffmpeg- https://github.com/FFmpeg/FFmpeg/commit/b1e0e216462a989a39e7b413aef6d32f8cedc154

  


  and also in zvbi :
https://github.com/zapping-vbi/zvbi/commit/40a6ab0200c46b67b059b5b1fb15793ce780892a

  


  I understand the root cause and fix in zvbi library, and how it is ignoring c4 flag in case the same page no.

  


  But i want to understand how the ffmpeg fix is working ?
ffmpeg fix : https://github.com/FFmpeg/FFmpeg/commit/b1e0e216462a989a39e7b413aef6d32f8cedc154

  


  what is the rule for repeated page headers ?