Recherche avancée

Sur d’autres sites (4768)

• Revision 122868 : Toujours sur ces vues de saisies fieldset. Champ extra se base sur le ...

  7 mars 2020, par Maïeul Rouquette — Log

  Toujours sur ces vues de saisies fieldset.
  Champ extra se base sur le classe vide pour ne pas afficher les saisies
  sans réponse. Le masquage est en css, contrairement à ce que je pensais
  (suis-je bête !)
  On ajoute donc la classe vide à la vie d’une saisie si l’ensemble de ses
  sous saisies sont vide.
  Permet d’éviter avec champ extra d’avoir un titre de fieldset qui se
  balade sans aucune réponse en dessous.

• mediastreamsegmenter stops sending id3 metadata using HLS

  9 mai 2013, par alawson421

  I am using a combination of ffmpegand Apple's mediastreamsegmenter and id3taggenerator to create a HLS stream with metadata (ID3) embedded in it. I have all of the applications running and am able to retrieve the metadata out of the stream on the client side, but the issue is that after what seems like a random amount of time, the client stops receiving metadata on the stream.

  Here is what I have working now :

  This is the ffmpeg and mediastreamsegmenter part :

  ffmpeg -i udp://@:5010 -c:v libx264 -crf:v 22 -b:v 500k -preset:v veryfast -c:a libfdk_aac -b:a 64k -f mpegts - | mediastreamsegmenter -b http://localhost/stream -f /usr/local/nginx/html/stream/ -t 10 -s 4 -S 1 -D -y id3 -m -M 4242 -l log.txt

  This is taking a udp stream on the localhost on port 5010 and encoding the video and audio to H.264 and AAC, respectively. It is pipping the MPEG-2 transport stream segments to the mediastreamsegmenter, which in turn is generating the .M3U8 file and associated .ts files and placing them on a NGINX webserver. The mediastreamsegmenter is also listening on port 4242 for tcp traffic from the id3taggenerator which I will show how I am using it now :

  id3taggenerator -text '{"x":"12","y":"36"}' -a localhost:4242

  As you can see, I am sending some simple x-y coordinates in JSON as text to the mediastreamsegmenter listening on port 4242. I am actually using a Python script to send some test data at the moment. Here is the script :

  #!/usr/local/bin/python3
  
  from subprocess import call
  
  from time import sleep
  
  
  
  for i in range(0, 10):
  
      for j in range(0, 10):
  
          x = str(20 * i)
  
          y = str(20 * j)
  
          print("X: " + x + " - Y: " + y)
  
          call(["id3taggenerator", "-text", "{\"x\":\"" + x + "\",\"y\":\"" + y + "\"}", "-a", "localhost:4242"])
  
          sleep(1.0)

  This script is just sending a bunch of x-y coordinates using the id3taggenerator CLI. I have adjusted the sleep amount from 0.1 to 15 and it doesn't seem to make any difference. The client application that is getting the metadata out of the stream only prints out anywhere from 4 to 35 points, but never has it ever received all of the x-y coordinates.

  I am wondering if there is a limit to the amount of metadata that can sent per frame, but I cannot seem to find anything that specifies a hard-limit value for any of these tools...

• FFmpeg - add watermark, concat files and output mp4, optimize ?

  17 août 2021, par user892134

  I use the following code, which works to :

  


  

  1. add watermark

  


  2. convert (preroll,video,postroll) to .ts

  


  3. create concat txt file from new .ts files

  


  4. create mp4 ouput

  


  


  variables

  


  $filePath = '/home/mywebsite/public_html/assets/videos/';
$convertfile = $filePath.'convert-videos-2.txt';
$watermark = $filePath.'watermark.png';
$preroll = $filePath.'preroll.mp4';
$video = $filePath.'Video.mp4';
$video_watermark = str_replace(".mp4","-watermark.mp4",$video);
$newVideo = $filePath.'Video-New.mp4';
$postroll = $filePath.'postroll.mp4';
$input1 = $filePath.'input1.ts';
$input2 = $filePath.'input2.ts';
$input3 = $filePath.'input3.ts';


  


  Add watermark to original mp4

  


  $mark = "ffmpeg -y -i '".$video."' -i '".$watermark."' -filter_complex \"overlay=10:10\" '".$video_watermark."'";
exec($mark);


  


  convert to ts (video transport stream)

  


  exec("ffmpeg -i '".$preroll."' -c copy -bsf:v h264_mp4toannexb -f mpegts '".$input1."'");
exec("ffmpeg -i '".$video_watermark."' -c copy -bsf:v h264_mp4toannexb -f mpegts '".$input2."'");
exec("ffmpeg -i '".$postroll."' -c copy -bsf:v h264_mp4toannexb -f mpegts '".$input3."'");


  


  create concatenate txt file, with new .ts files

  


  $cmd = "echo \"file '".$input1."'\n";
$cmd.="file '".$input2."'\n";
$cmd.="file '".$input3."'\" > ".$convertfile;
exec($cmd);


  


  create output mp4

  


  exec("ffmpeg -f concat -safe 0 -y -i '".$convertfile."' -c copy -bsf:a aac_adtstoasc '".$newVideo."'");


  


  This works but takes way to long. For an mp4 that is 1GB or more, can take upto 3 hours ? The watermark alone takes 1hour plus. How can i optimize this code ? combine commands ? How do i solve ?

  


  I've tried adding the watermark when generate the second .ts file below, but not working, returns this error Streamcopy requested for output stream 0:0, which is fed from a complex filtergraph. Filtering and streamcopy cannot be used together. :

  


  exec("ffmpeg -y -i '".$video."' -i '".$watermark."' -filter_complex \"overlay=10:10\" -c copy -bsf:v h264_mp4toannexb -f mpegts '".$input2."'");