Recherche avancée

Sur d’autres sites (11413)

• How to append fMP4 chunks to SourceBuffer ?

  24 octobre 2020, par Stefan Falk

  I have finally managed to create an fMP4 but now I am not able to seek or play the file depending on what I do in the file.

  


  On my backend I am taking the file and convert it to MP4 or fragmented MP4.

  


  The file gets send to the clients chunk-wise but this approach does not seem to work as it used to work on Chrome (bot not on Firefox) when using MP3.

  


  How I played MP3

  


  Say we have a 10 seconds track that is 1 MB in size which I want to start playing from second five. I want to load chunks of 1 second.

  


  Thus, I have offset = 5 / 10 * file_size and chunkSize = 1 / 10 * file_size`.

  


  With this I just started loading the MP3-file at an offset of 0.5 MB and loaded the chunks as needed where each chunk was 0.1 MB in size.

  


  This worked because before actually playing the file, I loaded the first bytes of the file and appended it to the SourceBuffer as well s.t. it was able to load the meta-information of the file. However, this approach is just not working for fMP4.

  


  What I tried with fMP4

  


  So, I have been converting MP3 to fMP4 with the MP3-approach ..

  


  .. using +dash (can play but not seek)

  


  ffmpeg -i input.mp3 -acodec aac -b:a 256k -f mp4 -movflags +dash output.mp4


  


  .. using frag_keyframe+empty_moov (cannot play on Chrome)

  


  ffmpeg -i input.mp3 -acodec aac -b:a 256k -f mp4 -movflags frag_keyframe+empty_moov output.mp4


  


  On the client the chunks get appended to a SourceBuffer (as explained above) after creating it with the Mime-Type audio/mp4; codecs="mp4a.40.2" :

  


  this.sourceBuffer = this.mediaSource
                        .addSourceBuffer('audio/mp4; codecs="mp4a.40.2"');


  


  and

  


  private appendSegment = (chunk) => {
  try {
    this.sourceBuffer.appendBuffer(chunk);
  } catch {
    return;
  }
}


  


  The problem is that I can only play the +dash converted file if I start reading it from the start and continue adding chunks.

  


  However, if I start reading the file from further down, the audio gets never played.

  


  playTrack(track, 0.0);  // Start at second 0 works
playTrack(track, 10.0); // Start at second 10 does not work


  


• FFMPEG Encoding in Multiple resoultions for adaptive streaming

  9 janvier 2020, par thatman

  I am using the following ffmpeg script for encoding mp4 video into different resolutions for adaptive HLS/DASH streaming :

   ffmpeg -y -nostdin -loglevel error -i INPUT.mp4 \
  
      -map 0:v:0  -map 0:v:0 -map 0:v:0  -map 0:v:0  -map 0:v:0  -map 0:v:0 -map 0:a\?:0  \
  
      -maxrate:v:0 350k -bufsize:v:0 700k -c:v:0 libx264 -filter:v:0 "scale=320:-2"  \
  
      -maxrate:v:1 1000k -bufsize:v:1 2000k -c:v:1 libx264 -filter:v:1 "scale=640:-2"  \
  
      -maxrate:v:2 3000k -bufsize:v:2 6000k -c:v:2 libx264 -filter:v:2 "scale=1280:-2" \
  
      -maxrate:v:3 300k -bufsize:v:3 600k -c:v:3 libvpx-vp9 -filter:v:3 "scale=320:-2"  \
  
      -maxrate:v:4 1088k -bufsize:v:4 2176k -c:v:4 libvpx-vp9 -filter:v:4 "scale=640:-2"  \
  
      -maxrate:v:5 1500k -bufsize:v:5 3000k -c:v:5 libvpx-vp9 -filter:v:5 "scale=1280:-2"  \
  
      -use_timeline 1  -use_template 1 -adaptation_sets "id=0,streams=v  id=1,streams=a" \
  
      -threads 8 -seg_duration 5 -hls_init_time 1 -hls_time 5 -hls_playlist true -f dash OUTPUT.mpd

  But the script is giving this error :

  Only ’-vf scale=320:640’ read, ignoring remaining -vf options : Use ’,’ to separate filters
  Only ’-vf scale=640:1280’ read, ignoring remaining -vf options : Use ’,’ to separate filters
  Only ’-af (null)’ read, ignoring remaining -af options : Use ’,’ to separate filters

  Please help in resolving the issue. Thanks in advance !

• How To Implement FFMPEG LL-HLS

  8 septembre 2022, par Devin Dixon

  How is Low Latency HLS achieved with FFMPEG ? From my understanding thus far, I am seeing changes around the -f option. For example :

  


  -f dash -method PUT http://example.com/live/manifest.mpd


  


  But there isn't much information researching on LL-HLS with ffmpeg. Making smaller segments I am finding comes at the cost of choppiness in the stream. Has anyone done this ? And is the protocol actually adopted or just in "theory".