Recherche avancée

Sur d’autres sites (5888)

• FFMPEG Custom Read function reads all the data

  15 juillet 2014, par Hafedh Haouala

  I’m trying to implement a custom read function for ffmpeg that will retrieve a buffer from local video ( from device in the future) and then deocde this buffer, etc..

  So, here’s my read function

  int IORead(void *opaque, uint8_t *buf, int buf_size)
  
  {
  
  FileReader* datrec = (FileReader*)opaque;
  
  int ret = datrec->Read(buf, buf_size);
  
  return ret;
  
  }

  As for the FileReader :

   class FileReader { 
  
   protected:
  
    int fd;
  
   public:
  
    FileReader(const char *filename){ //, int buf_size){
  
        fd = open(filename, O_RDONLY);
  
        };
  
  
  
   ~FileReader() {
  
         close(fd);
  
       };
  
  
  
   int Read(uint8_t *buf, int buf_size){
  
     int len = read(fd, buf, buf_size);
  
     return len;
  
        };
  
   };

  and for the my execution :

  FileReader *receiver = new FileReader("/sdcard/clip.ts");
  
  
  
  AVFormatContext *avFormatContextPtr = NULL;
  
  this->iobuffer = (unsigned char*) av_malloc(4096 + FF_INPUT_BUFFER_PADDING_SIZE);
  
  avFormatContextPtr = avformat_alloc_context();
  
  avFormatContextPtr->pb = avio_alloc_context(this->iobuffer, 4096, 0, receiver, IORead, NULL, NULL);
  
  avFormatContextPtr->pb->seekable    = 0;
  
  
  
  int err = avformat_open_input(&avFormatContextPtr, "", NULL, NULL) ;
  
  if( err != 0)
  
   {...}
  
  // Decoding process
  
    {...}

  However, once the avformat_open_input() is called, the read function IORead is called and keeps reading the file clip.ts until it reaches its end and only then it exit and the decoding process is reached with no data to decode ( as all of it was consumed)

  I don’t know what is the problem especially that this code

   AVFormatContext *avFormatContextPtr = NULL;
  
   int err = avformat_open_input(&avFormatContextPtr, "/sdcard/clip.ts", NULL, NULL) ;

  isn’t blocking untill the end of the file is reached.

  Am I missing something ?
  I appreciate your help.

• how to create an http live stream ( broadcast ) from a segmented mp4 without sound defects

  18 août 2020, par user3621958

  tl ;dr - concatenating independent short videos to an hls playlist creates sound defects. A 'Tick' sound at the concatenation point. How do I avoid it ?

  



  I have a process that emits a 4 second MOV video every 3 seconds.
Those 4s videos are segments of a longer video. But they are completely independent. 
I want to be able to play the Longer video before it exists.
In order to achieve that I'm creating an HLS even playlist
hls event playlist
I first convert the first 4s video to hls using ffmpeg
by running :

  



  fmpeg -y -i  output0.mov -c:a aac -b:a 128000 -r:a 44100 -c:v libx264  -x264-params keyint=150:scenecut=0  -hls_flags omit_endlist -hls_playlist_type -b:t 800k -maxrate 2000k -bufsize 1200k -f hls -pix_fmt yuv420p -g 5 -hls_time 5 one.m3u8


  



  this will create an .ts file and and an .m3u8 playlist

  



  #EXT-X-PLAYLIST-TYPE:EVENT
#EXT-X-TARGETDURATION:4
#EXT-X-VERSION:4
#EXT-X-MEDIA-SEQUENCE:0
#EXTINF:4.00,
part1.ts


  



  notice the : -hls_flags omit_endlist -hls_playlist_type event flags. They make sure the playlist is an EVENT playlist and that it has no EXT-X-ENDLIST at the end of the list ( this makes the player to start polling the .m3u8 file in order to "wait" for new .ts chunks to be appended to the playlist )

  



  When the next 4s video is available I'm running the ffmpeg command again.
and then "merge" the new .m3u8 file to the old .m3u8 file. 
by appending the new .ts file to the playlist.

  



  #EXT-X-PLAYLIST-TYPE:EVENT
#EXT-X-TARGETDURATION:4
#EXT-X-VERSION:4
#EXT-X-MEDIA-SEQUENCE:0
#EXTINF:4.00,
part1.ts <- first 4 seconds chunk
#EXTINF:4.00,
part2.ts <- second ( new ) 4 second chunk


  



  since the player is polling the .m3u8 file , when its updated it fetches the second .ts file (part2.ts)
and plays it.

  



  The only issue is that sound defect. exactly at the concatenation point of the two videos. ( at second 4 ) in this example.
The video has no visible defects.
I suspect this happens because of an encoding problem. ( every mov video is encoded to hls separately ) 
But I must play the video after 4 seconds. ( can't wait for the whole duration to be ready )

  



  What am I doing wrong ? How to encode the hls chunks to avoid the sound defect ?

  



  In ffmpeg documentation I saw :
Segment the input file, and create an M3U8 live playlist (can be used as live HLS source) :

  



  ffmpeg -re -i in.mkv -codec copy -map 0 -f segment -segment_list playlist.m3u8 \
-segment_list_flags +live -segment_time 10 out%03d.mkv


  



  My problem is that i don't have the entire video. Its generated chunk by chunk.

  


• Révision 99151 : Fix https://core.spip.net/issues/3520 : déplier les blocs des referers uniquement...

  13 août 2016, par brunobergot@gmail.com

  Opera est un peu concon car il déclenche l’event onclick si on utilise le clic milieu