Recherche avancée

Sur d’autres sites (15854)

• IO.copy_stream performance in ruby

  19 juillet 2016, par stiller_leser

  I am trying to continously read a file in ruby (which is growing over time and needs to be processed in a separate process). Currently I am archiving this with the following bit of code :

  r,w = IO.pipe
  
  pid = Process.spawn('ffmpeg'+ffmpeg_args, {STDIN => r, STDERR => STDOUT})
  
  Process.detach pid
  
  while true do
  
    IO.copy_stream(open(@options['filename']), w)
  
    sleep 1
  
  end

  However - while working - I can’t imagine that this is the most performant way of doing it. An alternative would be to use the following variation :

  step = 1024*4
  
  copied = 0
  
  pid = Process.spawn('ffmpeg'+ffmpeg_args, {STDIN => r, STDERR => STDOUT})
  
  Process.detach pid
  
  while true do
  
    IO.copy_stream(open(@options['filename']), w, step, copied)
  
    copied += step
  
    sleep 1
  
  end

  which would only continously copy parts of the file (the issue here being if the step should ever overreach the end of the file). Other approaches such a simple read-file led to ffmpeg failing when there was no new data. With this solution the frames are dropped if no new data is available (which is what I need).

  Is there a better way (more performant) to archive something like that ?

  EDIT :

  Using the method proposed by @RaVeN I am now using the following code :

  open(@options['filename'], 'rb') do |stream|
  
    stream.seek(0, IO::SEEK_END)
  
    queue = INotify::Notifier.new
  
    queue.watch(@options['filename'], :modify) do
  
      w.write stream.read
  
    end
  
    queue.run
  
  end

  However now ffmpeg complaints about invalid data. Is there another method than read ?

• Calculate Bitrate from m3u8 ts files ffmpeg

  16 janvier 2017, par John

  I im using ffmpeg to connect to source and then make hls chunks (ts files) so that i can restream.

  I use following ffmpeg command to connect to source and generate hls segments :

  ffmpeg -y -nostdin -hide_banner -fflags +genpts -async 1 -i http://livestream.com/stream_name.ts -acodec copy -vcodec copy -scodec copy -f hls -hls_time 10 -hls_list_size 6 -hls_flags delete_segments /tmp/chunks/1_.m3u8

  Files are generated in /tmp/chunks/ folder ==>♦ 1_.m3u8 index file and 1_0.ts 1_1.ts 1_2.ts 1_3.ts

  So question is how can i calulate bitrate of stream ?

  If i use -f mpegts i can see from ffmpeg bitrate in terminal...but i don’t have chunks and opening stream is slow because stream is written in one file 1_.m3u8 on HDD and is growing..so this solution is not good.

  I use to calulate this way :

  When 1_1.ts is generated i read file size of 1_0.ts, read duration of 1_0.ts and calulate bitrate....let’s say that file size is : 2360904 bytes and file duration is : 12seconds) :

  I calucalte it using forumula :

  BITRATE = ((file_size / 1024) / file_duration) * 8 [kbit/s]

  So in my example this is :

  BITRATE = ((2360904 / 1024) / 12) * 8 = 1537 [kbit/s]

  Is this correct calulation and is this valid for ffmpeg ? Because in ffmpeg when i don’t use hls segmenter only -f mpegts i get bitrate around 1770kbit/s so would need to ask if this is correct way of calculating bitrate of stream or if is not how is correct way and formula...thanks.

  Also i need to use HLS segmenter from ffmpeg because this way streams opens very quick and cpu usage is lower (ram usage is 1-2MB per connection witch is good)

• Playing fragmented mp4 doesnt continue playing

  18 mai 2017, par adviner

  Using ffmpeg I create an mp4 using my video camera as the source.

  ffmpeg -f dshow -i video="Integrated Webcam":audio="Microphone (Realtek High Definition Audio)"^
  
   -g 52^
  
   -vcodec libx264 -pix_fmt yuv420p -profile:v baseline -level 3^
  
   -f mp4 -movflags empty_moov+default_base_moof+frag_keyframe^
  
   %OUTPUT%\video.mp4

  Works with IE11, Chrome and Firefox

  And my html video tags :

  <video controls="controls" autoplay="autoplay" style="width:640px;height:360px;">
  
    <source src="http://localhost/video.mp4" type="video/mp4;codecs=&quot;avc1.42E01E, mp4a.40.2&quot;"></source>
  
  </video>

  The node web server version just has src="http://localhost/" for the

  using

  I’ve tried nginx and a node version that I got from this site
  The nginx doesnt do anything special. I just basically point the root to the folder so it sees the mp4.

  The problem is it only plays however much the webserver sees when the page is loaded. The file is still continuously growing. And if I refresh the page I can see the time length is now longer since the last time.

  My question is how can I make the video tag continuously play the fragmented video data without stopping and refreshing the page (which starts from the beginning again) ?