Recherche avancée

Sur d’autres sites (7113)

• FFmpeg : Is it better to make a application then using ffmepg directly

  14 juillet 2020, par Mirabeau

  For all my IP camera streams, I use, under Linux, ffmpeg by a bash script which allows me at the same time to manage :

  


  1/ Convert the RTSP streams to HTTP (HLS) for the "Live" and therefore which generates "m3u8" file and *.ts segments

  


  2/ Backup the stream for archiving (in increments of 5 minutes), a cron remove older files (xx days)

  


  ffmpeg -i "rtsp://[IP_CAM01]" -rtsp_transport tcp -c copy -map 0 -f segment -segment_time 300 -segment_atclocktime 1 -segment_format mkv "cam01-% 03d.mkv" -c copy -f segment -segment_list cam01.m3u8 -segment_list_flags + live -segment_time 2 -segment_list_size 20 -segment_wrap 20 cam01-% 03d.ts


  


  The question I ask myself, and the reason for this message is as follows :

  


  

  • would there be an interest (memory / cpu / speed) to develop a program (C/C++/other ?) to do the same thing by using the libraries of ffmpeg ?

  


  • or the "gain" and the interest would be so minimal that it is not worth the expenditure of energy and time ?

  


  


  I appeal to your feedback, your opinions, your tips !, and if you had leads (sample) to attack this kind of development, I am interested.

  


  Thank you very much in advance for your feedback.
(this is my fist question on stackoverflow, Champagne ! ;))

  


• subprocess.run output is empty (python 3.8)

  26 juillet 2022, par Marino Linaje

  I'm trying to capture the output of a command with the following code :

  


  lines = subprocess.run(['ffmpeg', '-hide_banner', '-nostats', '-i', in_filename, '-vn', '-af', 'silencedetect=n={}:d={}'.format(silence_threshold, silence_duration), '-f', 'null', '-'], capture_output=True, text=True, shell=True, check=True, encoding='utf-8').stdout 
print (lines)


  


  But lines is an empty string and nothing is printed.
When capture_output=True is removed, the correct output is showed (without printing it).

  


  I tested many combinations, including removing all the subprocess.run parameters and only include capture_output=True with the same result.

  


  I also tested with few argument for a minimal example : subprocess.run(['ffmpeg', '-version'], capture_output=True, text=True, shell=True, check=True, encoding='utf-8').stdout

  


  Also tested stdout=subprocess.PIPE and stderr=subprocess.PIPE as subprocess.run arguments instead of capture_output=True

  


  I can't figure out what is happening. Thanks in advance !

  


• sws_scale() does not convert image simply copying it from source to target

  14 mars 2020, par Slav

  Trying to read arbitrary video as plain RGB24 pixels so convert frame with sws_scale() this way :

      //...
  
      AVFrame* pic_out = av_frame_alloc();
  
      pic_out->format = AV_PIX_FMT_RGB24;
  
      pic_out->width  = 1920;
  
      pic_out->height = 1080;
  
      av_frame_get_buffer( pic_out, 32 );
  
  
  
      struct SwsContext * img_convert_ctx = sws_getContext(
  
          1920, 1080, AV_PIX_FMT_YUV420P,
  
          1920, 1080, AV_PIX_FMT_RGB24,
  
          SWS_BICUBIC,
  
          NULL, NULL, NULL
  
      );
  
      //...
  
      sws_scale(
  
          img_convert_ctx,
  
          pic_src->data,     //pic_src is from avcodec_receive_frame()
  
          pic_src->linesize,
  
          0,
  
          1080,
  
          pic_out->data,
  
          pic_out->linesize
  
      );

  Everything goes without any errors, but pic_out ends up having the same data as pic_src.
  What could be the problem ?

  Full minimal example is here (supposed to be RGB24 image is there as 2.bmp which looks like actually being YUV-something)