Recherche avancée

Sur d’autres sites (7258)

• lavu/hwcontext_qsv : silence the warning

  24 juillet 2023, par Haihao Xiang

  lavu/hwcontext_qsv : silence the warning
  libavutil/hwcontext_qsv.c : In function ‘qsv_map_to’ :
  
  libavutil/hwcontext_qsv.c:1905:47 : warning : cast from pointer to integer
  
  of different size [-Wpointer-to-int-cast]
  Signed-off-by : Haihao Xiang <haihao.xiang@intel.com>
  

  • [DH] libavutil/hwcontext_qsv.c

• FFmpeg what is the correct way to manually write silence through pipe:0 ?

  19 juillet 2023, par Bohdan Petrenko

  I have an ffmpeg process running with this parameters :

  


  ffmpeg -y -f s16le -ac {Channels} -ar 48000 -re -use_wallclock_as_timestamps true -i pipe:0 -f segment -segment_time {_segmentSize} -segment_list \"{_segmentListPath}\" -segment_format mp3 -segment_wrap 2 -reset_timestamps 0 -af aresample=async=1 \"{_filePath}\"


  


  I also have a DateTimeOffsetwhich represents the time when the recording was started. When an FFMpeg process is created, I need to add some some amount of silence that equals to the delay between current time and when the recording was started. This delay may be bigger than ffmpeg segments, so I calculate it relatively to the time when last ffmpeg segment should begin.
I store silence in a static byte array with length of two ffmpeg segments :

  


  _silenceBuffer ??= new byte[_segmentSize * 2 * Channels * SampleRate * 2];


  


  I tried two ways of writing silence :

  


  First code I tried is this :

  


  var delay = DateTimeOffset.UtcNow - RecordingStartDateTime;

var time = CalculateRelativeMilliseconds(delay.TotalMilliseconds); // this returns time based on current segment. It works fine.

var amount = (int)(time * 2 * Channels * SampleRate / 1000);

WriterStream.Write(_silenceBuffer, 0, amount);


  


  As the result, I have a very loud noise everywhere in output from ffmpeg. It brokes audio, so this way doesn't work for me.

  


  Second code I tried is this :

  


  var delay = DateTimeOffset.UtcNow - RecordingStartDateTime;

var time = CalculateRelativeMilliseconds(delay.TotalMilliseconds); // this returns time based on current segment. It works fine.

var amount = (int)time * 2 * Channels * SampleRate / 1000;

WriterStream.Write(_silenceBuffer, 0, amount);


  


  Difference between first and second code is that now I cast only time to int type, not the result of the whole expression. But it also doesn't work. This time at the beginning I have no silence I wrote, the recording begins with voice data I piped after writing silence. But if I use this ffmpeg command :

  


  ffmpeg -y -f s16le -ac {Channels} -ar 48000 -i pipe:0 -f segment -segment_time {_segmentSize} -segment_list \"{_segmentListPath}\" -segment_format mp3 -segment_wrap 2 -reset_timestamps 0 \"{_filePath}\"


  


  Then it works as expected. Recording begins with silence what I need, and then goes voice data I piped.

  


  So, how can I manually calculate and write silence to my ffmpeg instance ? Is there some universal way of writing and calculating silence that will work with any ffmpeg command ? I don`t want to use filters and other ffmpeg instances for offsetting piped voice data, because I do it only once per session. I think that I can write silence with byte arrays. I look forward to any suggestions.

  


• avcodec/dts2pts_bsf : Eliminate some 64bit corner cases

  21 novembre 2022, par Michael Niedermayer

  avcodec/dts2pts_bsf : Eliminate some 64bit corner cases
  Fixes : negation of -2147483648 cannot be represented in type 'int' ; cast to an unsigned type to negate this value to itself
  
  Fixes : 53364/clusterfuzz-testcase-minimized-ffmpeg_BSF_DTS2PTS_fuzzer-4693772269387776
  Found-by : continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
  
  Signed-off-by : Michael Niedermayer <michael@niedermayer.cc>
  

  • [DH] libavcodec/dts2pts_bsf.c