Recherche avancée

Sur d’autres sites (4296)

• FFmpeg massive data loss when writing large data and swapping segments

  25 avril 2023, par Bohdan Petrenko

  I have an ffmpeg process running which continiously writes audio data to two 30 seconds (for testing, I'm actually planning to use 5 minutes) segments. The problem is that when I write some audio data with length more than size of two segments (60 seconds), 8-17 seconds of audio is lost. Here is how I run FFmpeg and write data :

  


      _ffmpeg = Process.Start(new ProcessStartInfo
    {
        FileName = "ffmpeg",
        Arguments = 
            $"-y -f s16le -ar 48000 -ac {Channels} -i pipe:0 -c:a libmp3lame -f segment -segment_time {BufferDuration} -segment_format mp3 -segment_wrap 2 -reset_timestamps 1 -segment_list \"{_segmentListPath}\" \"{segmentName}\"",
        UseShellExecute = false,
        RedirectStandardInput = true
    })!;
    // Channels is usually 1, BufferDuration is 30


  


  And here is how I write data :

  


  public async Task WriteSilenceAsync(int amount)
{
    if (amount > _size) amount = _size; // _size is 48000 * 1 * 2 * 30 * 2 = 5760000 (size of 1 minute of audio)
    
    var silence = _silenceBuffer.AsMemory(0, amount);
    await _ffmpeg.StandardInput.BaseStream.WriteAsync(silence);
}


  


  I tried to change the ffmpeg parameters and ways I write data. But I haven't found the solution.

  


  I'm sure that the problem is caused by ffmpeg segments, because if I disable segmenting and write audio to a single file, there are no problems with data loss or audio missmatch. I also sure that amount of silence to add in WriteSilenceAsync() method is calculated right. I'm not sure if the problem appears with data length more than 30 seconds but less then 1 minute, but I think it doesn't.

  


  I don't know how to solve this problem and would be glad to see any suggestions or solutions.

  


• Better examples/language for supported vs. ok(), onready()/ontimeout()

  30 décembre 2010, par Scott Schiller

  m doc/getstarted/index.html m doc/index.html Better examples/language for supported vs. ok(), onready()/ontimeout()

• How to convert natural language to structural description like FFmpeg filter graph strings

  11 juillet 2022, par 张无敌

  I'm working on a program that takes natural language as input, and outputs valid FFmpeg filter graph description, for example :

  


  Input string :

  


  rotate 1.mp4 90deg clockwise, downscale to 480p, output 2.mpeg


  


  Output description :

  


  rotate=a=90,scale=640*480


  


  It seems impractical with complete natural language, especially in the case of complex filter graphs, but still doable with slightly formatted input.

  


  I am new to NLP, wondering which areas or methods could help implement this.