Recherche avancée

Sur d’autres sites (14136)

• Failed to decode HLS by FFMpeg command. Invalid NAL unit 0

  9 mars 2024, par Fyodor Khruschov

  On front-end I create stream with chrome.tabCapture.capture or navigator.mediaDevices.getDisplayMedia methods. Then send chunks generated by MediaRecorder to server. On the server I have FFMpeg command which decodes chunks into .mp4 file. This is the command :

  


  ffmpeg -y -i - -preset veryfast -tune zerolatency -filter_complex [0:v]split=3[v1][v2][v3];[v1]scale=w=-2:h=1080,fps=30[v1out];[v2]scale=w=-2:h=720,fps=30[v2out];[v3]scale=w=-2:h=480,fps=30[v3out] -map [v1out] -maxrate:0 6M -bufsize:0 12M -keyint_min 100 -g 100 -map [v2out] -maxrate:1 3M -bufsize:1 6M -keyint_min 100 -g 100 -map [v3out] -maxrate:2 1.5M -bufsize:2 3M -keyint_min 100 -g 100 -c:v libx264 -map a:0 -c:a:0 aac -b:a:0 128k -ac 2 -map a:0 -c:a:1 aac -b:a:1 96k -map a:0 -c:a:2 aac -b:a:2 96k -f hls -hls_time 2 -hls_playlist_type vod -hls_flags independent_segments+temp_file -hls_segment_type fmp4 -hls_segment_filename ./output/ready/output_%v_%03d.m4s -var_stream_map v:0,a:0 v:1,a:1 v:2,a:2 -master_pl_name master.m3u8 ./output/ready/stream_%v.m3u8 -map 0:v:0 -map 0:a:0 -c:v copy -c:a aac ./output/download/video.mp4 -map 0:a:0 -ar 16000 -ac 1 -c:a pcm_s16le ./output/captions/audio.wav -loglevel info


  


  During the process of decoding I have these errors in logs :

  


  [extract_extradata @ 0x60000264b250] Invalid NAL unit 0, skipping.
[h264 @ 0x13ff04e60] Invalid NAL unit 0, skipping.
[h264 @ 0x13ff04e60] co located POCs unavailable
[h264 @ 0x13ff04e60] negative number of zero coeffs at 17 0
[h264 @ 0x13ff04e60] error while decoding MB 17 0
[h264 @ 0x13ff04e60] concealing 3388 DC, 3388 AC, 3388 MV errors in B frame
[h264 @ 0x13ff04e60] missing picture in access unit with size 24158
[h264 @ 0x13ff04e60] Invalid NAL unit 0, skipping.
[h264 @ 0x13ff04e60] data partitioning is not implemented. Update your FFmpeg version to the newest one from Git. If the problem still occurs, it means that your file has a feature which has not been implemented.
[h264 @ 0x13ff04e60] If you want to help, upload a sample of this file to https://streams.videolan.org/upload/ and contact the ffmpeg-devel mailing list. (ffmpeg-devel@ffmpeg.org)
[h264 @ 0x13ff04e60] no frame!
[h264 @ 0x13ff04e60] Unknown SAR index: 18.
[h264 @ 0x13ff04e60] Invalid NAL unit 0, skipping.
[h264 @ 0x13ff04e60] Unknown SAR index: 18.
[h264 @ 0x13ff04e60] number of reference frames (2+4) exceeds max (5; probably corrupt input), discarding one
[h264 @ 0x13ff04e60] number of reference frames (3+3) exceeds max (5; probably corrupt input), discarding one
[h264 @ 0x13ff04e60] number of reference frames (4+2) exceeds max (5; probably corrupt input), discarding one
[h264 @ 0x13ff04e60] FMO is not implemented. Update your FFmpeg version to the newest one from Git. If the problem still occurs, it means that your file has a feature which has not been implemented.
[h264 @ 0x13ff04e60] sps_id 4 out of range


  


  This issue is very inconsistent and happen in rare cases (I can't understand the logic). Most of the time chunks decoded successfully, but sometimes not.

  


  How to understand where the issue is coming from ? Is it possible for FFMpeg to skip wrong data and generate mp4 file anyway even with glitches, but don't crush ?

  


• How to save audio chunks from client to ffmpeg readable file ?

  22 septembre 2023, par LuckOverflow

  I am live recording audio data from a TS React front-end and need to send it to the server, where it can be saved to a file so that ffmpeg can mix it. The front-end saves the mic data to a blob with type "mimeType : "audio/webm ; codecs=opus" when printed in the browser terminal. I send the exact object that I printed to the server, where logging it indicates it is a, or was passed as a, "Buffer" object.

  


  I have tried saving that Buffer as a webm file, but when I pass that file as an input to ffmpeg ffprobe, I get the error "Format matroska,webm detected only with a low score of 1..." and "EBML header parsing failed.." "Invalid data found when processing input." I have tried several other formats to no success.

  


  I need a way to transform this Buffer object to an audio file that can be mixed by ffmpeg. When I am finished, I also need to be able to do the reverse operation to send it in the same format to another client for playback, which is currently working.

  


  Code that records and sends the audio (TS React) :

  


  React Record

  


  const startRecording = async function () {
    inputStream = await navigator.mediaDevices.getUserMedia({ audio: true });
   
    mediaRecorder.current = new MediaRecorder(inputStream, { mimeType: "audio/webm; codecs=opus" });

    mediaRecorder.current.ondataavailable = e => {
      console.log(e.data)
      if (e.data.size > 0) {
        socket.emit("recording", e.data);
        console.log("Audio data recorded. Transmitting to server via socketio...");
      }
    };

    mediaRecorder.current.start(1000);
  };



  


  Code that receives and tries to save the Buffer to a file (JS Node.js) :

  


  Server Receive

  


  socket.on("recording", (chunk) => {
    console_log("Audio chunk recieved. Transmitting to frontend...");
    socket.broadcast.emit('listening', chunk);

    fs.writeFileSync('out.webm', chunk.toString());
    if (counter > 3) {
      console.log("Trying ffmpeg...");

      ffmpegInstance
        .input('out.webm')
        .complexFilter([
          {
            filter: 'amix'
          }])
        .save('./Music/FFMPEGSTREAM.mp3');
    }

    counter++;
  });


  


  fluent-ffmpeg interface package is includued in the server code, but I have been using ffmpeg in the terminal (Pop OS) to debug. The goal is to save the file to a ram disk and use fluent ffmpeg to mix before sending to a different client for playback. Currently I am just trying to save it to disk and get ffmpeg command line to work on it.

  


  Update :
Problem was that the chunk I was analyzing didn't have the header info. MediaRecorder encodes, then slices it up, not slices it up into your specified time slot and encodes. I have not found a good solution to this. Saving the file, without toString I believe, results in a playable webm when the header is properly included.

  


• Permissions issue with Python and ffmpeg on a Mac

  13 avril 2020, par EventHorizon

  I am fairly new to Python ( 4 weeks), and I have been struggling with this all day.

  



  I am using MacOS 10.13, Python 3.7 via Anaconda Navigator 1.9.12 and Spyder 4.0.1.

  



  Somehow (only a noob, remember) I had 2 Anaconda environments. I don't do production code, just research, so I figured I would make life simple and just use the base environment. I deleted the other environment.

  



  I had previously got FFmpeg working and was able to do frame grabs, build mpeg animations, and convert them to gifs for blog posts and such. I had FFmpeg installed in the directories associated with the deleted environment, so it went away.

  



  No worries, I got the git URL, used Terminal to install it in /opt/anaconda3/bin. It's all there and I can run FFmpeg from the Terminal.

  



  My problem : When I attempt to run a module that previously worked fine, I get the following message :

  



  [Errno 13] Permission denied : '/opt/anaconda3/bin/ffmpeg'

  



  In my module I set the default location of FFmpeg : plt.rcParams['animation.ffmpeg_path'] = '/opt/anaconda3/bin/ffmpeg'

  



  In my module I have the following lines :

  



  writer = animation.FFMpegWriter(fps=frameRate, metadata=metadata)
writer.setup(fig, "animation.mp4", 100)


  



  This calls matplotlib's 'animation.py', which runs the following :

  



  def setup(self, fig, outfile, dpi=None):
    '''
    Perform setup for writing the movie file.

    Parameters
    ----------
    fig : `~matplotlib.figure.Figure`
        The figure object that contains the information for frames
    outfile : str
        The filename of the resulting movie file
    dpi : int, optional
        The DPI (or resolution) for the file.  This controls the size
        in pixels of the resulting movie file. Default is fig.dpi.
    '''
    self.outfile = outfile
    self.fig = fig
    if dpi is None:
        dpi = self.fig.dpi
    self.dpi = dpi
    self._w, self._h = self._adjust_frame_size()

    # Run here so that grab_frame() can write the data to a pipe. This
    # eliminates the need for temp files.
    self._run()

def _run(self):
    # Uses subprocess to call the program for assembling frames into a
    # movie file.  *args* returns the sequence of command line arguments
    # from a few configuration options.
    command = self._args()
    _log.info('MovieWriter.run: running command: %s', command)
    PIPE = subprocess.PIPE
    self._proc = subprocess.Popen(
        command, stdin=PIPE, stdout=PIPE, stderr=PIPE,
        creationflags=subprocess_creation_flags)


  



  Everything works fine up to the last line (i.e. 'command' looks like a well-formatted FFmpeg command line, PIPE returns -1) but subprocess.Popen() bombs out with the error message above.

  



  I have tried changing file permissions - taking a sledgehammer approach and setting everything in /opt/anaconda3/bin/ffmpeg to 777, read, write, and execute. But that doesn't seem to make any difference. I really am clueless when it comes to Apple's OS, file permissions, etc. Any suggestions ?