Recherche avancée

Sur d’autres sites (6970)

• ffmpeg concatenating videos of different fps while keeping the total length not changed

  23 novembre 2017, par A_Matar

  I wanna pad an mp4 video stream with another video clip of a static image that I created using :

  def generate_white_vid (duration):
  
      output_filename = os.path.join(p_path,'white_vid_'+" 0:.2f}".format(duration)+'.mp4')
  
      ffmpeg_create_vid_from_static_img = 'ffmpeg -loop 1 -i /path/WhiteBackground.jpg -c:v libx264 -t %f -pix_fmt yuv420p -vf scale=1920:1080 %s' % (duration, output_filename)
  
      p = subprocess.Popen(ffmpeg_create_vid_from_static_img, shell=True)
  
      p.communicate()
  
      return output_filename

  I use the following to concatenate :

  def concat_vids(clip_paths):
  
      filenames_txt = open('clips_to_join.txt','w')
  
      for clip in clip_paths:
  
          filenames_txt.write ('file \''+ clip+'\'\n')
  
      filenames_txt.close()
  
      output_filename = clip_paths[0].split('.', 2)[0]
  
      output_file_path = os.path.join(root_path, output_filename+'-padded.mp4')
  
      # join the clips
  
      ffmpeg_command = ["ffmpeg", "-f", "concat", "-safe", "0", "-i", "clips_to_join.txt", "-codec", "copy", output_file_path] # output_filename = ch0X-start_time-end_time
  
      p = subprocess.Popen(ffmpeg_command)
  
      p.communicate() # wait till the subprocess finishes. You can send commands to process as well.
  
      return output_file_path

  When I check the length of the resulting video after concatenation, I find that it is not equal to the sum of the two segments that I concatenated, and sometimes it is even less by some seconds !!

  Here is how I get the video length in seconds :

  def ffmpeg_len(vid_path):
  
      '''
  
      Returns length in seconds using ffmpeg
  
      '''
  
      ffmpeg_get_mediafile_length = ['sh', '-c', 'ffmpeg -i "$1" 2>&1 | grep Duration', '_', vid_path]
  
      p = subprocess.Popen(ffmpeg_get_mediafile_length, stdout=subprocess.PIPE, stderr=subprocess.PIPE)    
  
      output, err = p.communicate()
  
      length_regexp = 'Duration: (\d{2}):(\d{2}):(\d{2})(\.\d+),'
  
      re_length = re.compile(length_regexp)
  
      matches = re_length.search(output)
  
      if matches:
  
          video_length = int(matches.group(1)) * 3600 + \
  
                          int(matches.group(2)) * 60 + \
  
                          int(matches.group(3)) + float(matches.group(4))
  
          return video_length
  
      else:
  
          print("Can't determine video length.")
  
          print err
  
          raise SystemExit

  My guess is that maybe the concatenation unifies the fps rate for the all the clips to be joined, if this is the case, how to prevent this from happening ? How can I get a video of the desired length exactly.

  Maybe it worth mentioning that the video to padded is very short 0.42 second, the original video is 210.58 and the resultant video is 210.56 !

  I have verified that ffmpeg does generate the desired padding region and it is of the desired length 0.42 I got a 0.43 segment when I forced 30 fps but it is okay.

• How to save image from the middle of a video ?

  29 octobre 2017, par puppon -su

  I need to make a thumbnail for a video, to seek to the 25% of a video and save the image. Here is what I’m doing right now, but it only saves black image.

  #include 
  
  
  
  #include <libavformat></libavformat>avformat.h>
  
  #include <libavutil></libavutil>dict.h>
  
  
  
  int main (int argc, char **argv)
  
  {
  
  
  
      av_register_all();
  
  
  
      AVFormatContext *pFormatCtx = avformat_alloc_context();
  
  
  
      int res;
  
  
  
      res = avformat_open_input(&amp;pFormatCtx, "test.mp4", NULL, NULL);
  
      if (res) {
  
          return res;
  
      }
  
  
  
  
  
      avformat_find_stream_info(pFormatCtx, NULL);
  
  
  
      int64_t duration = pFormatCtx->duration;
  
  
  
  
  
      // Find the first video stream
  
      int videoStream=-1;
  
      for(int i=0; inb_streams; i++) {
  
          if(pFormatCtx->streams[i]->codec->codec_type==AVMEDIA_TYPE_VIDEO) {
  
              videoStream=i;
  
              break;
  
          }
  
      }
  
      if(videoStream==-1) {
  
          return -1;
  
      }
  
  
  
      AVCodecContext *pCodecCtxOrig = NULL;
  
  
  
      // Get a pointer to the codec context for the video stream
  
      pCodecCtxOrig=pFormatCtx->streams[videoStream]->codec;
  
  
  
  
  
      AVCodec *pCodec = NULL;
  
      // Find the decoder for the video stream
  
      pCodec=avcodec_find_decoder(pCodecCtxOrig->codec_id);
  
      if(pCodec==NULL) {
  
          fprintf(stderr, "Unsupported codec!\n");
  
          return -1; // Codec not found
  
      }
  
  
  
  
  
      AVCodecContext *pCodecCtx = NULL;
  
      // Copy context
  
      pCodecCtx = avcodec_alloc_context3(pCodec);
  
      if(avcodec_copy_context(pCodecCtx, pCodecCtxOrig) != 0) {
  
          fprintf(stderr, "Couldn't copy codec context");
  
          return -1; // Error copying codec context
  
      }
  
  
  
  
  
      // Open codec
  
      if(avcodec_open2(pCodecCtx, pCodec, NULL)&lt;0) {
  
          return -1; // Could not open codec
  
      }
  
  
  
  
  
      AVFrame *pFrame = NULL;
  
  
  
      pFrame=av_frame_alloc();
  
  
  
      AVFrame *pFrameRGB = NULL;
  
  
  
      pFrameRGB=av_frame_alloc();
  
  
  
  
  
  
  
      // Determine required buffer size and allocate buffer
  
      int numBytes=avpicture_get_size(AV_PIX_FMT_RGB24, pCodecCtx->width,
  
                                  pCodecCtx->height);
  
  
  
      uint8_t *buffer = NULL;
  
      buffer=(uint8_t *)av_malloc(numBytes*sizeof(uint8_t));
  
  
  
  
  
      // Assign appropriate parts of buffer to image planes in pFrameRGB
  
      // Note that pFrameRGB is an AVFrame, but AVFrame is a superset
  
      // of AVPicture
  
      res = avpicture_fill((AVPicture *)pFrameRGB, buffer, AV_PIX_FMT_RGB24,
  
                      pCodecCtx->width, pCodecCtx->height);
  
      if (res&lt;0) {
  
          return;
  
      }
  
  
  
  
  
  
  
      // I've set this number randomly
  
      res = av_seek_frame(pFormatCtx, videoStream, 20.0, AVSEEK_FLAG_FRAME);
  
      if (res&lt;0) {
  
          return;
  
      }
  
  
  
  
  
  
  
      AVPacket packet;
  
      while(1) {
  
          av_read_frame(pFormatCtx, &amp;packet);
  
          if(packet.stream_index==videoStream) {
  
              int frameFinished;
  
              avcodec_decode_video2(pCodecCtx, pFrame, &amp;frameFinished, &amp;packet);
  
              if(frameFinished) {
  
                  SaveFrame(pFrameRGB, pCodecCtx->width,
  
                      pCodecCtx->height);
  
                  break;
  
              }
  
  
  
          }
  
      }
  
  
  
  
  
      avformat_close_input(&amp;pFormatCtx);
  
      return 0;
  
  }
  
  
  
  
  
  
  
  void SaveFrame(AVFrame *pFrame, int width, int height) {
  
    FILE *pFile;
  
    char szFilename[] = "frame.ppm";
  
    int  y;
  
  
  
    // Open file
  
    pFile=fopen(szFilename, "wb");
  
    if(pFile==NULL)
  
      return;
  
  
  
    // Write header
  
    fprintf(pFile, "P6\n%d %d\n255\n", width, height);
  
  
  
    // Write pixel data
  
    for(y=0; ydata[0]+y*pFrame->linesize[0], 1, width*3, pFile);
  
  
  
    // Close file
  
    fclose(pFile);
  
  }

  I was following this tutorial http://dranger.com/ffmpeg/tutorial01.html http://dranger.com/ffmpeg/tutorial07.html . It says that it was updated in 2015, but there already are some warnings about deprecated code, for example here : pFormatCtx->streams[i]->codec.

  I got video duration (in microseconds), but I don’t understand what I should send to av_seek_frame. Can I somehow use frame number for both duration and seeking, instead of time ?

• How to set comma delimited values from stdout ?

  28 octobre 2017, par gregm

  I have a batch file that processes the output of an ffprobe query. It retrieves several bits of data that I use to determine some ffmpeg directives. In particular I’m converting h264 videos into h265 if the video frame height is 720 or greater. I also convert the audio stream to aac if it isn’t already and if that stream is higher than 128 kbps I convert it down to 128.

  I can do all of that by calling ffprobe a number of times and use if statements to decide what my ffmpeg command will be.

  I’d like my batch file to be more efficient so I was thinking that if I could take the output of one (maybe two) ffprobe queries and then stick that output into a for /f token=.... loop then I could set each ffprobe data point to a variable and then just check the variables to decide what the resulting ffmpeg command will be.

  Here’s what I have right now to simply check if the video stream is hevc. If it isn’t then ffmpeg converts the video to hevc and copies the audio to aac.

  for %%a in ("*.*") do (
  
  ffprobe -v quiet -show_entries stream=index,codec_name,height -of csv "%%a" 2>&1 | findstr "hevc"
  
  if errorlevel 1 (
  
      ffmpeg.exe -hwaccel cuvid -i "%%a" -pix_fmt p010le -c:v hevc_nvenc -preset slow -rc vbr_hq -b:v 4M -maxrate:v 10M -c:a aac "%%~na.h265-convert.mp4"
  
  ))

  That ffprobe query output looks like this :

  stream,0,h264,480

  I was thinking if I could tokenize that output with something like :

  for /f "tokens=1,2,3,4 delims= " %%a in ("______") do set codec=%%b&set fheight=%%d

  I don’t know what to put in the spot where I have the _______. I really don’t want to create a temp file unless that’s the only option though.

  1) Is this an efficient way to achieve what I’m trying to do ?

  2) What do I use where I have a blank line above ________ to call the output of the ffprobe query to use in my for loop ?