Recherche avancée

Sur d’autres sites (11382)

• probing individual klv streams for specific signature/header

  17 juin 2019, par J Heyman

  Currently, the software I support processes the different streams within a video container (.ts, .mp4, .mpg, etc) without any issues as long as there is only one(1) type of each codec stream.

  I’ve recently encountered a video sample that actually contains three(3) identified AV_CODE_ID_SMPTE_KLV streams. As I loop through the three streams, one of them is the stream I need.
  I haven’t been able to figure out an easy way to do the specific query I need (check for known header bytes in the stream).

      ...
  
      for (i = 0; i < nb_streams; i++) {
  
         int real_stream_index = program ? program[i] : i;
  
         AVStream *st          = ic->streams[real_stream_index];
  
         AVCodecParameters *par = st->codecpar;
  
         if (par->codec_type != type)
  
            continue;
  
         if (id != AV_CODEC_ID_NONE) {
  
            if (par->codec_id != id)
  
               continue;
  
         }
  
         if (wanted_stream_nb >= 0 && real_stream_index != wanted_stream_nb)
  
            continue;
  
         if (type == AVMEDIA_TYPE_AUDIO && !(par->channels && par->sample_rate))
  
            continue;
  
         disposition = !(st->disposition & (AV_DISPOSITION_HEARING_IMPAIRED | AV_DISPOSITION_VISUAL_IMPAIRED));
  
         count = st->codec_info_nb_frames;
  
         bitrate = par->bit_rate;
  
         multiframe = FFMIN(5, count);
  
         if ((best_disposition >  disposition) ||
  
             (best_disposition == disposition && best_multiframe >  multiframe) ||
  
             (best_disposition == disposition && best_multiframe == multiframe && best_bitrate >  bitrate) ||
  
             (best_disposition == disposition && best_multiframe == multiframe && best_bitrate == bitrate && best_count >= count))
  
            continue;
  
         best_disposition = disposition;
  
         best_count   = count;
  
         best_bitrate = bitrate;
  
         best_multiframe = multiframe;
  
         ret          = real_stream_index;

  My thought was to add another || to the complex if{} above, but I haven’t been able to figure out how to do the comparison I need (looking for the header bytes).

  I’ve looked into existing documentation, and thought that accessing the probe_data structure within the AVStream contained within the AVFormatContext structure would give me the first few bytes of the stream. No such luck, as the probe_data structure is empty even though we’ve done a probe on the file itself.

  fprintf(stderr, "Filename: %s\t buf_size: %d\n", st-> probe_data.filename, st-> probe_data.buf_size);

• FFmpeg individual image zoom-in transition

  19 juillet 2024, par The Somebody

  I'm stuck with an ffmpeg command that needs to generate a video. Originally, I had an ffmpeg command that would loop through images and change after a certain amount of time. Now, to improve the video a little bit, I wish to add a zoom in effect on each image.

  


  I am facing an issue that when I generate the video, the images no longer change. That is, it is constantly first image. I can the zoom effect replay as it should (for every changing image), but the images do not actually change (it is always the first picture). Any advice/suggestion would be appreciated. I am sure, it will be something wrong with the syntax, but now I am at a loss..

  


  Original code, that works fine (without the zoom in) :

  


  ffmpeg -n -loop 1 -t 26.01 -i "\image_0.png" -loop 1 -t 26.01 -i "\image_1.png"
 -loop 1 -t 26.01 -i "\image_2.png" -loop 1 -t 26.01 -i "\image_3.png" 
 -i "/speech.mp3" 
 -filter_complex "[0:v]scale=1080x1920,setpts=PTS-STARTPTS[v0]; 
 [1:v]scale=1080x1920,setpts=PTS-STARTPTS[v1]; 
 [2:v]scale=1080x1920,setpts=PTS-STARTPTS[v2]; 
 [3:v]scale=1080x1920,setpts=PTS-STARTPTS[v3]; 
 [v0][v1][v2][v3]concat=n=4:v=1:a=0,subtitles='"C\:/transcription.ass"'[v]" -map "[v]" -map 4:a -c:v libx264 -c:a aac -b:a 192k
 -shortest "C:\Videos/Video.mp4" -loglevel verbose


  


  Code that does not work as intended :

  


  ffmpeg -n -loop 1 -t 26.01 -i "\image_0.png" -loop 1 -t 26.01 -i "\image_1.png" -loop 1 -t 26.01 -i "\image_2.png" -loop 1 -t 26.01 -i "\image_3.png" -i "/speech.mp3" -filter_complex "[0:v]scale=1080x1920,zoompan=z='zoom+0.001':d=650:s=1080x1920:x=iw/2-(iw/zoom/2):y=ih/2-(ih/zoom/2),setpts=PTS-STARTPTS[v0]; [1:v]scale=1080x1920,zoompan=z='zoom+0.001':d=650:s=1080x1920:x=iw/2-(iw/zoom/2):y=ih/2-(ih/zoom/2),setpts=PTS-STARTPTS[v1]; [2:v]scale=1080x1920,zoompan=z='zoom+0.001':d=650:s=1080x1920:x=iw/2-(iw/zoom/2):y=ih/2-(ih/zoom/2),setpts=PTS-STARTPTS[v2]; [1:v]scale=1080x1920,zoompan='z=zoom+0.001':d=650:s=1080x1920:x=iw/2-(iw/zoom/2):y=ih/2-(ih/zoom/2),setpts=PTS-STARTPTS[v3]; [v0][v1][v2][v3]concat=n=4:v=1:a=0,subtitles='C\:/transcription.ass'[v]" -map "[v]" -map 4:a -c:v libx264 -c:a aac -b:a 192k -shortest "C:\Videos/Video.mp4" -loglevel verbose


  


  I tried to change and see if it really is just the first image that's used through out the video and confirmed it. Also tried to play around with the symbols '"', ' ' ', etc.

  


• ffmpeg - convert video into individual frame image files ? [closed]

  2 janvier 2021, par MrFunreal

  i am trying to turn every single frame of a video file into png files using ffmpeg.
i have found several guides on how to do that, most of which look the same, by doing this :

  


  


  ffmpeg -i vid.mp4 frames/out%03d.png

  


  


  However, this does not work. FFmpeg for some reason just replaces the "%0" of the output file with the entire path of the bat file and then tells me that's wrong, like so :

  


  


  [image2 @ 000002c6aa87f680] Could not open file :
frames/outG :\Videos\vid_to_frames\3_video_to_frames.bat3d.jpg
av_interleaved_write_frame() : I/O error

  


  


  If i remove the "%0" of the output file it creates a single frame and gives me an error, telling me i need to put the "%03d" into the name, which doesn't work.

  


  


  [image2 @ 000001dc9a09fd40] Could not get frame filename number 2 from
pattern '$filename3d.png'. Use '-frames:v 1' for a single image, or
'-update' option, or use a pattern such as %03d within the filename.
av_interleaved_write_frame() : Invalid argument

  


  


  Is this just broken ?
I found about 20 guides on how to do it, all of which use this exact same method and it just doesn't work.

  


  Does anyone know another way to do this via ffmpeg ? (it must be via ffmpeg)