Recherche avancée

Sur d’autres sites (9399)

• Green Stripe Artifacts When Encoding Video With FFMPEG & Python

  5 novembre 2014, par trianta2

  I’m attempting to encode (VP8 codec) and write a video using FFMPEG (version 2.3.3) and Python. However, I get these diagonal green stripe artifacts after I’ve finished encoding my video and I cannot find why.

  I have a sequence of frames in memory in the form of numpy ndarrays, which I generate synthetically for test purposes using a method :

  def _generate_test_images(self, samples=50):
  
      '''
  
      Creates an image array gradually changing from black to white
  
      '''
  
      pixelValues = np.linspace(0, 255, samples)
  
  
  
      imageList = [np.full((100, 100, 3), pixelValue, dtype=np.uint8)
  
                   for pixelValue in pixelValues]
  
  
  
      return np.array(imageList)

  I then use the Python subprocess module to open a pipe to FFMPEG and write the image data. I have tried using stdin.write and communicate but both produce the green stripe problem. Here is how I interact with FFMPEG :

      import subprocess as sp
  
      params = get_params() #provides info like threads, frame size, etc.
  
  
  
      command = list()
  
      command.extend(['/opt/ffmpeg/bin/ffmpeg'])
  
      command.extend(['-y'])
  
      command.extend(['-f', 'rawvideo'])
  
      command.extend(['-vcodec', 'rawvideo'])
  
      command.extend(['-s', params['frameSize']])
  
      command.extend(['-pix_fmt', 'bgr24'])
  
      command.extend(['-r', '1'])
  
      command.extend(['-an'])
  
      command.extend(['-i', '-'])
  
      command.extend(['-an'])
  
      command.extend(['-codec:v', 'libvpx'])
  
      command.extend(['-quality', 'good'])
  
      command.extend(['-cpu-used', '0'])
  
      command.extend(['-b:v', params['bitrate']])
  
      command.extend(['-qmin', '4'])
  
      command.extend(['-qmax', '42'])
  
      command.extend(['-maxrate', '1M'])
  
      command.extend(['-bufsize', '2M'])
  
      command.extend(['-threads', params['threads']])
  
      command.extend(['-f', 'webm'])
  
      command.extend([params['target']])
  
  
  
      pipe = sp.Popen(command, stdin=sp.PIPE, stdout=sp.PIPE, stderr=sp.PIPE, bufsize=-1)
  
      pipe.communicate(input=frameArray.tostring())

  However when my video is done encoding, this is what I see :

  

  What is causing this ?

• ffmpeg write while converting video

  6 septembre 2016, par Ahmed Alashrafy

  I am trying to watermarking the video but after size of output video become larger than 800MB the ffmpeg stopped working without error. So what I think is that the buffer size is full. So my question is how to make ffmpeg release the finished part each 10 minutes instead of waiting after the full video being converted and then write it to the disk.

  ffmpeg.exe -i input.mkv -i logo.png -report -filter_complex "overlay = 20:15,subtitles = 'sub.srt'" -codec:a copy output.mkv

• Converting problematic 9.99fps video to something standard (ffmpeg ?)

  22 octobre 2014, par Sai

  Question moved over to SuperUser.

  I recently got some CCTV video released to me via a FOIA/PA request. I’d like to put up the full version on YouTube, as well as version edited w/ Final Cut Pro. However, Handbrake, Quicktime, Final Cut Pro, & YouTube are all unable to open the original .avi file. :-(

  ffmpeg -i says for avi input #0 (there is no audio part) :

   Duration: 00:48:48.73, start: 0.000000, bitrate: 2497 kb/s
   Stream #0:0: Video: h264 (Main) (V264 / 0x34363256), yuv420p, 1280x960, 2495 kb/s, 9.99 fps, 9.99 tbr, 9.99 tbn, 19.98 tbc

  Handbrake’s log showed this before giving up with "no valid source found" (note the tbr discrepancy vs ffmpeg -i) :

   Duration: 00:48:48.72, start: 0.000000, bitrate: 2497 kb/s
   Stream #0.0: Video: V264 / 0x34363256, 1280x960, 10 fps, 29.97 tbr, 9.99 tbn
[13:54:47] scan: decoding previews for title 1
[13:55:28] scan: could not get a decoded picture
[13:55:38] scan: could not get a decoded picture
[13:55:43] scan: could not get a decoded picture
[13:55:49] scan: could not get a decoded picture
[13:55:54] scan: could not get a decoded picture
[13:55:58] scan: could not get a decoded picture
[13:56:03] scan: could not get a decoded picture
[13:56:06] scan: could not get a decoded picture
[13:56:07] Warning: Could not read data for preview 9, skipped
[13:56:07] Warning: Could not read data for preview 10, skipped
[13:56:07] libhb: scan thread found 0 valid title(s)
[13:56:07] macgui: ScanDone state received from fHandle

  VLC is able to play it just fine, but I’ve not managed to get VLC to transcode it into something usable ; an attempt at transcoding to m4v and importing that to FCP resulted in a weirdly stuttering and moonwalky video when FCP tried to convert it to a normal FPS.

  How can I convert this into something that’s of normal size / fps / codec so that I can edit it in FCP and upload it to YouTube specs ?

  My guess is that ffmpeg should be able to do the job, but I’m not sure how to go about it.