Recherche avancée

Sur d’autres sites (7133)

• Make ffplay leave video on GPU

  6 février 2019, par tmlen

  Is there a way to use ffplay to decode a video on GPU (using the h264_cuvid decoder, for example), and then show the output onscreen, without copying each frame to the host first ?

  I’m using the following command to launch ffplay decoding+playing a sequence of h264 files on the GPU through CUVID :

  ffplay -v 64 -loop 0 -vcodec h264_cuvid video.mkv

  It plays the video properly, but if I run the NVidia Profiler (nvpp), the timeline looks like this :

  

  It seems that it does the decoding on the GPU, but then copies the output to host memory (Memcpy DtoH) for each frame.

  Is it possible to have ffplay skip this step, and instead directly show the image on GPU using its graphics context ?

• How to create UI for Terminal [y/n] Prompts ?

  22 juin 2023, par itsRits

  I am working with FFmpeg and I want to create a responsive user interface for terminal prompts.

  


  Here's the main function.

  


  #converter.py

def convert_video(input_video, output_video=None, new_fps=None, new_container=None):

   if not output_video:
      if new_container is not None:
         output_video = "output." + new_container
      else:
         output_video = "output." + input_video.split('.')[-1]
   
   command = f"ffmpeg -i {input_video}"

   if new_fps:
      command += f' -r {new_fps}'

   command += f' {output_video}'

   subprocess.call(command, shell=True)


  


  On running the above code, if the video at the output destination with the same name already exists.
the terminal prompts :File 'output.mp4' already exists. Overwrite? [y/N].
I want to implement same using the message box in pyside2. I can create a message box in such a fashion

  


  #ui.py
    message_box = QMessageBox()
    message_box.setText("Do you want to proceed?")
    message_box.setStandardButtons(QMessageBox.Yes | QMessageBox.No)
    message_box.setDefaultButton(QMessageBox.No)


  


  But I am unable to figure how to make the message box pop once this prompt is found in terminal. Further how to send response back to it ?

  


• Process Multiple Streams using ffmpeg-python

  27 novembre 2019, par C Dorman

  I have a MPEG-2 TS with a stream of video and a steam of KLV metadata. I can use ffmpeg-python to process each of these streams independently. For example, I can get the video stream, process each frames using numpy as shown in the docs. I can also get the data stream and show it in following way (using the library klvdata) :

  process = (
  
      ffmpeg
  
          .input(in_filename)
  
          .output('pipe:', format='data', codec='copy', map='data-re')
  
          .run_async(pipe_stdout=True, pipe_stderr=True)
  
  )
  
  
  
  for packet in klvdata.StreamParser(process.stdout.read()):
  
      packet.structure()
  
  
  
  process.wait()

  How do I do these at the same time ? I need to split the TS data into its streams and process them both, keeping them in sync. ffmpeg by itself can demultiplex the streams into separate files, but how do I handle the streams in python. The KLV has information that I want to show on top of the video stream (recognition boxes).