Recherche avancée

Sur d’autres sites (7174)

• Adds ESPCN super resolution filter merged with SRCNN filter.

  13 juin 2018, par Sergey Lavrushkin

  Adds ESPCN super resolution filter merged with SRCNN filter.
  Signed-off-by : Pedro Arthur <bygrandao@gmail.com>
  

  • [DH] configure
  • [DH] libavfilter/Makefile
  • [DH] libavfilter/allfilters.c
  • [DH] libavfilter/dnn_backend_native.c
  • [DH] libavfilter/dnn_backend_tf.c
  • [DH] libavfilter/dnn_espcn.h
  • [DH] libavfilter/dnn_interface.h
  • [DH] libavfilter/dnn_srcnn.h
  • [DH] libavfilter/vf_sr.c
  • [DH] libavfilter/vf_srcnn.c

• MoviePy on EC2 Instance Super Slow

  27 janvier 2020, par connorvo

  No matter what size EC2 instance I use, MoviePy.write_videofile() takes forever.

  I ran it on a t2.micro, c5.large, and c4.2xlarge and they all projected to take the same amount of time ( 7 hours).

  I have a 720mb mp4 file that I am adding a 30mb mp4 file to the beginning of and then putting a text overlay on the entire video (just static text saying the company name in the bottom corner).

  final_video.write_videofile(
  
      f"{FINAL_VIDEO_FILENAME}.mp4", 
  
      fps=60,
  
      codec='libx264', 
  
      audio_codec='aac', 
  
      temp_audiofile=f'{FINAL_VIDEO_FILENAME}_temp_audio.mp4', 
  
      remove_temp=True,
  
      threads = 8, # this was for c4.2xlarge instance
  
  )

  def __get_intro_clip(self, title, size, fps):
  
          clip = ( VideoFileClip('ytbot/intro_video/intro_video.mp4')
  
              .subclip(0,9.5)
  
              .resize(size)
  
              .set_fps(fps)
  
          )
  
  
  
          # make bigger then resize() so stroke works properly
  
          text_clip1 = ( TextClip(f'{TEXTTEXT}',font='Helvetica-Narrow-Bold',fontsize=self.DEFAULT_TEXT_SIZE*self.TITLE_MULTIPLIER,color='white',stroke_color='black',stroke_width=self.DEFAULT_STROKE_SIZE*self.TITLE_MULTIPLIER).resize(self.DEFAULT_RESIZE_MULTIPLIER)
  
              .margin(bottom=175, opacity=0)
  
              .set_position('center')
  
              .set_duration(4) 
  
              .set_start(1)
  
              .fadein(0.25)
  
              .fadeout(0.25)
  
          )
  
          text_clip2 = ( TextClip(f'presents',font='Helvetica-Narrow-Bold',fontsize=self.DEFAULT_TEXT_SIZE*self.TITLE_MULTIPLIER*0.7,color='white',stroke_color='black',stroke_width=self.DEFAULT_STROKE_SIZE*self.TITLE_MULTIPLIER).resize(self.DEFAULT_RESIZE_MULTIPLIER)
  
              .set_position('center')
  
              .set_duration(3.5) 
  
              .set_start(1.5)
  
              .fadein(0.25)
  
              .fadeout(0.25)
  
          )
  
          text_clip3 = ( TextClip(title,font='Helvetica-Narrow-Bold',fontsize=self.DEFAULT_TEXT_SIZE*self.TITLE_MULTIPLIER,color='white',stroke_color='black',stroke_width=self.DEFAULT_STROKE_SIZE*self.TITLE_MULTIPLIER).resize(self.DEFAULT_RESIZE_MULTIPLIER)
  
              .margin(top=175, opacity=0)
  
              .set_position('center')
  
              .set_duration(3) 
  
              .set_start(2)
  
              .fadein(0.25)
  
              .fadeout(0.25)
  
          )
  
  
  
          return CompositeVideoClip([clip, text_clip1, text_clip2, text_clip3]).fadein(0.5)
  
  
  
      def edit_yt_video(self, video_path, name):
  
          video_clips = []
  
  
  
          clip = VideoFileClip(video_path)
  
  
  
          video_clips.append(self.__get_intro_clip(name, clip.size, clip.fps))
  
  
  
          text_clip = ( TextClip('Company: {name}',font='Helvetica-Narrow-Bold',fontsize=self.DEFAULT_TEXT_SIZE,color='white',stroke_color='black',stroke_width=self.DEFAULT_STROKE_SIZE).resize(self.DEFAULT_RESIZE_MULTIPLIER)
  
              .margin(bottom=15, left=15, opacity=0)
  
              .set_position(('left', 'bottom'))
  
              .set_duration(clip.duration)
  
          )
  
  
  
          video_clips.append(CompositeVideoClip([clip, text_clip]).fadeout(0.5).crossfadein(0.5))
  
  
  
          return concatenate_videoclips(video_clips, padding=-1, method='compose')

• Using ffmpeg and aresample with super low sample rate

  23 mars 2020, par Jurijs Kovzels

  I am trying to use ffmpeg to get data to generate list of peaks with resolution of 30 peaks per second.

  the params I’m using are :

  ffmpeg -y -i audio.wav -filter_complex "[0]aresample=30[resampled]" -map [resampled] output_30.wav

  which kills almost all information and resulting data array contains only very small values. This waveform is before resampling
  

  versus after resampling to 60kHz

  

  My question is if it is possible to use ffmpeg and get maximums (peaks) over each timespan (second or 1/20 second) of audio ?