Recherche avancée

Sur d’autres sites (12247)

• A query on ffmpeg's DASH muxer and setting timeShiftBufferDepth values in MPEG-DASH Media Presentation Description (MPD) files

  18 février 2019, par 林正浩

  The documentation for the webm_dash_manifest muxer mentions the time_shift_buffer_depth option, defined as "Smallest time (in seconds) shifting buffer for which any Representation is guaranteed to be available. "

  What’s the equivalent option/setting for the same in the dash muxer (,if implemented) ?

  Documentation for the webm_dash_manifest muxer

• MPEG DASH : Playing video segments from .m4s files instead of byte ranges in the MPD File using Simple DASH Player

  1er novembre 2018, par Trycoder

  I have followed the tutorial from the link below to create an MPEG DASH player using HTML5 and javascript.
  Building a simple MPEG dash player using HTML5 and JS.

  In the given tutorial, it is possible to play the video segments using byte ranges. But for my application, I need the following features.

  1. The video player should read the .m4s segment files and play the video instead of byte ranges.
  2. The amount of data in the MediaSource buffer should be calculated periodically, also the amount of space left in the buffer.

  Are these possible in the given player or Is there a better MPEG DASH Player with the above features ?

  PS : The MPD File is created using FFmpeg.

  ffmpeg  -f  avfoundation -video_size 1280x720 -framerate 30 -i 0 -vcodec libx264  -acodec aac  -b:v 800k  -f dash  -use_template 0  -min_seg_duration 4000 -single_file 1  -start_at_zero -live 1  ffmpeg.mpd

  Also, I tried using the DASH.js player but the documentation is very vast in it. Can we get the size of the source buffer in dash.js (The space available in the source buffer and also the amount of space filled) ? This is the main feature which is required for my project.

  Edit :
  Code which I tried

• Optimizing preprocessing of data

  12 mai 2018, par WhatAMesh

  I have to preprocess 1-2,5 Terabytes of 1920x1080 videos (in .mp4 format) and try to find the fastest way to do so.

  For every video I have to :

  • Sample at 1 fps
  • Crop two images out of every frame
  • Save the two frames (output format needs to be uncompressed)

  Im currently doing it the following way using ffmpeg :

  def cropVideo(vidInPath, vidOutPath):
  
              crop = 'ffmpeg -i ' + vidInPath + ' -filter_complex "[0:v]crop=400:200:0:484[face];[0:v]crop=300:50:1624:300[number]" -map [face] -map 0:a ' + vidOutPath + '/face/face.mp4 -map [number] -map 0:a ' + vidOutPath + '/number/number.mp4'
  
              subprocess.call(crop)
  
              return
  
  
  
  def videoToImageSequence(vidInPath, vidOutPath):
  
      conv = 'ffmpeg -i ' + vidInPath + ' -vf fps=1 ' + vidOutPath + 'video-frame%05d.png'
  
      subprocess.call(conv)
  
      return
  
  
  
  cropVideo(dirPath + a + '.mp4', dirPath + a + '/')
  
  videoToImageSequence(dirPath + a + '/face/face.mp4', dirPath + a + '/face/')
  
  videoToImageSequence(dirPath + a + '/number/number.mp4', dirPath + a + '/number/')

  Is there a faster way to do this ?