Recherche avancée

Sur d’autres sites (10217)

• Create MPEG-DASH Initialization segment

  5 janvier 2016, par Mahout

  I am looking to convert between HLS and MPEG Dash. I do not access to the original fully concatenated video file, only the individual HLS segments.

  In doing this transformation to MPEG Dash I need to supply an initialziation segment for the Dash manifest .mpd file.

  My questions are :

  1. What is the structure of a Dash video initialization segment ?
  2. How can I generate/create one without the need for the original full file ?

  Perhaps a solution would involve getting MP4Box to convert the ’.ts’ HLS segments to Dash ’.m4s’ segments which are self initializing, but I am unsure how to go about this this ?

  Any ideas are much appreciated.

  Many thanks.

  UPDATE :
  Snippet to stream using original hls segments. Video plays all the way through but is just black.

    <representation width="426" height="238" framerate="25" bandwidth="400000">
  
      <segmentlist timescale="25000" duration="112500">
  
             <segmenturl media="video_0_400000/hls/segment_0.ts"></segmenturl>
  
             <segmenturl media="video_0_400000/hls/segment_1.ts"></segmenturl>
  
            <segmenturl media="video_0_400000/hls/segment_2.ts"></segmenturl>
  
      </segmentlist>
  
     </representation>

• How can I best utilize an AWS service to segment a video into smaller chunks and then combine them back to together ? [on hold]

  19 avril 2018, par Justin Malin

  I am trying to do processing on videos uploaded to AWS S3 using an AWS Lambda function in Python. However, FFmpeg and ffmpeg-python (as far as I am aware) are unable to process objects and must do processing on stored files. Lambda only allows for 500 MB of storage in the /tmp/ folder, thus limiting the size of video that I can do processing on.

  If there is an alternative to FFmpeg that allows me to work on object files that I am unaware of, that would be a reasonable solution because I can scale up the memory of the Lambda function (although there is still a limit).

  Alternatively, I have looked into segmenting the video using AWS Elastic Transcoder, but I do not think I can dynamically segment the video using that service. If there is a service similar to this that could segment the video into individual frames (and back), that would be even better.

  I have also considered using AWS EC2, but I would only be using the EC2 service to segment videos sporadically, so it would be a waste to constantly have a server that capable running. If I use the AWS Elastic Beanstalk, would it automatically start a more powerful instance of EC2 to do the video segmentation (and reformation) when that is called and revert back to a much smaller instance when dormant ?

  Essentially, I would like to know if there are any services (preferably within AWS) that allow me to segment a video into shorter videos or into each frame at-will.

• bash : receive single frames from ffmpeg pipe

  30 août 2014, par manu

  I’m trying to achieve single-frame handling in a pipe where the the j2c encoder "kdu_compress" (Kakadu) only accepts single files. To save harddrive space. I didn’t manage to pipe frames directly, so I’m trying to handle them via a bash script, by creating each picture, process it, and overwrite it with the next.

  Here is my approach. Thanks for your advice, I really want to climb this mountain, though I’m a bit fresh here thanks.

  ---

  Is it possible to pipe an ffmpeg output to a bash script and save the individual frame,
  do further commands with the file before the next frame is handled ?

  Best result so far is, that ALL frames are added into the intermediate file, without recognizing the end of a frame.

  I used this ffmpeg setting to pipe, example with .ppm :

  ffmpeg -y  -i "/path/to/source.mov" -an -c:v ppm -updatefirst 1 -f image2 - \
  
   | /path/to/receiver.sh

  and this script as a receiver.sh

  #!/bin/bash  
  
  
  
  while read a; 
  
  do 
  
      cat /dev/null > "/path/to/tempfile.ppm"; #to empty the file first
  
      cat $a >> "/path/to/tempfile.ppm";        #to fill one picture
  
  
  
      kdu_compress -i /path/to/tempfile.ppm -otherparams   #to process this intermediate
  
  
  
  done
  
  exit;

  Thank you very much.