Recherche avancée

Sur d’autres sites (12487)

• MPEG-4 and alpha transparency - the ongoing saga

  6 décembre 2022, par OrangeDog

  I'd like to reach some sort of definitive answer for the following questions :

  



  

  1. Is alpha transparency supported in MPEG-4-based codecs ?

  


  2. Is there any way to be reasonably certain that there is no alpha-channel from the output of ffprobe ?

  


  



  Some links I've found on the subject :

  



  

  • https://stackoverflow.com/a/5057631/476716

  


  



  


  Also, transparent alpha layers, an optional part of the H.264 specification, are now supported in H.264-based QuickTime movies.

  


  



  

  • http://software.intel.com/en-us/articles/alpha-channel-support-in-the-h264-decoder-in-ipp/

  


  



  


  The support for alpha channel in the H.264 decoder is now available in Intel® Integrated Performance Primitives 6.0 version.

  


  



  

  • http://forums.creativecow.net/readpost/335/30475

  


  



  


  The MPEG4 Simple Studio Profile supports auxiliary channels, which could be used for alpha or depth — and presumably even displacement, motion vectors, normals, etc.

  


  



  However, I've never seen a concrete example, or indeed any details on how you use these announced features.

  


• Concatenate mkv files keeping original timestmap with gaps

  13 août 2020, par Filip

  16 days ago I had this problem : Concatenating mka files but keeping timestamp which I fixed by using amix, a delay by using start_pts from ffprobe.

  


  Today I have a new challenge like this, but with video.

  


  So I have a bunch of mkv videos. Each time a person joins a chat, a mkv is recorded, but if the person refreshes the page a new mkv is created with start_pts and start_time to what it actually is. Also if the meeting started and a person joins after a minute, the start_time is set to 1 minute. I need to merge all those mkv and pad them with blank screen when there is no feed.

  


  Like in the above example, if a person joins after a minute, the first minute is a blank screen. Also if the participant leaves and re-joins after 10 seconds, those 10 seconds are blank again.

  


  Any ideas on how to do that with ffmpeg ?

  


  Concrete example of files :

  


  0PA84c5c3f412769b311d44b159941b2d22.mkv - start_pts: 742 start_time: 0.742000
2PA73d94e8cb0f41c3002fadd6c04b4a88f.mkv - start_pts: 30761 start_time: 30.761000
3PAcd35e470325618fa8a3fb8bb5a41403e.mkv - start_pts: 50940 start_time: 50.940000
4PAddccde7b8847ecc43d5e8643b7903dba.mkv - start_pts: 69243 start_time: 69.243000


  


  The end file would result in a file with length 69.243000, first 0.742 seconds are blank and also the gaps between should also be blank.

  


  So far i've tried :

  


  ffmpeg -i 0PA84c5c3f412769b311d44b159941b2d22.mkv -i 2PA73d94e8cb0f41c3002fadd6c04b4a88f.mkv -i 3PAcd35e470325618fa8a3fb8bb5a41403e.mkv -i 4PAddccde7b8847ecc43d5e8643b7903dba.mkv -filter_complex "[0:v] [1:v] [2:v] [3:v] concat=n=4:v=1 [v]" -map "[v]" test.mkv

  


  This works but without those gaps i mentioned.

  


• FFMpeg- Raw compressed data to video

  21 août 2012, par p.streef

  I'm trying to use FFMpeg to create a video. So far i've been playing with a multiplexing example :
  http://ffmpeg.org/doxygen/trunk/muxing_8c-source.html, and i'm able to create a compressed video from an already existing video.

  Because my program is going to run on an embedded platform I would like to use some custom code (generated by a colleague) to compress the video data and place it into the video file.
  So I'm looking for a way to create a video file in c/c++ using ffmpeg in which i have full control over the compression part (to basically circumvent ffmpeg from doing the compression for me and inserting my own code).

  To clarify i'm planning to use this to save film from an intelligent camera into a compressed h264 mpeg-4 file.