Sur d’autres sites (7105)

• Absolute timestamp as MP4 start time

  14 juin 2016, par galbarm

  I’d like to store the exact start time a video was recorded on, inside its mp4 container.
  I need a millisecond accuracy (i.e. year,month,day,hour,sec,milli).
  Such an accuracy requires 8 bytes.

  The only standard way I found to store a video creation time is to use the mvhd/tkhd/mdhd boxes creation_time field.
  But according to the base media file format spec, the field only gives a granularity of seconds :

  creation_time is an integer that declares the creation time of this
  track (in seconds since midnight, Jan. 1, 1904, in UTC time)

  In version 0 the field size was 4 bytes, while in version 1 it was increased to 8 bytes. But the description remained unchanged so it can still only reflect a timestamp in up to second granularity. (for maintaining backward compatibility maybe ?)

  So finally, is there a standard way to store a single absolute timestamp with millisecond accuracy in a mp4 container ?
  If the only way to do it, is to store it as a custom metadata, is there an agreed common way to do it according to ?

• Transcoding fMP4 to HLS while writing on iOS using FFmpeg

  29 avril 2017, par bclymer

  TL ;DR

  I want to convert fMP4 fragments to TS segments (for HLS) as the fragments are being written using FFmpeg on an iOS device.

  Why ?

  I’m trying to achieve live uploading on iOS while maintaining a seamless, HD copy locally.

  What I’ve tried

  1. Rolling AVAssetWriters where each writes for 8 seconds, then concating the MP4s together via FFmpeg.

     What went wrong - There are blips in the audio and video at times. I’ve identified 3 reasons for this.

     1) Priming frames for audio written by the AAC encoder creating gaps.

     2) Since video frames are 33.33ms long, and audio frames 0.022ms long, it’s possible for them to not line up at the end of a file.

     3) The lack of frame accurate encoding present on Mac OS, but not available for iOS Details Here

  2. FFmpeg muxing a large video only MP4 file with raw audio into TS segments. The work was based off the Kickflip SDK

     What Went Wrong - Every once in a while an audio only file would get uploaded, with no video whatsoever. Never able to reproduce it in-house, but it was pretty upsetting to our users when they didn’t record what they thought they did. There were also issues with accurate seeking on the final segments, almost like the TS segments were incorrectly time stamped.

  What I’m thinking now

  Apple was pushing fMP4 at WWDC this year (2016) and I hadn’t looked into it much at all before that. Since an fMP4 file can be read, and played while it’s being written, I thought that it would be possible for FFmpeg to transcode the file as it’s being written as well, as long as we hold off sending the bytes to FFmpeg until each fragment within the file is finished.

  However, I’m not familiar enough with the FFmpeg C API, I only used it briefly within attempt #2.

  What I need from you

  1. Is this a feasible solution ? Is anybody familiar enough with fMP4 to know if I can actually accomplish this ?
  2. How will I know that AVFoundation has finished writing a fragment within the file so that I can pipe it into FFmpeg ?
  3. How can I take data from a file on disk, chunk at a time, pass it into FFmpeg and have it spit out TS segments ?

• ffmpeg : is vidstab multithreaded, and/or is there a way to make it perform better on a very high # of cores ?

  30 juillet 2016, par ljwobker

  I’m working on a project where I use the vidstab ffmpeg plugin to stabilize some videos. I’m lucky enough to have access to an extremely fast x86 machine (2 socket, 16 core, 32 thread) but I can’t seem to keep it busy and I’m trying to figure out if it’s a limitation of the toolchain or the config/commands. The workflow is basically 3 steps :

  1. crop the video in terms of both time and dimension (ffmpeg "crop"
     filter)
  2. run vidstabdetect to identify the transformation corrections
  3. run vidstabtransform to apply the transformation and output the
     final video

  When I run the transcode script on this machine, the "crop" pass executes extremely fast, and the htop output from the machine clearly shows all 32 cores(threads) running at nearly 100%.

  When I run the pass with vidstabdetect, htop clearly shows one core running at/near 100%, with all of the other cores hovering in the "few percent" range, and total CPU utilization for the parent PID hovers near 130%. This leads me to believe there must be only one main processing thread, but also several other smaller threads that are consuming at least some parallel time.

  the vidstabtransform pass looks similar, with one core constantly near 100%, and the rest of the cores hovering in the few percent.

  As far as I can tell, there is no way to parallelize the two vidstab processes, as the transform step is completely dependent on the results of the detection pass. There is a single pass option described in the vidstab docs, but the quality isn’t as good so I’m trying to avoid that.