Recherche avancée

Sur d’autres sites (4810)

• Transcode HLS Segments individually using FFMPEG

  27 mai 2013, par rayh

  I am recording a continuous, live stream to a high-bitrate HLS stream. I then want to asynchronously transcode this to different formats/bitrates. I have this working, mostly, except audio artefacts are appearing between each segment (gaps and pops).

  Here is an example ffmpeg command line :

  ffmpeg -threads 1 -nostdin -loglevel verbose \
  
     -nostdin -y -i input.ts -c:a libfdk_aac \
  
     -ac 2 -b:a 64k -y -metadata -vn output.ts

  Inspecting an example sound file shows that there is a gap at the end of the audio :

  

  And the start of the file looks suspiciously attenuated (although this may not be an issue) :

  

  My suspicion is that these artefacts are happening because transcoding are occurring without the context of the stream as a whole.

  Any ideas on how to convince FFMPEG to produce audio that will fit back into a HLS stream ?

  ** UPDATE 1 **

  Here are the start/end of the original segment. As you can see, the start still appears the same, but the end is cleanly ended at 30s. I expect some degree of padding with lossy encoding, but I there is some way that HLS manages to do gapless playback (is this related to iTunes method with custom metadata ?)

  
  

  ** UPDATED 2 **

  So, I converted both the original (128k aac in MPEG2 TS) and the transcoded (64k aac in aac/adts container) to WAV and put the two side-by-side. This is the result :

  
  

  I'm not sure if this is representative of how a client will play it back, but it seems a bit odd that decoding the transcoded one introduces a gap at the start and makes the segment longer. Given they are both lossy encoding, I would have expected padding to be equally present in both (if at all).

  ** UPDATE 3 **

  According to http://en.wikipedia.org/wiki/Gapless_playback - Only a handful of encoders support gapless - for MP3, I've switched to lame in ffmpeg, and the problem, so far, appears to have gone.

  For AAC (see http://en.wikipedia.org/wiki/FAAC), I have tried libfaac (as opposed to libfdk_aac) and it also seems to produce gapless audio. However, the quality of the latter isn't that great and I'd rather use libfdk_aac is possible.

• Frame Accurate Seeking in WebM

  11 janvier 2016, par SapphireSun

  I’m trying to do a somewhat tricky thing with WebM. I am trying to encode a stack of 256 biological images as a WebM. The time dimension of motion is very similar to the space dimension of the image stack so the compression ratios are insanely good. However, I am having trouble decoding the movie frames. I know that WebM uses an IPB predictive coding scheme, but I was reading several blog posts and discussion groups from WHATWG from 2011, and they said that frame accurate seeking was working in Chrome at that time.

  When I do video.currentTime = 0, I correctly get this :

  

  However, if I do video.currentTime = 0.34 (for example) I get something that looks like this :

  

  It looks like I’m getting a random poorly predicted frame. Am I just encoding the video wrong ? When I play it normally it looks fine.

  I encoded the video using 256 pngs using ffmpeg compiled with libvpx using the VP8 codec.

  ffmpeg -y -framerate 60 -start_number 0 -pattern_type glob -i '*.png' -qmin 10 -qmax 42 out.webm

  References to the WHATWG and some other info from 2011 :

  WHATWG discusses frame accuracy :

  https://lists.w3.org/Archives/Public/public-whatwg-archive/2011Jan/0372.html

  BBC Tech Director talking about frame accuracy :

  http://www.bbc.co.uk/blogs/bbcinternet/2011/02/frame_accurate_video_in_html5.html

• Posthoc connect FFMPEG to opencv-python binary for Google Cloud Dataflow job

  16 juillet 2017, par bw4sz

  I would like to run some video processing jobs on linux based compute engines on Google Cloud DataFlow. Cloud DataFlow requires you to build a setup.py file, or supply dependencies in a requirements.txt.

  https://cloud.google.com/dataflow/pipelines/dependencies-python

  My video process requires opencv in python with FFMPEG support. I would like to avoid building opencv from source, as this takes nearly 35 minutes for each worker to git clone/make/make install.

  There is a linux python binary .whl that works great. But its specifically compiled without FFMPEG.

  From https://pypi.python.org/pypi/opencv-python

  "IMPORTANT NOTE

  MacOS and Linux wheels have currently some limitations :

  video related functionality is not supported (not compiled with FFmpeg)"

  Is it possible to post-hoc connect FFMPEG to the binary ? That is download FFMPEG and its libraries separately and still read video ? I know this is contrived, but are are there any options here besides building opencv from source for every new worker ?