Recherche avancée

Sur d’autres sites (9137)

• movenc : Add support for writing sidx atoms for DASH segments

  5 octobre 2014, par Martin Storsjö

  movenc : Add support for writing sidx atoms for DASH segments
  A flag "dash" is added, which enables the necessary flags for
  
  creating DASH compatible fragments.
  When this is enabled, one sidx atom is written for each track
  
  before every moof atom.
  Signed-off-by : Martin Storsjö <martin@martin.st>
  

  • [DH] Changelog
  • [DH] libavformat/movenc.c
  • [DH] libavformat/movenc.h
  • [DH] libavformat/version.h

• How to remove silence from an audio file while leaving a bit of the otherwise deleted portion ?

  26 février 2020, par A Question Asker

  Pardon the title, it’s a bit difficult to explain concisely !

  I have a bunch of short (1s-4s) recordings of people reading sentences. They were recorded themselves. Often, there is a bit of silence at the beginning or end, and I would like to automate removing that.

  As a first pass, I used this command :

  sox original.mp3 edited.mp3 silence 1 0.1 2% reverse silence 1 0.1 2% reverse

  The problem is that when people speak, sometimes the very beginning of their speech is quiet and quickly ramps up. This command thus clipped the very beginning of a number of sentences.

  So what I would like to do is essentially the same as what the sox command does, but ideally once it detects the boundary, it leaves 50ms before the boundary, in recognition that there may be some quiet but important sounds.

  I think that ffmpeg might be a good tool for this, as I don’t think sox can be configured like that. I’m tool agnostic. I don’t know ffmpeg at all though, so I appreciate any help putting together a command given ffmpeg’s rather arcane syntax !

  Another nice to have over the aforementioned sox command is for it only to remove silence at the beginning and end (if it exists) — not from the middle of the recording.

  UPDATE : based on time spent with ffmpeg, it looks like I want something like this

  ffmpeg -i input.mp3 -af "silenceremove=start_periods=1:start_threshold=0.02:start_silence=0.1:detection=peak,areverse,silenceremove=start_periods=1:start_threshold=0.02:start_silence=0.1:detection=peak,areverse" output.mp3

  Oddly, this comes out garbled. I thought that it wasn’t reversing properly, but when I try to do something like

  ffmpeg -i input.mp3 -af areverse output.mp3

  It doesn’t work — I just get the input back. I’m not sure what is going wrong ?

  If I just do the beginning,

  ffmpeg -i input.mp3 -af "silenceremove=start_periods=1:start_threshold=0.02:start_silence=0.1:detection=peak" output.mp3

  It works fine — for the beginning. But not sure why the reverse trick is resulting in a garbled output.

  UPDATE2 : it actually looks like the command works, but it changes the metadata in a way that makes itunes wonky. If I use VLC instead, it works fine.

• aarch64 : vp9 : use alternative returns in the core loop filter function

  9 janvier 2017, par Janne Grunau

  aarch64 : vp9 : use alternative returns in the core loop filter function
  Since aarch64 has enough free general purpose registers use them to
  
  branch to the appropiate storage code. 1-2 cycles faster for the
  
  functions using loop_filter 8/16, ... on a cortex-a53. Mixed results
  
  (up to 2 cycles faster/slower) on a cortex-a57.
  This is cherrypicked from libav commit
  
  d7595de0b25e7064fd9e06dea5d0425536cef6dc.
  Signed-off-by : Michael Niedermayer <michael@niedermayer.cc>
  

  • [DH] libavcodec/aarch64/vp9lpf_neon.S