Recherche avancée

Sur d’autres sites (11141)

• avcodec/h2645 : allocate film grain metadata dynamically

  13 août 2024, par Dale Curtis

  avcodec/h2645 : allocate film grain metadata dynamically
  Film grain support adds a huge amount of overhead to the H264Context
  
  structure for a feature that is rarely used. On low end devices or
  
  pages that have lots of media this bloats memory usage rapidly.
  This changes the static film grain metadata allocations to be dynamic
  
  which reduces the H264Context size from 851808 bytes to 53444 bytes.
  Bug : https://crbug.com/359358875
  
  Signed-off-by : Dale Curtis <dalecurtis@chromium.org>
  
  Signed-off-by : Niklas Haas <git@haasn.dev>
  

  • [DH] libavcodec/h2645_sei.c
  • [DH] libavcodec/h2645_sei.h
  • [DH] libavcodec/h264_picture.c
  • [DH] libavcodec/h264_sei.c
  • [DH] libavcodec/h264_slice.c
  • [DH] libavcodec/h264dec.c
  • [DH] libavcodec/h264dec.h
  • [DH] libavcodec/hevc/hevcdec.c
  • [DH] libavcodec/hevc/hevcdec.h

• fate/cbs : update the two film grain cbs_av1 tests

  12 décembre 2019, par James Almer

  fate/cbs : update the two film grain cbs_av1 tests
  They were missed in the previous commit.
  Signed-off-by : James Almer <jamrial@gmail.com>
  

  • [DH] tests/ref/fate/cbs-av1-av1-1-b10-23-film_grain-50
  • [DH] tests/ref/fate/cbs-av1-av1-1-b8-23-film_grain-50

• How do you make a modern bleep censor ? (With ffmpeg or other ready-made software)

  1er août 2023, par nick carraway

  I have always been a fan of live caller radio shows. These (sports) shows allow callers to call a hotline and talk to the radio host directly.

  


  One of the oldest problems with this show format is what if the caller curses, or says something highly inappropriate ? To keep the show clean (and legal), the radio shows broadcast with a 7 second delay. They also use a "bleep" censor, which historically allowed them to wipe the incriminating phrase with a "Beeeeeep" sound. These days, however, they completely cut out the caller's sentence before it even begins. ("Ah, we had to let you go there pal. Can't say that on the radio"). In the modern method, the transition is seamless, almost like they shorten the 7 second delay to a 4 second delay as they remove the start of the caller's reply entirely, and overwrite it with the host's explanations. The caller does not appear to be "interrupted" at all, the start of their sentence leading to the bad phrase is never even broadcasted.

  


  I've been thinking about how to do this in software. I found a project that looked promising. It adds a 7 second delay to your streams, and allows you to convert X amount of those seconds into silence assuming a caller says something inappropriate. While not ideal (since it's a few seconds of dead silence and would interrupt the caller mid-sentence), how can you do something like this in ffmpeg ? It is a good starting point before implementing the more modern features.

  


  


  1. How do you use ffmpeg to livestream a video/audio stream with a delay ?

     


  


  2. How do you overwrite the last 3 seconds of that stream with silence or a "bleep", when you need to ?

     


  


  3. Are you able to easily switch your stream to overwrite those 3 seconds with new audio (e.g. the host's explanation for why the caller was hung-up on) ? And how can you go back from a 4 second delay to a 7 second delay ?

     


  


  4. OR, is there a ready-made way to get flawless "radio-like" hang-ups on bad callers ?