Recherche avancée

Sur d’autres sites (12347)

• Improving Google Cloud Speech-to-Text accuracy

  6 juillet 2020, par lr_optim

  I'm working on a project where I need to perform these steps :

  


  

  1. Record a voice call (.webm -file)

  


  2. Split the webm -file into chunks with ffmpeg and convert the file into wav

  


  3. Transcribe the chunks using SpeechRecognition -library and Google Cloud API

  


  


  I've faced problems with the transcription accuracy and wondering if there is something I could do to improve it. At the time I'm splitting the original file into 30s chunks. I thought there might be one problem, that I might be missing words because of splitting so I've tried also with longer chunks under 60s but didn't notice any improve in accuracy.
Reading trough the speechRecognition docs I decided to set r.energy_threshold = 4000, I also tried to set the energy_treshold dynamically like this :

  


  with sr.AudioFile(name) as source:
    r.dynamic_energy_threshold = True
    r.adjust_for_ambient_noise(source, duration = 1)
    audio = r.record(source)


  


  I've also tested en-US and en-GB to see if there's some difference but there isn't as much as I'd want. The program is supposed to work with english language spoken by nordic people. If someone has experience about choosing a right language model for people speaking with accent, please let me know.

  


  This is the ffmpeg command is use to split the webm file into chunks : command = ['ffmpeg', '-i', filename, '-f', 'segment', '-segment_time', '30', parts_dir + outputname + '%09d.wav']

  


  Is there somethig I could do better ? I'm wondering if the quality is not good enough an Google is having hard time because of that ?

  


  The main problem is I'm getting bad results (lots of wrong words) from Google and wondering if there is something I could do about it.

  


• How to insert bullet screen comments into a video by using ffmpeg ?

  7 septembre 2019, par Saeron Meng

  I would like to add some real-time comments in my video but I do not know how to use ffmpeg to realize this. The comments are like screen bullets through the screen, scrolling from right to left.

  My thought is to count the length of the comments and define speeds for them to move. Don’t worry about the source of the comments, I have already gotten them saved as an xml file. Also, I can transfer it into srt. For instance :

  <chat timestamp="671.195">
  
      <ems utctime="1562584080" sender="Bill">
  
          <richtext></richtext>
  
      </ems>
  
  </chat>
  
  <chat timestamp="677.798">
  
      <ems utctime="1562584086" sender="Jack">
  
          <richtext></richtext>
  
      </ems>
  
  </chat>

  The final result is like this (I did not find an example displayed in English), these fancy characters can move horizontally :

  

  I have searched some solutions on the Internet, most of which talk about how to write ass/srt files and add motionless subtitles. Like this :

  3
  
  00:00:39,770 --> 00:00:41,880
  
  When I was lying there in the VA hospital ...
  
  
  
  4
  
  00:00:42,550 --> 00:00:44,690
  
  ... with a big hole blown through the middle of my life,
  
  
  
  5
  
  00:00:45,590 --> 00:00:48,120
  
  ... I started having these dreams of flying.

  ffmpeg -i infile.mp4 -i infile.srt -c copy -c:s mov_text outfile.mp4

  But I need another kind of "subtitles" which can move.

  So my question is, how to modify the ffmpeg command and the template of srt file so as to arrange the subtitles from the top to the bottom and let them move from right to left ?

• VLC/ffplay can't read some audiotracks created by ffmpeg

  19 juin 2021, par skanarr

  I have two different .mkv files of the same movie. One contains the English, German, Italian, Spanish and French audio and subtitle tracks, the other contains Japanese Audio and Subtitle tracks. Since I want to have all tracks in one file I tried to 'merge' them using :

  


  $ ffmpeg -i "Assassin's Creed (2016).mkv" -i "./Other/Assassin's Creed_t13.mkv" -map 0:v -map 0:a -map 1:a:2 -map 0:s -map 1:s:1 -map 1:s:2 -map 1:s:3 -c:v copy -c:s copy -c:a copy tmp2.mkv


  


  Full Command Output from ffmpeg

  


  However playing it in VLC none of the Tracks from ./Other/Assassin's Creed_t13.mkv are working properly.
There are short periods where there is audio and then it is gone again for some minutes.
Also tmp2.mkv freezes at 4:44 and freezes VLC for some time, before only audio continues top play.
Looking under Messages this is what I got : All Message from VLC
(Messages were set to display warnings and errors)

  


  I don't know what I did wrong. The Japanese tracks are working in ./Other/Assassin's Creed_t13.mkv.
The all non-japanese Tracks are still working in tmp2.mkv.
Also there are working Tracks with the same codecs on tmp2.mkv (dts) as well as subtitles (pgs).

  


  I then tried playing it with ffplay and got the same result.
Whenever i switched the audiotrack to stream 7 (the japanese audio)
It said non A-V: non in the little progress bar instead of the usual numbers.