Recherche avancée

Sur d’autres sites (3746)

• Any equivalent to '-c:v copy' for changing container of webm to mp4 video client-side in JS ?

  15 mars 2020, par programmingisphun

  Exploring the Mediarecorder API and really hope to get an mp4 output instead of webm. Prefer to stay client-side to reduce server resources and ffmpeg.js at 17mb doesn’t seem viable for online use.

  Discovered that one can create (Chrome/Firefox) a webm/h.264, which converts without re-encoding to an mp4 using FFMPEG. The resulting file opens fine in Quicktime MacOS (snippet source) :
  ffmpeg -i _inputfile_.webm -c:v copy _outputfile_.mp4

  Whereas this clever hack works to display the video maybe as an mp4 in the browser, but the downloaded file won’t open in Quicktime (just VLC, same for webm) :

  new Blob(recordedBlobs, {type: 'video/mp4'});
  
  video.src = window.URL.createObjectURL(superBuffer);

  Therefore, wondering if there’s another JS hack to do what FFMPEG’s -c:v copy does, but to the blob bytes of the MediaRecorder’s recordedBlobs and give it the same mp4 container that’s accepted by Quicktime/etc ?

  Pastebin of current setup

• First HLS segment overlapping when renditions have different framerates

  7 septembre 2023, par drajvver

  I have a multi resolution m3u8 file, created from 1080p60 source file. Resolutions lower than 720p are converted to 30fps, while 720p+ stay at 60fps.

  


  The problem is that the first segment is broken between 30 and 60fps renditions, causing an overlap, which is detected by hls.js as a gap between segments for some reason.

  


  I have checked the segment files manually and it looks like they are fine, except for single audio frame missing between them, 60fps one has 187 'frames', while 30fps one has 186 'frames'.

  


  Videos are segmented in separate ffmpeg runs, but using the same settings, differing only in FPS, resolution and bitrate. Source file is always the same.

  


  Encoding parameters are : -async 30 -fps_mode cfr -vf "scale=640:-2,framerate=fps=30" -ac 2 -preset faster -crf 23 -b:a 128k -maxrate 1160k -bufsize 1600k -force_key_frames "expr:if(isnan(prev_forced_n),1,eq(n,prev_forced_n+30))" -g 30 -keyint_min 30 -sc_threshold 0 -hls_time 4 -hls_list_size 0 -hls_playlist_type vod -hls_flags independent_segments+discont_start -max_muxing_queue_size 9999 -profile:v main -pix_fmt yuv420p -fflags +genpts -r 30 - it's the same for 60fps, just with different resolution, bitrate and 30 gets replaced by 60

  


  Is there a way to fix this ?

  


  Thanks !

  


  


  


• Remove Static Pixels from a Video to Mimic a Green Screen Effect [closed]

  2 août 2021, par nKrkan

  I have a video that's 50 seconds in length, resolution of 480x480 and 16 frames per second.

  


  There is a person talking in it, with the background being static I thought if there's a way
  
to remove those static pixels (background) and just extract the moving pixels (foreground)
  
and possibly mimic a green screen effect ?

  


  I was thinking on writing a picture-by-picture comparison tool to do such thing but I don't
  
believe I'm up to the task, or maybe It's laziness.

  


  And now I know, some of you will point out that the video has compression artifacts and that
  
might cause some problems but It doesn't have to be Studio quality stuff.

  


  I tried the ffmpeg command from this question : Remove random background from video using ffmpeg or Python
  
And it does mask the person, but... I couldn't quite get it to work, apparently putting the
  
reference image in the input makes that image burned into the video, thus having no way to
  
remove it, but it did mask the background as black and the person as greenish, so still not a
  
viable way to do it.

  


  Have also tried some Python projects I've found on the GitHub but none of them worked as
  
I expected.

  


  So, what I thought to do is simply compare the first and the second frame of the video, check
  
all the pixels by comparing them with the two sources, and change those that stay within a
  
certain range of the initial pixel value.

  


  I should point out I'm not very knowledgeable with mathematics and the majority of the
  
methods used in these type of things, but perhaps someone could point me to an interesting
  
source to read and learn, or by providing an alternative to the methods aforementioned above.