Recherche avancée

Sur d’autres sites (8774)

• ffmpeg for merging vfr videos

  20 avril 2021, par Rupsha Chaudhuri

  I'm trying to generate a timelapse with images taken roughly at 10 Hz and for about an hour. Maintaining the relative timestamps is important. So I've split the images into roughly 1 min chunks and I generated 60 small videos in parallel (I've handled the duration for the boundary images). After that I stitch the videos. However when I look at the final contiguous video I realize that at the chunk boundaries after stitching the video either has an extra frame or falls behind the images.. the timestamps don't line up.

  


  The smaller videos are variable frame rate, so I convert them to mpeg-ts before applying the concat protocol to merge them.

  


  


  


  Generate small video:
ffmpeg -f concat -safe 0 -i input.txt -vsync vfr -pix_fmt yuv420p output.mp4

Convert to mpeg-ts
ffmpeg -i output.mp4 -c copy -bsf h264_mp4toannexb output.ts

Merge multiple small mpeg-ts videos
ffmpeg -i input.txt -c copy final_output.mp4

  


  


  



• What are good settings for transcoding videos uploaded to my app ?

  14 mai 2020, par Dmitry Minkovsky

  I am working on an app that allows users to share videos. The problem is that many videos are very high bitrate. For example, A 4-minute H264 video from an old iPhone is encoded at 1080p and runs 17,000 kb/s ( 500 megabytes). Accepting and distributing such videos at this bitrate/resolution is not practical for a social application.

  



  I have been playing with ffmpeg to transcode videos to smaller sizes and higher compression, but have not achieved acceptable results. For example :

  



  ffmpeg \
    -i in.mov \
    -vf scale=w='if(gt(iw\,ih)\,780\,-2)':h='if(gt(iw\,ih)\,-2\,780)' \ 
    -c:v libx264 \
    -crf 28 \
    -preset medium \
    -pix_fmt yuv420p \
    -movflags +faststart \
    out.mp4


  



  This command transcodes the above-mentioned 500MB file down to 70MB. It scales the larger dimension of the video to 780 pixels and compresses the video quite a bit. The results are okay, but the file is still large.

  



  Taking the longer dimension down to 480 pixels, the file is reduced to 40MB. Still quite large, and now significantly degraded. Also, the transcoding still takes quite a long time : about 1-1.5x on my 4 year old i7 Macbook Pro with 16GB RAM.

  



  I'm not sure how to improve on this. H265 is not supported in browsers. I am wondering :

  



  

  • How can I reduce size further ?

  


  • How can I transcode faster than 1x without significantly reducing quality ? Even 2-3x doesn't seem great ?

  


  



  Is this as good as it gets ?

  


• Livestream of prerecorded flv videos with ffmpeg and red5

  6 octobre 2014, par user2060812

  My goal is to acheive the following steps :-

  1. rtmp livestream of prerecoreded flv videos using ffmpeg.
     
  2. videos should be played continuously just like a tv station.
     

  We are currently using red5 and ffmpeg to acheive this goal and we have successfully published the live stream of prerecorded single video to jwplayer using the following conversion command :

  for i in *.avi; do ffmpeg -i $i -acodec copy -vcodec copy -f flv rtmp://localhost/oflaDemo/livestream

  But the problem comes when we need to livestream two videos one after the other. User have to click play button again in order to stream second video which is not a tv-station thing instead we need to continuously play stream for user instead of clicking on play button on ending of each stream.