Recherche avancée

Sur d’autres sites (8934)

• How can I convince ffserver to save a locally-sourced webcam stream to a file in high resolution AND stream it in lower resolution ?

  2 novembre 2015, par Dominic Jacobssen

  We have a remote Linux machine, accessible over VPN, which has a USB webcam. We want to use this for video conferencing, but we also want to store the stream for archiving.

  Since the streaming bandwidth is limited, it makes sense to capture the stream on the same machine as the webcam and rsync that across after-the-fact, rather than trying to capture the streamed content, which is necessarily going to be poor quality.

  We’re trying to use ffmpeg and ffserver to achieve this, but with little success. Most of the articles on the internet either deal with just streaming a webcam, or rebroadcasting a remote stream. We found we had to recompile ffserver because of a missing "my_addr->sin_family = AF_INET ;" in the version of ffserver.c we had been using, since fixed in git.

  Here’s the ffserver.conf we’re trying to use :

  Port 43688
  
  BindAddress 127.0.0.1
  
  MaxHTTPConnections 2000
  
  MaxClients 1000
  
  MaxBandwidth 1000
  
  CustomLog -
  
  NoDaemon
  
  <feed>
  
  ReadOnlyFile /tmp/feed.ffm
  
  FileMaxSize 20M
  
  ACL allow 127.0.0.1
  
  </feed>
  
  <stream>
  
  Feed feed.ffm
  
  Format mp4
  
  VideoSize qvga
  
  VideoGopSize 12
  
  VideoHighQuality
  
  Video4MotionVector
  
  VideoCodec libx264
  
  VideoBitRate 100
  
  VideoBufferSize 40
  
  VideoFrameRate 5
  
  VideoQMin 3
  
  VideoQMax 31
  
  AudioCodec libfaac
  
  AudioBitRate 32
  
  AudioChannels 2
  
  AudioSampleRate 22050
  
  ACL allow localhost
  
  </stream>

  When we fire this up, we get the error :

  Unable to create feed file '/tmp/feed.ffm' as it is marked readonly

  Fair enough, but this is not what is implied in the docs. Changing the directive to :

  File /tmp/feed.ffm

  allows ffserver to fire up and appear to sit and wait for ffmpeg to connect to it. However, when we fire up ffmpeg with the command :

  ffmpeg -f alsa -i pulse -r 16000 -f video4linux2 -s qvga -i /dev/video0 -r 5 -f mp4 -vcodec libx264 -sameq -acodec libfaac -ab 32k http://127.0.0.1:43688/feed.ffm

  then the webcam lights up and ffserver acknowledges the connection with the messages :

  New connection: POST /feed.ffm
  
  [POST] "/feed.ffm HTTP/1.1" 200 0

  but after a few seconds we get the errors :

  [mp4 @ 0x264b160] muxer does not support non seekable output
  
  Could not write header for output file #0 (incorrect codec parameters ?)

  We’ve tried various other formats (mpeg, mpegts, avi) and codecs (mpeg1video, mpeg2video, mpeg4), all without success.

  Moreover, we were under the impression that ffserver could reencode input format to a lower resolution for streaming, but if the stream resolution doesn’t match the feed resolution, we get an error about the resolutions not matching.

  Has anyone ever managed to get this working correctly ? I’ve read about vlc being able to do something like this, but the vlc command lines are well nigh impenetrable.

  Thanks !

  Dominic

• Theatrical quality ffmpeg/x264 encoding of a high-motion 1080p video

  2 décembre 2011, par Ian

  I've been struggling with encoding videos using FFMPEG and x264. The output stutters when played back in Quicktime, while in VLC it shows a lot of compression artifacts at the same places Quicktime stutters. So it seems like Quicktime is stuttering because it's trying to suppress the corruption/artifacts.

  The videos have a lot of random motion in them, including frames where 75% of the pixels will change at a random interval (the video is software generated so it's truly pseudo-random). The compression seems to be choking in these places where it's likely detecting a "scene cut" incorrectly. It also seems to choke at regular intervals where I guess it's doing a keyframe.

  I've based my encoding preset off of the x264-hq preset that comes with FFMPEG. I've tried turning off scene cut detection, and playing with the keyint/g and keyint_min options. Setting g to 1 makes it work, but blows out the filesize. I've tried the lossless presets, but they won't playback at all in Quicktime. Oddly, I haven't had any problems when working with a lower-resolution test video (1440x810).

  Here's the preset I have right now, which works, but yields a file that's approximately 60% larger than the (non-working) hq preset yields. Is there any way to improve upon this ? The filesize doesn't matter much, I just want something that will playback anywhere and be very high quality.

  coder=1
  flags=+loop
  cmp=+chroma
  partitions=+parti8x8+parti4x4+partp8x8+partp4x4+partb8x8
  me_method=umh
  subq=8
  me_range=16
  g=1
  keyint_min=1
  sc_threshold=0
  i_qfactor=0.71
  b_strategy=1crf=20
  qcomp=0.6
  qmin=20
  qmax=51
  qdiff=4
  bf=16
  refs=4
  trellis=1
  flags2=+dct8x8+wpred+bpyramid+mixed_refs
  wpredp=2
  

  Here's the command :

  ffmpeg \
    -r 60 -i "frame-%06d.tiff" \
    -vcodec libx264 -vpre my_preset \
    -threads 0 \
    -r 60 -an -f out.mp4
  

• Theatrical quality ffmpeg/x264 encoding of a high-motion 1080p video

  2 décembre 2011, par Ian

  I've been struggling with encoding videos using FFMPEG and x264. The output stutters when played back in Quicktime, while in VLC it shows a lot of compression artifacts at the same places Quicktime stutters. So it seems like Quicktime is stuttering because it's trying to suppress the corruption/artifacts.

  The videos have a lot of random motion in them, including frames where 75% of the pixels will change at a random interval (the video is software generated so it's truly pseudo-random). The compression seems to be choking in these places where it's likely detecting a "scene cut" incorrectly. It also seems to choke at regular intervals where I guess it's doing a keyframe.

  I've based my encoding preset off of the x264-hq preset that comes with FFMPEG. I've tried turning off scene cut detection, and playing with the keyint/g and keyint_min options. Setting g to 1 makes it work, but blows out the filesize. I've tried the lossless presets, but they won't playback at all in Quicktime. Oddly, I haven't had any problems when working with a lower-resolution test video (1440x810).

  Here's the preset I have right now, which works, but yields a file that's approximately 60% larger than the (non-working) hq preset yields. Is there any way to improve upon this ? The filesize doesn't matter much, I just want something that will playback anywhere and be very high quality.

  coder=1
  flags=+loop
  cmp=+chroma
  partitions=+parti8x8+parti4x4+partp8x8+partp4x4+partb8x8
  me_method=umh
  subq=8
  me_range=16
  g=1
  keyint_min=1
  sc_threshold=0
  i_qfactor=0.71
  b_strategy=1crf=20
  qcomp=0.6
  qmin=20
  qmax=51
  qdiff=4
  bf=16
  refs=4
  trellis=1
  flags2=+dct8x8+wpred+bpyramid+mixed_refs
  wpredp=2
  

  Here's the command :

  ffmpeg \
    -r 60 -i "frame-%06d.tiff" \
    -vcodec libx264 -vpre my_preset \
    -threads 0 \
    -r 60 -an -f out.mp4