Recherche avancée

Sur d’autres sites (12395)

• Pipe between two child processes in Node.js ?

  4 mai 2014, par Joey Morani

  I’m trying to capture video using FFmpeg with Node.js, and send it to a browser via websockets for playing using the MediaSource API. What I have so far works in Firefox but doesn’t decode properly in Chrome. Apparently, from reading this question I need to use the sample_muxer program to ensure each ’cluster’ starts with a keyframe.

  Here’s the code I’m using :

  var ffmpeg = child_process.spawn("ffmpeg",[
  
      "-y",
  
      "-r", "30",
  
  
  
      "-f","dshow",           
  
      "-i","video=FFsource:audio=Stereo Mix (Realtek High Definition Audio)",
  
  
  
      "-vcodec", "libvpx",
  
      "-acodec", "libvorbis",
  
  
  
      "-threads", "0",
  
  
  
      "-b:v", "3300k",
  
      "-keyint_min", "150",
  
      "-g", "150",
  
  
  
      "-f", "webm",
  
  
  
      "-" // Output to STDOUT
  
  ]);
  
  
  
  ffmpeg.stdout.on('data', function(data) {
  
      //socket.send(data); // Just sending the FFmpeg clusters works with Firefox's 
  
                           // implementation of the MediaSource API. No joy with Chrome.
  
  
  
      // - - - This is the part that doesn't work - - -
  
      var muxer = child_process.spawn("sample_muxer",[
  
          "-i", data, // This isn't correct...
  
  
  
          "-o", "-" // Output to STDOUT
  
      ]);
  
  
  
      muxer.stdout.on('data', function(muxdata) {
  
          socket.send(muxdata); // Send the cluster
  
      });
  
  });
  
  
  
  ffmpeg.stderr.on('data', function (data) {
  
      console.log("" + data); // Output to console
  
  });

  Obviously I’m not piping it correctly and I’m unsure how I would while also including the arguments. Appreciate any help getting this working. Thanks !

• swscale : add two spatially stable dithering methods

  23 mars 2014, par Øyvind Kolås

  swscale : add two spatially stable dithering methods
  Both of these dithering methods are from http://pippin.gimp.org/a_dither/ for
  
  GIF they can be considered better than bayer (provides more gray-levels), and
  
  spatial stability - often more than twice as good compression and less visual
  
  flicker than error diffusion methods (the methods also avoids error-shadow
  
  artifacts of diffusion dithers).
  These methods are similar to blue/green noise type dither masks ; but are
  
  simple enough to generate their mask on the fly. They are still research work
  
  in progress ; though more expensive to generate masks (which can be used in a
  
  LUT) like ’void and cluster’ and similar methods will yield superior results
  

  • [DH] doc/scaler.texi
  • [DH] libswscale/options.c
  • [DH] libswscale/output.c
  • [DH] libswscale/swscale_internal.h
  • [DH] libswscale/utils.c

• Node.js Webm live stream server : issues with tag

  10 décembre 2013, par breathe0

  I'm using Node.js as stream server to stream realtime Webm videos that is sent by FFMPEG (executed from another application, the stream is done via HTTP) and received by a webapp that uses the tag.

  This is what I'm doing : FFMPEG streams the received frames using the following command :

  ffmpeg -r 30 -f rawvideo  -pix_fmt bgra -s 640x480 
  
  -i \\.\pipe\STREAM_PIPE -r 60 
  
  -f segment  -s 240x160 -codec:v libvpx  -f webm 
  
  http://my.domain.com/video_stream.webm

  (the stream comes from an application that uses the Kinect as source and communicates with FFMPEG through a pipe, sending one frame after another)

  When the webapp connects, it receives immediately this response from the server :

  HTTP/1.1 200 OK
  
  X-Powered-By: Express
  
  content-type: video/webm
  
  cache-control: private
  
  connection: close
  
  Date: Fri, 06 Dec 2013 14:36:31 GMT

  and a Webm header (previously stored on the server, with the same resolution and frame rate of the source stream and tested as working on VLC) is immediately appended. Then the webapp starts to receive the data streamed by FFMPEG. Here is a screenshot of Mkvinfo GUI showing the fields of the header :

  

  However, even if the Network tab of the Chrome console shows that there is an actual stream of data (meaning that what is streamed is not completely garbage, otherwise the connection would be dropped), the player doesn't display anything. We tried to manually prepend our header to the dumped video received by the webapp and VLC plays it just fine, but this is not happening with the tag.

  What can cause this problem ? Are we missing something about the encoding on the FFMPEG side or we stored wrong values on the header (or they're not enough) ?

  PS : I cannot rely on an extern stream server.

  PPS : We tried the following experiments :

  • substituting the video header with the one stored in the server makes the video playable on both vlc and video tag
  • if we dump the video that is already started (without an header) and we prepend the video header stored in the server or even its original header, the video is playable in VLC but not on the tag (we're carefully prepending the header just before the beggining of the cluster).