Recherche avancée

Sur d’autres sites (7033)

• Building A livestreaming server like youtube from scratch

  9 décembre 2022, par Dipo Ahmed

  I am trying to build a live streaming server like youtube where I can watch the video live or if I want to I can play the video from any duration I want.

  


  What I have tried so far.
I have built a node js WebSocket server where I push the video blob that I receive from the browser via MediaRecorder API every 2 seconds. This blob is then getting converted to hls by a ffmpeg process which generates 2 seconds *.ts files and a .m3u8 file which I am playing with video.js in browser.

  


  This is my ffmpeg command

  


   spawn('ffmpeg', [
        '-i', '-',
        // '-re',
        '-fflags', '+igndts',

        '-vcodec', 'h264',
        '-acodec', 'aac',

        '-preset', 'slow',
        '-crf', '22',
        // You can also use QP value to adjust output stream quality, e.g.: 
        // '-qp', '0',
        // You can also specify output target bitrate directly, e.g.:
        '-b:v', '1500K',
        '-b:a', '128K', // Audio bitrate

        '-f', 'hls',
        '-hls_time', '1',
        // '-hls_playlist_type', 'vod',
        '-hls_list_size', '2',
        '-hls_flags', 'independent_segments',
        '-hls_segment_type', 'mpegts',
        '-hls_segment_filename', `${path}/stream%02d.ts`, `${path}/stream.m3u8`,
    ]);


  


  The problem is that the video js player duration is not updating like in youtube where the video duration increases every second.

  


  Any direction will be appreciated. Please tell me if my approach is wrong and what needs to be learned for me to build this system.

  


• images->video->web canvas : RGB/YUV issues

  5 février 2016, par nrob

  We’ve written an web app which :

  1. takes 3D, time dependent weather data
  2. tiles each 3D time point to make a 2D frame (written out as a png image)
  3. stitches these frames together into a video (using ffmpeg/avconv)
  4. streams this video into a web app
  5. polls the canvas for frames
  6. sends the frames to the GPU where they are converted back to 3D and ray traced

  You can see the app here, code here and you can see the data video here

  Currently the pngs are written as RGB images, the video codec is in YUV and getting frames from the canvas returns RGB. As such there is a significant loss of information due to the conversion between image spaces.

  Does anyone have suggestions what is the best way round this ?

  I’ve tried a bunch of RGB video codecs, but I can’t get any to work, and I don’t know if the web browser will support it anyway. Can anyone suggest a good RGB codec (both lossy and lossless would be great)

  Also, is it possible to write to YUV images/read them from a video canvas in HTML5 ?

  Ultimately, I don’t even want anything to do with images/videos, I’m just hacking the codecs to stream/compress large animated 3D data volumes

• FFMPEG reports invalid video resolution, video headers container seem to be different from actual frames !

  20 décembre 2011, par Emiliano

  Quick question, i have a movie, which was cut and rendered with Sony Vegas from its original format to a .wmv file. Here comes the tricky part, movie when played, either with VLC or WMP, has a resolution of 656x480 ... BUT when i run a ffmpeg -i on it, it says it has a resolution of 600x480 ....

  I took the time of actually capturing a frame and croping it with photoshop and its 656 and not 600 like ffmpeg its reporting, why would this could be happening ? How could i fix the headers resolution ? Would that have any impact on video re-rendering ? As i said, VLC and WMP seems not to care about the incorrect headers and are playing it right, BUT, jwplayer seems to be using the header information, which i don't blame him, its correct to do that, but why the video headers could be wrong ?

  ffmpeg -i trailer.wmv

  Input #0, asf, from 'trailer.wmv' :
  Duration : 00:01:04.93, start : 3.000000, bitrate : 2144 kb/s
  Stream #0.0 : Audio : wmav2, 44100 Hz, mono, 32 kb/s
  Stream #0.1 : Video : wmv3, yuv420p, 600x480 [PAR 59:54 DAR 295:216], 2065 kb/
  s, 25.00 tb(r)

  And yeah, the PAR/DAR parameters are also wrong, but honestly, i don't understand that technical shit, usually watch video and make sure it look good, any feedback would be appreciated :P

  Is there a way to change the container information with ffmpeg so applications that actually do use the container information don't render video incorrectly ?