Recherche avancée

Sur d’autres sites (7623)

• Video Overlay not returning to black once video stopped

  24 septembre 2020, par Matt Nelson

  I am combining multiple videos from a webrtc call and aligning the "user-terminal" videos to the left, and the "user-visitor" videos to the right. As the visitor feed can start and strop, they have multiple videos so I'm offsetting them by the timestamp.

  


  This is working, however crazy it looks !

  


  The one last issue I have is that when the first visitor video stop on the right, it shows the last frame of that video until the next video starts on the left. Can I have it return o the black background ?

  


  Here is the command passed to ffmpeg :

  


  ffmpeg -y 
-i /recordings/process/5f6c9c3/videoroom-5f6c9c3-user-terminal-1600953586531366-audio.mjr.opus 
-i /recordings/process/5f6c9c3/videoroom-5f6c9c3-user-visitor-1600953592694430-audio.mjr.opus 
-i /recordings/process/5f6c9c3/videoroom-5f6c9c3-user-visitor-1600953609873223-audio.mjr.opus 
-i /recordings/process/5f6c9c3/videoroom-5f6c9c3-user-visitor-1600953628668227-audio.mjr.opus 
-i /recordings/process/5f6c9c3/videoroom-5f6c9c3-user-visitor-1600953663905342-audio.mjr.opus 
-i /recordings/process/5f6c9c3/videoroom-5f6c9c3-user-MasterTerminal-52350116-1600953681107272-audio.mjr.opus 
-i /recordings/process/5f6c9c3/videoroom-5f6c9c3-user-visitor-1600953697832165-audio.mjr.opus 
-i /recordings/process/5f6c9c3/videoroom-5f6c9c3-user-MasterTerminal-52350116-1600953723320364-audio.mjr.opus 
-i /recordings/process/5f6c9c3/videoroom-5f6c9c3-user-visitor-1600953725307043-audio.mjr.opus 
-filter_complex [1:a]adelay=6163|6163[1adelay];[2:a]adelay=23341|23341[2adelay];[3:a]adelay=42136|42136[3adelay];[4:a]adelay=77373|77373[4adelay];[5:a]adelay=94575|94575[5adelay];[6:a]adelay=111300|111300[6adelay];[7:a]adelay=136788|136788[7adelay];[8:a]adelay=138775|138775[8adelay];[0:a][1adelay][2adelay][3adelay][4adelay][5adelay][6adelay][7adelay][8adelay]amix=inputs=9:duration=longest[a] 
-map [a] -ac -2 /recordings/process/5f6c9c3/5f6c9c3.mp3


  


• FFMPEG : Recording a video stream to disk real-time

  31 mars 2022, par Евгений Федоров

  I made a video recording with ffmpeg

  


   var inputArgs = string.Format(CultureInfo.InvariantCulture, "-framerate {0} -f rawvideo -pix_fmt {3} -video_size {1}x{2} -i -", VideoFPS, VideoXRes, VideoYRes, VideoPXLFormat);

 var outputArgs = string.Format(CultureInfo.InvariantCulture, "-vcodec mpeg4 -crf {2} -pix_fmt yuv420p -preset {3} -shortest -r {1} \"{0}\"", outputPath, VideoFPS, VideoCRFValue, CompressionRate);

 var ffmpegProcess = new Process
        {
            StartInfo =
            {
                FileName = ffmpegPath,
                Arguments = $"{inputArgs} {outputArgs}",
                UseShellExecute = false,
                CreateNoWindow = true,
                RedirectStandardInput = true
            }
        };

        ffmpegProcess .Start();
        var VideoRecordingFFmpegInput = VideoRecordingFFmpegProcces.StandardInput.BaseStream;


  


  Now I write data bytes from the image into the stream from time to time

  


  VideoRecordingFFmpegInput.Write(framesByteArray, 0, sizeOfpack);


  


  When I finished recording the frames :

  


    VideoRecordingFFmpegInput.Flush();
  VideoRecordingFFmpegInput.Close();
  ffmpegProcess.WaitForExit();


  


  Everything works fine, and it creates the file I want.

  


  When the process starts, it creates a file that has a very small size


  


  And during the recording of image bytes in the video stream - this size does not change (Probably it is written to RAM ?)

  


  But as soon as the ffmpeg process finishes - flush / close
the file size becomes normal.

  


  

  


  The problem is that if I record for several hours it can cause an OutOfMemory exception

  


  Is there any way to have the frames record and immediately increase the file size ? (Without infinitely filling RAM ?)

  


• Why UA sniffing is generally bad, real-world example #1048577 : "That time a missing /s on navigator.userAgent (based on single-digit UA version assumptions) broke SM2 on iOS 8+ devices."

  1er septembre 2014, par scottschiller

  Why UA sniffing is generally bad, real-world example #1048577 : "That time a missing /s on navigator.userAgent (based on single-digit UA version assumptions) broke SM2 on iOS 8+ devices."
  The upside : The fix is adding \s to a UA regex check targeting legacy iOS devices with broken HTML5 Audio support, so that the "OS 10 " portion of the UA string does not match the pattern "OS 1 ".
  This is a pre-emptive fix for the pending iOS 8 release, which reportedly includes the string "iPhone OS 10_10 like Mac OS X". Due to SM2 checking for /os ([12|3_0|3_1])/i which had broken HTML5 Audio() support, and not considering the far future possibility of "iOS 10" being included in the userAgent string at the time, HTML5 audio is incorrectly disabled on iOS 8 beta 3 and is likely to apply to the final iOS 8 release.
  This is a good example and reminder of why UA checks are generally bad, and dangerous ; similar issues hit other libraries doing UA and plugin checks both client and server-side when IE hit 10, and Flash hit version 10, so this iOS pattern change may affect some other libraries targeting iOS as well.