Recherche avancée

Sur d’autres sites (4421)

• ffmpeg and Red5 Issue : Increase in number of ffmpeg simultaneous streams to Red5 resulting in packet loss

  30 octobre 2014, par kajarigd

  I have a screen sharing app written in flex, using which one person can share his screen with another person via Red5 server (Version : 1.0.3). Platform is Windows Server 2008. Now, I want to load test this Red 5 server to find out maximum how many simultaneous screen sharing session it can allow, without any quality compromise. By quality I mean, speed of transmission and no data loss during transmission. I simulated the load using ffmpeg command.

  For this, instead of transmitting a live captured screen, I am transmitting (uploading) a FLV file stored in my local to the Red5 server using ffmpeg command. In the receiving client side, I am starting to download (transmitting) this same FLV file after 5 secs since the upload has started. This is working fine when I am running this test for less than 10 pairs of upstreaming-downstreaming sessions. But, when the number is increasing beyond 10, I am observing significant packet loss in transmission.

  Here are the commands I am running in a loop. The loop count is the number of streaming pairs.

  1. upstreaming : ffmpeg -re -i  -f flv -ar 22050 "rtmp://" -report
  2. downstreaming : ffmpeg -re -i "rtmp:// live=1"  -report

  The and are set in such a way, that in the downstream I will download the same uploaded file. "rtmp ://" are the same in both the cases. I am not doing the upstream in record mode, hence, no physical file is getting saved in the server side. When I am analyzing the file I received in the receiving client side, it is a poor quality video due to frame loss. Uploading and downloading machines are two different machines. I ran the test for many hours, repeating the same 10 simultaneous streaming sets. Each set is consistently giving the same results.

  What is puzzling me is, this is working fine without any packet loss for less that 10 simultaneous streaming. I searched about it in various forums, but none of the answers were applicable for this scenario. For a while I was thinking that Red5 has limited capacity, but I found many posts saying Red5 can easily scale up to take very big load. Does that mean, the problem is in my configuration ? I am not sure which are to focus on.

  An example log snippet :

  Lots of missing data at downstream side. For e.g. between frames 101 and 102 there is a difference of 25 sec. On replaying the video there is a stoppage for this much time.In this time gap all the frames are lost.

  frame=  101 fps=1.0 q=14.5 size=    2650kB time=00:01:41.00 bitrate= 214.9kbits/s
  
  frame=  102 fps=1.0 q=13.2 size=    2763kB time=00:02:06.00 bitrate= 179.6kbits/s

  Any help is appreciated !

• Image generated by System.Drawing.Image is 10x larger than original file

  13 mai 2021, par Ninbura

  I've written a Powershell script that extracts one frame from a dshow capture device using FFmpeg, and then copies the output image from a file to my clipboard :

  


  $screenshot = [System.Drawing.Image]::FromFile((Get-Item -Path $outputFilePath))
[System.Windows.Forms.Clipboard]::SetImage($screenshot)
$screenshot.Dispose()


  


  The original file is a 1MB JPEG, but the image in my clipboard is an 11MB PNG. When I go to paste this file into something like Discord the file size is too large, unless you have Discord Nitro.

  


  Is there any way to avoid this file size inflation while using [System.Drawing.Image]::FromFile() ?

  


  Here's an easy test template :

  


  Add-Type -AssemblyName System.Drawing
Add-Type -AssemblyName System.Windows.Forms

$screenshot = [System.Drawing.Image]::FromFile((Get-Item -Path "Path to picture"))
[System.Windows.Forms.Clipboard]::SetImage($screenshot)
$screenshot.Dispose()


  


  Edit 2021/05/13 :

  


  Interestingly, it's starting to seem like this is a limitation of copying an image to a clipboard. I decided to dig deeper and see if the image size was increasing when I drew the image with System.Drawing.Image or when I pushed it in my clipboard with System.Windwos.Forms.Clipboard. If I save the image before putting it my clipboard like so, it retains the original file size, even if I change the extension to png :

  


  Add-Type -AssemblyName System.Drawing
Add-Type -AssemblyName System.Windows.Forms

$screenshot = [System.Drawing.Image]::FromFile((Get-Item -Path "C:\Users\gabri\Pictures\FFScreenshots\2021-05-12 19-04-26.845.jpeg"))
$screenshot.Save("C:\Users\gabri\Pictures\FFScreenshots\test.jpeg") # test.png makes no difference
$screenshot.Dispose()


  


  So it would seem that the image grows exponentially in size specifically when I push it to my clipboard. I verified this by simply uploading an image to Discord, copying it to my clipboard (click image > Open original > right click and copy result), and uploading it again. Sure enough, the new image pasted from my clipboard was around 10x the size of the original image (1MB vs 12MB). I wondered if this was a "limitation" of Windows so I ran the same test on Ubuntu 20.04. Uploaded an image to Discord, copied the image to my clipboard, and uploaded it again. Once again the image was about 10x larger, though slightly smaller than then when I uploaded the image from my clipboard on my Windows machine (11.4MB VS 12MB). Just to make sure this wasn't limited to Discord, I ran the same test with Gmail in-browser on both operating systems, same result.

  


  Alas, there appears to be something happening when you copy an image to a clipboard that drastically increases file size. One thing's for sure, whatever's happening is definitely over my head.

  


• AAC : Fix M/S stereo encoding

  3 mars 2015, par Claudio Freire

  AAC : Fix M/S stereo encoding
  This patch fixes a pointer arithmetic bug in adjust_frame_information that resulted in heavily corrupted audio when using M/S encoding. Also, a backup copy of untransformed coefficients has to be kept around or attempts at re-processing the frame (which happens when hevavily overspending bits during transients) will result in re-encoding of the coefficients and subsequent corruption of the resulting stream.
  A/B testing shows the bug as corrected, but still cannot prove that M/S coding is a win at least in numbers. Limited listening tests do show improvement on M/S encoded samples in lower bitrates, but they’re hidden among the other artifacts that remain to be corrected in the encoder.
  Some of the regressions flagged in the report do show poor stereo image (but not buggy), so M/S encoding is clearly not good enough yet to be defaulted to auto.
  In numbers, Patched against Unpatched, stereo_mode auto :
    Files : 114
  
    Bitrates : 6
  
    Tests : 683
    Serious Regressions : 0 (0%)
  
    Regressions : 0 (0%)
  
    Improvements : 227 (33%)
  
    Big improvements : 92 (13%)
  
    Worst regression - mybloodrusts.wv - 256k
  
      - StdDev : 28.61       pSNR : -0.43     maxdiff : 1372.00
  
    Best improvement - 60.wv - 384k
  
      - StdDev : -369.57     pSNR : 45.02     maxdiff : -13322.00
  
    Average          - StdDev : -80.56       pSNR : 2.49      maxdiff : -8858.00
  Patched against Unpatched stereo_mode ms_off shows no difference.
  Patched stereo_mode auto vs Unpatched stereo_mode ms_off shows a small average improvement, just not too significant :
    Serious Regressions : 0 (0%)
  
    Regressions : 10 (1%)
  
    Improvements : 45 (6%)
  
    Big improvements : 2 (0%)
  
    Worst regression - Illinois.wv - 256k
  
      - StdDev : 33.20       pSNR : -2.03     maxdiff : 477.00
  
    Best improvement - song_of_circomstances.flac - 384k
  
      - StdDev : -3.97       pSNR : 7.61      maxdiff : -826.00
  
    Average          - StdDev : -10.25       pSNR : 0.20      maxdiff : -281.00
  Signed-off-by : Michael Niedermayer <michaelni@gmx.at>
  

  • [DH] libavcodec/aac.h
  • [DH] libavcodec/aaccoder.c
  • [DH] libavcodec/aacenc.c