Writing Video conference software for Windows

[Paolo Pantaleo]

Hi,

I need to write a software that allow to see the desktop and hear the
microphone capture of a remote PC across a network. I need to do that
for a unviresity assignement. The software must run on Windows. Since
I like Python very much I am thinking to write that software in
Python. Do you thinkit is a good choice? Are there libraries for audio
compression (OGG or MP3 or maybe GSM or something like realaudio) and
video compression (btw what can be some good libraries to transmit
images of a desktop in a bandwidth-efficent way?). What about capture
of audio and screen? (Probably i will need some Win32 system call,
there are bindings in Python, aren't they?)

If I needed to write some Python modules in C, would it be difficult?

Can some language like C# or C++ may be better?

Thnx
PAolo

 

[Jordan]

If you're going to need win32 system access use the win32all python
extension (very, very good extension). Do you need single frame image
capture, or constant video stream? PIL can be used for the first, it
might also be usable for video, I'm not sure. For sound, python comes
with some built in libraries, but you should also take a look at
pysonic http://www.cs.unc.edu/Research/assist/developer.shtml. For the
bandwidth efficiency issue, what type of connection are you using? The
socket module is quite capable of transmiting whatever data you have,
so unless you're thinking of implementing some mini bittorrent like
network in an attempt to save bandwidth I don't know what you can do
about that. There's an extension called IPqueue which might give you
somewhere to start for packet/bandwidth manipulation. Check out The
Vaults of Parnassus, which has a lot of stuff (including ogg/mp3
converters last time a check). Big question, is this supposed to act
like a remote desktop, or just show what's happening? Start by
searching Google, it's very useful.

 

[Paolo Pantaleo]

If you're going to need win32 system access use the win32all python
extension (very, very good extension). Do you need single frame image
capture, or constant video stream? PIL can be used for the first, it
might also be usable for video, I'm not sure.

Well I need something like 5-10 fps. An issue is the comression
method: MPEG and friends aren't good (I think) for compressing stuff
with sharp borders. Maybe I could use A sequence of PNG images, but it
isn't a great solution.

For sound, python comes

with some built in libraries, but you should also take a look at
pysonic http://www.cs.unc.edu/Research/assist/developer.shtml. For the
bandwidth efficiency issue, what type of connection are you using? The
socket module is quite capable of transmiting whatever data you have,
so unless you're thinking of implementing some mini bittorrent like
network in an attempt to save bandwidth I don't know what you can do
about that. There's an extension called IPqueue which might give you
somewhere to start for packet/bandwidth manipulation. Check out The
Vaults of Parnassus, which has a lot of stuff (including ogg/mp3
converters last time a check). Big question, is this supposed to act
like a remote desktop, or just show what's happening? Start by
searching Google, it's very useful.

Well the bandwidth issue is most of all related to video compression
(see above). Well maybe 256 kbps would be nice.

It should just show what's happening.


PAolo

 

[sturlamolden]

Well I need something like 5-10 fps. An issue is the comression
method: MPEG and friends aren't good (I think) for compressing stuff
with sharp borders. Maybe I could use A sequence of PNG images, but it
isn't a great solution.

You probably want to use is VideoCapture for capturing digital video
from e.g. USB webcameras (videocapture.sourceforge.net), Twisted for
asynchronous sockets (twistedmatrix.com), and PyMedia for video and
audio codec (pymedia.org). For GUI and graphics there are a number of
candidates, e.g. PyGTK or PyGame.

Video conference is an I/O bound task (except for the audio/video
codec), and you will consequently gain nothing by using C++ over
Python. The only part that is CPU-bound part (where C++ could
theoretically be better) is the media codec, but you can leave that to
PyMedia's C extensions. Python will most likely not be the bottleneck!

Sure, you could use C++ instead. However, you would quickly be
frustrated with Windows' extremely complex native APIs. E.g. DirectShow
for digital video capture, Winsock for asynchronous sockets,
Microsoft's multimedia APIs for codec, and perhaps MFC for GUI. Using
Python is a good advice!

 

[Ravi Teja]

I need to write a software that allow to see the desktop

That would be the VNC protocol. Don't reinvent the wheel.

VNC is relatively efficient in that it only updates the portions of the
screen that changed.

Maybe this project could be your starting point.
PyVNC
http://bdash.net.nz/blog/2005/07/17/pyvnc-the-beginnings-of-a-python-vnc-client/

and hear the
microphone capture of a remote PC across a network.
I need to do that
for a unviresity assignement. The software must run on Windows. Since
I like Python very much I am thinking to write that software in
Python. Do you thinkit is a good choice? Are there libraries for audio
compression (OGG or MP3 or maybe GSM or something like realaudio)

PyMedia can record and it works for mp3 and oggs.

and video compression (btw what can be some good libraries to transmit
images of a desktop in a bandwidth-efficent way?).

See above. You can lossy compress images in VNC.

(Probably i will need some Win32 system call,
there are bindings in Python, aren't they?)

The above modules should take care of everything. BTW, PIL can take
screen shots if you want to do this manually.

You need to create a streaming server for audio. Take a look at Edna, a
streaming MP3 server. http://edna.sourceforge.net/

Also of interest ... Twisted and Shtoom.

And finally... since you are on Windows - you can just try using the
NetMeeting ActiveX control through Python. It already does everything
you want.

Disclaimer: While I tinkered with many of the above, I have not used
most of the above projects recently. These are just leads for you to
explore.

 

[Paolo Pantaleo]

Thnx everybody for the precious help

Someone said about VNC... I'll take a look, but since it is an
exercise I need to do it, I can't just say someone else arelady did
that

Everything seems quite useful. I forgot two specifications:

1. Screen should be split in small squares and only the changing
squares must be transmitted (Ok it shouldn't be too difficult)

2. The comunication must be in multicast

I will spend some time testing the resources.

PAolo

 

[Paolo Pantaleo]

Thnx everybody for the precious help

Someone said about VNC... I'll take a look, but since it is an
exercise I need to do it, I can't just say someone else arelady did
that

Everything seems quite useful. I forgot two specifications:

1. Screen should be split in small squares and only the changing
squares must be transmitted (Ok it shouldn't be too difficult)

2. The comunication must be in multicast

Twisted supports multicast ( example
http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/425975)

 

[Ravi Teja]

Someone said about VNC... I'll take a look, but since it is an

exercise I need to do it,

Exercises typically need you to implement, not invent (leave that for a
thesis or a dissertation). Rather than invent VNC, you could just
implement it on your own from the specs.

http://realvnc.com/docs/rfbproto.pdf

I can't just say someone else arelady did that

No. But you can build upon it . Few construct software from the
scratch.