pyvnc2swf

back [Japanese]

Homepage: http://www.unixuser.org/~euske/vnc2swf/
Discussion: http://lists.sourceforge.net/lists/listinfo/vnc2swf-users

$Id: pyvnc2swf.html,v 1.15 2006/10/30 00:59:24 euske Exp $


Pyvnc2swf is a cross-platform screen recording tool. It captures screen motion through VNC protocol and generates a Shockwave Flash (SWF) movie. Pyvnc2swf suite comes with three Python programs:

For questions, please read the FAQ and list archives before sending me emails.

Terms and Conditions: Pyvnc2swf comes with ABSOLUTELY NO WARRANTY. This software is distributed under the GNU General Public License.


Installation

In all platforms, the following packages are required:

In most Linux distros, these packages are readily available. In Mac OS X, you would need an additional OS X build of Pygame package.

Also, you need at least one VNC server:


Recording

vnc2swf.py program captures a VNC sessions and records it in either SWF or VNCLog format. This is a VNC client and communicates directly with a VNC server. A user need to start a VNC server in advance.

vnc2swf.py runs in two different modes: GUI (Graphical User Interface) mode and CLI (Command Line Interface) mode. In the GUI mode, start recording by clicking the "Start" button. Then choose the "Save as..." command from the "File" menu to save the recorded movie to a file. In the CLI mode, a user needs to specify the output filename from command line. Recording is started immediately. In both modes, a user is prompted for a VNC password if the server requires authentication (and unless the user doesn't specify the password file). In the CLI mode, hit Control-C to stop recording. After finishing recording, it generates two files with the specified name: a .swf and .html file. The .html file contains an HTML tag and a javascript code to provide seek bar function.

A user can choose three different methods to encode movie image: "shape", "video" and "vnc". The first shape encoding (default) provides a reasonable movie size. The second encoding method, video provides a smaller SWF movie. This is, however, not recommended to use within vnc2swf.py for two reasons: This type of encoding is only supported by Flash Player version 7 or newer. Also, generating a movie with on-the-fly video encoding is slower so you might experience frame dropping. Actually, you can convert a shape-encoded movie into video-encoded one after recording by using edit.py, so anyway you don't need to use this method when recording. The third encoding method is vnc. This method generates a .vnc (VNCLog) file, which is compatible with vncrec output file. You can convert it to a SWF movie with edit.py. A .vnc file is not a SWF movie by itself, but its encoding is the fastest.

NOTE: Unlike the C version, vnc2swf.py doesn't handle any user interaction. If you want to control the server's desktop, you need to launch vncviewer or its equivalent separately.

Syntax

Example

(Set up a virtual screen)
$ vncserver -geometry 640x480
...
$ vnc2swf.py -o out.swf localhost:1

(Record an existing screen with audio recording)
$ x11vnc -localhost -viewonly -wait 10 -defer 10 &
...
$ vnc2swf.py -o out.swf -S "arecord -r 22050 voice.wav" localhost:0

If you're using x11vnc, see also recordwin. This is a convenient script to record a particular window.

Recording Tips

Options

-n
Console mode (no GUI).

-o outputfile
Specifies the output filename. This option is required in CLI mode. In GUI mode, when not specified it prompts the user for the filename. The movie encoding type is usually inferred from the filename, so the filename should end with either ".swf" or ".vnc". Otherwise, the user need to specify the output movie encoding with -t option (see below.)

-C clipping
Specifies the clipping rectangle. The geometry must be as form of "widthxheight+left+top" (e.g. "400x300+120+0"). Unlike other X11 applications, all rectangular components are required. Negative values are not supported.

-r framerate
Specifies the framerate in fps. (default=12.0)

-t encodingtype
Specifies the output movie encoding method ("shape", "video" or "vnc"). When omitted, the encoding type is automatically inferred from the filename (*.swf = shape, *.vnc = vnc).

-N
Disables cursor pseudo-encoding. Pyvnc2swf normally tries to use cursor pseudo-encoding to capture a mouse cursor position so that a cursor can be moved separately from a screen image and it reduces the movie size. However this might not work with some vnc servers. This option can be used for disabling the function.

-P passwdfile
Specifies a password file. If specified, its content is automatically loaded and supplied as password when a VNC server requires it. A password file contains an encrypted password string and can be created with vncpasswd. A user can directly specify ~/.vnc/passwd, which normally contains the password for the local vnc server.

-e vncencodings
Specifies the preferred encodings for VNC image transfer (this is different from movie encoding). Normally you don't need to change this option. Encodings are comma-separated integers (default="5,4,0"). Changing encodings might improve recording performance.

-S subprocess (Supported on Un*x only, Python 2.4 or above is required)
Set a command to run during recording. This option is useful for recording voice with a separated program. A command line is a space-separated sequence of arguments which are passed to a child process like a usual shell command line. However the command line is not passed to a shell and the arguments are directly passed to the child process without any substitution. A child process is started immediately after recording starts and sent SIGINT after recording finishes.

-d
Increases debug level.


Editing

edit.py program is for editing or reorganizing one or multiple movies generated by vnc2swf.py. This program also supports converting a .vnc file into .swf movie, changing the encoding method (shape->video), attaching MP3 audio file to a movie, extracting images from a movie and resampling/scaling/clipping a movie image.

edit.py currently supports only command line interface. The user must give one output filename and one or more input filename(s). Input movies are concatenated sequentially (in the specified order) and the desired effects are applied.

Syntax

$ edit.py -o outfile.swf [options] infile ...

Example

(Convert .vnc file into .swf with compressed video encoding)
$ edit.py -o out.swf -c -t video input.vnc

(Attach an mp3 file to .swf)
$ edit.py -o out.swf -a voice.mp3 in.swf

(Concatenate two movies and extract the frames into another movie)
$ edit.py -o out.swf -f 100-200,350- movie1.swf movie2.swf

(Clip the top left area of the movie and shrink it to half the size)
$ edit.py -o small.swf -C 320x240+0+0 -s 0.5 in.swf

(Convert .swf into MPEG)
$ edit.py -o out.mpg input.swf

Options

-o outputfile
Specifies the output filename. This option is always required.

-c
Compress a movie with zlib. Compressions is supported for both (shape and video) types of movies, but usually it's most effective when applied to video-encoded movies.

-t encodingtype
Specifies the encoding method for an output movie. When this option is not specified, the encoding type is inferred from the output filename extention.

TypeExtensionDescription
shape.swfSWF movie (default)
videoSWF movie with stream video encoding
mpeg.mpgMPEG movie (requires PyMedia)
bmp.bmpBMP image sequence
png.pngPNG image sequence

Note that video encoding is supported only with Flash Player version 7 or newer.

-V
Encodes a movie with video method. This is equivalent to -t video.

-f frames or
-F frames
Specifies a sequence of frames presented in the output movie. When omitted, all frames are presented in the original order (default). Frames can be specified with a comma-separated list of integers. A range of integers can be also specified by using - (hyphen) sign. For example: 10,200,300-400 specifies the frames whose number is 10 and 200, plus every frame between frame 300 and 400. The beginning (or ending) frame number can be omitted (e.g. -100 or 300-). In this case, the first (or last) frame number is used as the other end of the range.

-F option and -f option is same except that -F doesn't chop the audio while -f does. When you're putting audio on the movie, if you use -f and -a option at the same time, it chops an mp3 file according to the selected frames. However, when you want to cast a continuous sound (such as music) onto a whole movie, this might not be the desired effect. In such a case, use -F instead of -f.

-a mp3file
Attaches mp3 file(s) to the movie. Multiple mp3 files are concatenated in the specified order.
NOTE: When specifying multiple mp3 files, make sure every file has the same bitrate as the first one. And do NOT use "variable bitrate" (VBR) mp3 files, as the SWF format doesn't support them!

-s scaling
Rescale the movie image with a specified ratio which is given as a fraction.
NOTE: There will be noises with non-multiple scaling (e.g. 0.7).

-C clipping
Clips the movie into a specified rectangle. The geometry must be as form of "widthxheight+left+top" (e.g. "400x300+120+0").

-K keyinterval
Insert a keyframe in every N frames. Keyframes work as hints for Flash Player and are useful for seeking a frame within a long movie. When the number of total frames exceeds 10,000, inserting keyframes in every 500 frames (-K 500) is recommended.

-r framerate
Changes the frame rate of the movie. This option itself doesn't do any frame resampling and simply changes the movie speed. When this option is omitted, edit.py tries to keep the original frame rate in the output movie.

-R resampleframes
Resample (or "thin down") a movie by picking one from every N frames.

-S skipmp3frames
Skip the first N mp3-frames of the mp3 file. When 's' is appended to the number (such as '1.0s'), it indicates the number of seconds instead of frames. This option is useful when there is a time lag between a recorded image and audio.

-B blocksize
Sets the blocksize for video encoding method (default=32). This must be a multiple of 16.

-b
Suppress a seekbar in a generated html file.

-l
Disables movie loop in a generated html file.

-d
Increases debug level.


Previewing

play.py is a simple player for a .swf or .vnc file. This program might be useful for spotting the right frame number in a recorded movie. However its speed is awkward and audio output is not supported. It only supports vnc2swf-generated files and cannot play general SWF movies.

The player accepts the following keys:

Syntax

$ play.py [options] moviefile ...

Options

-r framerate
Specifies the framerate to play.

-s scaling
Specifies the scaling ratio with a fraction.

-C clipping
Specifies the clipping rectangle. The geometry must be as form of "widthxheight+left+top" (e.g. "400x300+120+0").

-d
Increases debug level.


Frequently Asked Questions

I get MemoryError in Windows. Why?
It is known that MemoryError occurs when there is a communication problem between vnc2swf.py and the VNC server or the screen size is too big. In the former case, make sure to use RealVNC version 4 or newer and allow local connections. This can be done by the following: Open the VNC server property window, click "Connections" tab, and check the "Only allow connections from the local machines" box. In the latter case, you can either choose the smaller desktop size or try to narrow the recording area by specifying a clipping rectangle with -C option. This problem doesn't exist in pyvnc2swf 0.7.2 or newer.

I found weird artifacts in a recorded movie. How to fix it?
Probably you played the SWF file directly. A SWF movie should be played in the exactly same size as it is recorded. Use an HTML file generated by vnc2swf.py or edit.py.

How to make a movie with audio?
Record your audio separately and encode it with an MP3 encoder like Lame. Then combine them with edit.py.

A browser freezes when I try to seek within a movie.
Put keyframes in every hundreds frame. Try -K option in edit.py. Normally putting keyframes in every 500 frames (-K 500) gives a reasonable result.

The mouse cursor doesn't appear on a movie.
With pyvnc2swf version prior to 0.8.0, the mouse cursor was not recorded with OSXVnc. This can be solved by upgrading to pyvnc2swf 0.8.0 or newer.


Structure of the Program

For those who are interested in learning or extending the program, I drew a simple figure to explain how data goes between objects within pyvnc2swf.


Yusuke Shinyama <yusuke at cs dot nyu dot edu>