Open Source Ecosystems

Record

The record.py script uses gtk's "get_from_drawable()" function to take screenshots and modified alsa's arecord to record the sound.

More details how it works:

The arecord and aplay alsa utilities are actually one and the same program, so it was stripped down, all signals removed, all playback removed, all header files merged into it, then a "capture_stop" global variable was added, which if set to 1, the main audio capture loop will end. Then it is called from cython using nogil in it's own thread and when the user wants to end it, the "capture_stop" is set to 1. The main python thread takes screenshots in periodic intervals (15 fps by default) and if it's late, it skips the frame, so that the next one is on time. All screenshots are saved to the "data" file in the temporary directory, which is later read and converted to a set of png images. The audio is saved to a wav file.

It is then your job to create a video from it (and to convert sound to any format you like, like ogg vorbis), it tells you some suggestions (see below). Also the amplifier.py script is provided to amplify volume of the wav file in case it is too quiet.

Usage

Install the following packages in Debian/Ubuntu:

sudo apt-get install python-gtk2 libasound2-dev

and run it:

./record.py

It produces a wav file in the tmp directory (it tells you which) and a set of png images. It also suggests you exact commands how to create videos from it, using either mencoder or ffmpeg2theora (it's up to you).

In my experience mencoder can't sync the sound and video correctly, while ffmpeg2theora just works. If the sound is too quiet, use the amplifier.py utility to amplify the wav file before converting it to ogg or the "normalize-audio" program (better results) from the Debian package of the same name (it operates inplace on the wav file).