lurker, an audio silence splitter
Mattias Wadman <mattias.wadman@gmail.com>

INSTALLATION:

You will need some kind of curses development packge installed.
apt-get install libncurses5-dev in debian.

Run "make" and copy the lurker binary somewhere if you want.


EXAMPLE USAGE:

Realtime audio splitting:
recoding program | lurker -t 0.15 -r 10 -f 0.5 -o "%F/%F_%H:%M:%S.wav"
Threshold 0.15, split if 10 seconds of silence, remove recording if less then
0.5 seconds long and output files relative to current directory.

Splitting an previously recorded file:
lurker -s "1970-10-10 07:00:00" -i long_old_recording.wav
With -s the amount of audio time offseted from the start time will be used
insted of system clock when generating output filenames.

OSS:
install bplay (apt-get install bplay in debian)
brec -w -b 16 | lurker ...

ALSA:
install arecord (apt-get install alsa-utils in debian)
arecord -q -t wav -c 1 -f S16_LE | lurker ...


OTHER HINTS:

lurker wants 16 bit mono PCM (not compressed) audio in wav format as input.

lurker automaticaly creates non-existing directories in the output path.

Read man page for brec and arecord if you want higher sample rate, record for a
limited amount of time etc.

The program should work fine on big endian machines (powerpc, SPARC, ...) but
have not been tested yet.

Wav filename is appended with "recording.wav" (configurable) while recording
to make it possible to exclude them while batch encoding, see batch_encode
template script.

If the rotating progress indicator stops the program providing audio data to
lurker has stopped or is blocking for some reason. If not, you have probably
found a bug in lurker.

If you dont trust your system clock for some reason you can try to use the -s
option with current date and time (or value "now" if system clock is correct
at program start).

If you see message of the type "overrun!!! (at least 4277.546 ms long)" they
are from arecord. I dont know the reason, high CPU usage maybe?

Dont use additional directories (includes "/") with the -a (append recording)
option, it will fail.

The slice divisor option decides the amount of audio data to be processed at a
time (how much to run thru the RMS function etc), if set low it will require
more "noise" to trigger.


And last, please let me know if you use this program for something interesting.