+The default sound API is the first of the ones listed below that are available.
+Usually this implies ALSA under Linux and Core Audio under OS X.
+.TP
+.B \-\-alsa\fR, \fB\-a
+Use ALSA to play sound.
+Only available on Linux.
+.TP
+.B \-\-oss\fR, \fB\-o
+Use OSS to play sound.
+Only available on Linux and FreeBSD.
+.TP
+.B \-\-core\-audio\fR, \fB\-c
+Use Core Audio to play sound.
+Only available on Macs.
+.TP
+.B \-\-device \fIDEVICE\fR, \fB\-D \fIDEVICE\fR
+Specifies the audio device to use.
+See
+.B "DEVICE NAMES"
+below for more information.
+.TP
+.B \-\-command \fICOMMAND\fR, \fB-e \fICOMMAND\fR
+Instead of sending to a physical audio device, invoke \fICOMMAND\fR using the
+shell and write audio samples to its standard input.
+Currently the input will be 44100KHz 16-bit signed stereo samples.
+If \fICOMMAND\fR exits it is re-executed; any samples that had been written to
+the pipe but not processed by the previous instance will be lost.
+.IP
+.B \-\-device
+is redundant with this option, but you might want to set
+.BR \-\-pause\-mode .
+.IP
+As an example,
+.B "-e \(aqcat > dump\(aq"
+would log audio data to a file for later processing.
+You could convert it to another format with, for instance:
+.IP
+.B "sox -c2 -traw -r44100 -s -w dump dump.wav"