The new Nyquist scripts can encode a speech recording to (almost) silence, and then decode back to "telephone quality" speech. Here's the encoder:



(defun audio-filter (sig)
  (setf sig (highpass8 sig 300))
  (lowpass8 sig 3000))

(defun modulate (sig)
  (mult sig (hzosc 16000)))

(defun sideband-filter (sig)
  (highpass8 sig 16300))

(defun subliminal (sig)
  (setf audio (audio-filter sig))
  (setf modulated (modulate audio))
  (sideband-filter modulated))


(subliminal *track*)