Hiding and recovering secret audio?

Hi. I’d like to be able to modulate as a carrier a subaudible tone or otherwise hide an audio signal so it can’t be heard.
I would also like to be able to make the signal audible again.
I know a lot of radios use supportable control codes and I think FM stereo is modulated in a subaudible way.
I just want to do it as an exercise to teach children about hidden messages.
As a newbie, I’d appreciate it if you’dplease give me some specific steps if it’s possible.

The magic-phrase is “silent subliminals”.

Children may be able to hear 18kHz, but they won’t be able to understand the modulated-speech, it’s effectively scrambled.

I think FM stereo is modulated in a subaudible way.

Super audible. It goes up. Regular audible sound stops (in FM) at about 15,000Hz. But wait, there’s more. There is a tiny pilot tone at 19,000Hz which is precisely half of the stereo separation information up at 38,000Hz. If your station is cursed with SCA (Elevator Music), there is yet another carrier at 67,000Hz.

The problem is going to be creating one of those demonstrations in Audacity. Most soundcards and audio music equipment won’t pass much higher than 20,000Hz. That’s considered the limit of human hearing. We had one radio station jock who could hear reliably up to 19200. We wanted to charge money to see him, but the university didn’t think that was the best idea.

I might take a giant step backward and see if anybody else has any ideas. You need to create modulation and then shove it so far up in pitch that the ear can’t find it. Then reverse it later.

I can tell one problem immediately. You can’t use 44100 (audio CD) sampling frequency. The digital sampling has to be at least twice the pitch of the highest sound. 44100 can handle music up to about 22,000Hz.

So, we start with AM radio quality voice at 5000Hz. Somehow make AM and that gives you a 10,000Hz channel Which oddly enough is the channel spacing for AM radio.

That means your normal sound has everything covered up to about 20,000Hz and plunk the Mystery Voice on top and that takes you out to 30,000Hz. Digitize that and you would need at least 60,000Hz Sampling. Studios use 96,000 sampling a lot and everything would fit.

So how are you going to make basic AM out of somebody’s voice.

Anyone? Anyone? Bueller ? Bueller?


Missed one. The carrier for this magic am radio channel is 25,000Hz. Way above hearing. The voice goes 5,000Hz up and 5,000Hz down.

There ought to be a way to force one of the regular Audacity tools to make AM.



The Vocoder effect modulates a voice on the Left stereo track onto a carrier on the Right stereo track.

I produced a 5000Hz (AM Voice) track and smashed it onto a 20,000 constant tone carrier.

Screen Shot 2020-02-23 at 18.29.28.png
This is the new work.

Screen Shot 2020-02-23 at 18.19.22.png
If you do a spectrum inspection, you find that everything is clustered around 20,000Hz and even though it’s perfectly loud, its completely silent.

Screen Shot 2020-02-23 at 18.15.51.png

Jury’s out how to decode that back to normal speech. I think this is where I got stuck last time.

I think I hear my mum calling.


This is interesting. Are you able to offer a more detailed walk-through?