Please point me to the right site if Sound Design is not the correct one.

I have been trying to hide a custom piece of audio within music so that it could not be heard, but will still show up when having a spectrogram run. I did my original testing on a Beatles track, but I threw together a quick siren so I can upload the audio here. *(1)

What I have tried to do with my track is I put in the music on tracks 1 and 2 and then I put the hidden audio on track 3. As for track 4, I put the inverse of track 3 which was created by the Inverse function from Audacity.

This works in Audacity itself and when playing on my Nexus 5's singular speaker. It does not work outside of that software or wearing any kind of earbud or headphones (that I have tested).

My goal is to make this work well enough to the point where I can include the music in a video game and eventually sell the DLC without negatively effecting the quality of experiencing the tracks.

I have been thinking of how it might be possible to make a song around the image sound, but since it looks like spectrograms are produced with frequencies and not amplitude (at least according to this image produced by Sox), so I don't know if it is possible outside of having a separate track.

Siren with hidden audio (Produced by Sox)

  • (1) I will still have to see about uploading the track or the Audacity project if it is necessary to answer the question.

Edit: To ask Tetsujin to confirm this is what he means by mixing the track to mono, I am putting this picture of my Audacity project. I am still having the issues with the sound though (nothing has changed).

The audacity project with mono sound

2 Answers 2


To expand on tetsujin's fantastic answer, i think a bit of depth on the subject of cancellation would be helpful. I will keep it simple, but i think it will make things click better.

So, when you zoom in on a waveform, you'll get to the point where you can see the compressions and rarefactions. When you see the waveform above the zero line, this denotes a compression of air. The speaker follows this by physically pushing out, which compresses the air in front of it and forms a compression wave (basically a high pressure zone). Then (when the waveform is below zero) the speaker pulls in and creates a rarefaction (or low pressure zone). The speaker does this really fast and blah blah creates sound. That's not the important part for cancellation. Focus on the speaker movement.

So, when you put a track out of phase with itself (the EXACT same track), the speaker is instructed to push and pull, the same amount, at the same time. This negates and the speaker stays still.

Now, as tetsujin said, you can't cancel if you have the tracks spread in stereo, because the speakers are free to move individually. The result would be either partial cancellation or none at all. You would need to have them panned the same direction (exactly).

Regarding your edit, that looks correct, but I can't see if the tracks are panned, which, as I said, will make it not work.

Lastly, as an aside, that picture DOES show amplitude. If you see the bar on the right labeled dBFS (decibel full scale, a digital realm amplitude scale), that shows the colors for amplitude. So if you look in the middle, you will see that the different colors actually represent the dBFS level at that frequency.

  • I had added the hidden audio again and made sure that it was not off by any amount of time and it still plays the sound on my earbuds. I am going to have to look into embedding the image directly into the music track, however, this does look promising for a potential application that is enhanced by being able to differentiate between different types of speakers. Nov 8, 2017 at 21:32
  • @SenorContento Ok so it's not off in time, but is it off even a little in panning? Being in mono does NOT mean they are panned the same. Hope that helps
    – user22688
    Nov 16, 2017 at 15:41

Total phase cancellation [which is what you appear to be looking for] only works in mono. Each time one channel has a positive voltage, the other has an identical negative voltage. When summed to mono, this equals a constant 0 volts, i.e. silence.

In stereo you can still hear both signals, because in open air there is a time-difference between the channels & on headphones each ear gets a totally different signal; being out of phase it will appear to come from 'outside the speakers' because your ears aren't really equipped to deal with that type of effect in everyday life, but it won't cancel out.

If you actually mix your phase-inverted track to mono within an otherwise stereo signal, it will, in effect, cease to exist at all.

  • 2
    +1 for understanding what the OP is talking about...
    – frcake
    Nov 7, 2017 at 8:27
  • I presume that the time-difference is true regardless if both the 3rd and 4th track are mono and that the only way is to merge the tracks into one? Nov 7, 2017 at 12:43
  • If you duplicate & phase-invert a stereo track, then in real terms you are phase-inverting 2 mono tracks, one either side & each will totally cancel the other out. No signal will result.
    – Tetsujin
    Nov 7, 2017 at 13:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.