Steganography and “Silent Subliminals”

AM Encoding is the most popular technique for creating “Silent Subliminal” recordings, which advocates claim can bring about positive shifts in one’s mindset. The technique was largely based on the 1989 US Patent US5159703A by Oliver M. Lowery for a “Silent subliminal presentation system”.

This plugin is an audio frequency AM encoder / decoder for speech. As I’ve lost so many links on this forum, I’ve decided to host the plugin on my own blog. More information, and the plugin itself may be downloaded HERE.

Hi Steve, thank you for your plugin. Is it also possible to convert a recording from 0-20.000Hz to 0-5.000Hz (or cut the the higher frequencies from 5.000-20.000) and to move this converted recording (without encoding) to carrier of 15.000Hz (or below)? So that I listen to some voices in radio quality instead of something nobody can understand (maybe birds as these encoded recordings sounds like this ;-). Thank you for reply…

Like, say, 12000Hz …

If you use 12000Hz as carrier, the recording could use larger range up to 8000Hz instead of only 5000Hz and should be better to understand.

Shifting the speech 12kHz is what makes it incomprehensible.
Increasing the bandwidth from 5kHz to 8kHz is not going to make it understandable.

I can only hear up to 8600Hz so I can not listen to your sample. How can I cut higher frequencies from my original recording (range of 12.400-20000Hz should be cut off) and shift this cut recording (range from 0-12.400) without any encoding to a carrier somewhere below 8600 Hz? So I can try on my own…

Or is this “shifting” (1:1 of my cut recording from 0-12,4kHz to 8,6-20kHz) only possible including encoding and change some parameters? Thank you.

Not possible : like trying to fit size-12 feet into size-8 shoes.
(12.4kHz into 8.6kHz).

Can morph the spectrum,
but if you push it too far speech will be incomprehensible …

after-before frequency magnet

So my hearing is pretty bad, I can hear whistle sounds at 11500 hz on the loudest volume (I tweaked the code so the carrier was 11500 hz) disregarding if subliminal messages work or not, would this still give the correct output as if it was on the default carrier, 16000 hz? or do I need to change something else in the code if I use a carrier hz of 11500?

To set the carrier frequency to 11500, change the line:

(psetq *cf* 16000     ; Carrier frequency


(psetq *cf* 11500     ; Carrier frequency

What would be the best way to improve the quality of the subliminal messages? Would changing the sample rate to 192000 hz or 384000 hz be a viable way to improve the quality of the messages so it isn’t telephone quality? Or maybe changing the “High band frequency”? I’ve changed this so it’s higher and the message is a lot clearer once decoded, however I have no idea if this would help when the audio file is encoded.

Audio formats are typically not capable of storing ~100,000Hz signals.
Audio transducers, (headphones/loudspeakers), and audio electronics are not designed to reproduce ~100kHz signals either.
and even if they did, humans could not hear it.

1 Like

so would 96000 hz be the maximum sample rate you could use essentially, since it’s still under 100,000 hz

Nominally 20,000Hz is the upper-limit for human hearing,
& for audio electronics designed for humans.

Only a minority of society can hear 20kHz: youngsters.
but even then they would not understand the frequency-shifted audio as it is effectively being scrambled.

1 Like

So if resampled my project to 192000 hz sample rate before encoding the audio with nyquist at a carrier of 17500 hz, what exactly would happen? Since we don’t have hardware that can reproduce 100,000 hz we’d basically be left with what we can hear up to if i’m understanding this correctly. What would be the perfect sample rate for a silent subliminal?
Just the standard 44100 hz? Thanks for the help by the way, I’m a novice at this stuff.

The plug-in from Steganography and “Silent Subliminals” | AudioNyq is optimised for a sample rate of 44100Hz or greater.
A higher sample rate (say 48000 Hz) may provide a marginal improvement to the quality, but raising the sample rate further has diminishing returns. A sample rate higher than 96kHz is just a waste of disk space.

As the author of that plug-in, I’d recommend using 44.1kHz or 48kHz sample rate.

1 Like

Ah thank you Steve!, I was thinking of using edgar rft’s original code from 2011 since I was getting confused on the equation he listed in this post.
Is this the correct equation to work out the total bandwidth of the audio?
So let’s say I used 44100 hz as recommended, it would just be a total of 19050 hz bandwidth?

Covertly inserting messages into people’s minds via “Silent subliminals” is a bogus concept.
Typically what they hear is nothing, i.e. silence,
or in a minority of cases incomprehensible tweeting.

No message is being communicated. It’s a scam.

I don’t know the details of edgar rtf’s 2011 code off the top of my head.

Technically, the issues are:

  1. Typically, human hearing is insensitive to frequencies above about 16 kHz. A few individuals (mostly children) can detect sounds up to 20 kHz or even a little higher. The upper limit tends to decline with age. This is a physiological limitation rather than psychological. Your subconscious can’t do anything with sounds that your ears can’t convert to synaptic messages because there is no way for the subconscious to even be aware of the existence of such frequencies.

  2. Digital audio has limited frequency bandwidth based on the sample rate. The theoretically maimum upper frequency limit is half the sample rate. In practice the limit is a little lower than half the sample rate.

  3. When applying amplitude modulation (the “silent subliminal encoding”), the bandwidth of the encoded signal must be limited so that it does not exceed the “Digital audio bandwidth”.

  4. The upper frequency of the encoded audio can be calculated as “carrier frequency + maximum input frequency”. For example, if the maximum input frequency is 4kHz, and the carrier frequency is 16 kHz, then the maximum encoded frequency is 16 + 4 = 20kHz, which is just under half of 44.1 kHz.

  5. Human speech has frequencies above 8kHz, but for a sample rate of 44.1 kHz and a carrier frequency of 16kHz, the maximum encoded frequency works out as 8 + 16 = 24kHz, but that is too high for a sample rate of 44.1 kHz, and would cause a “bad thing” to happen. The “bad thing” is called “aliasing”.

1 Like

How would I be able to calculate the maximum input frequency for my speech file? And also I’m wondering why is 24khz too high for a sample rate of 44100 hz, is it because it’s above half of the samplerate?

Yes, exactly that.

Plot Spectrum will show you the frequencies present.

The plug-in in this forum topic filters the audio with a 3kHz low-pass filter before encoding, which effectively eliminates frequencies above 4kHz.