I scraped a couple of webpages, then in java converted every letter into a number zero to nine dependent on how far along the alphabet it is. I then turned these into pairs of digits.

I imported this raw into audacity, as unsigned eight bit stereo signal at a 4410 sample rate. I repeated this as I didn't have time to go through and scrape more sites right now.

The result sounds like a sine wave, with some noise.


  • 1
    Hi Evanson, this is borderline 'off-topic' for a sound design site, perhaps I'm mistaking. Can you explain what makes this on topic? Perhaps add why you are creating the sound, which context it will be used? – Arnoud Traa Sep 6 '14 at 9:38
  • 3
    While I'd find this very exciting if it were true, I seriously doubt it's not just some artifact of the technique you used. Could you properly specify how you did it (for source code, link to GitHub Gist)? When I try to reproduce this I merely get noise, nothing sine-like. – leftaroundabout Sep 6 '14 at 15:36
  • @ArnoudTraa Why would timbre dynamics be off-topic in a sound design site? That makes no sense. This question is as on-topic as it gets. – Tom Cat Sep 6 '14 at 17:00
  • 1
    No worries JCPedroza, I'm definitely not saying timbre dynamics are off-topic. I was merely asking for a clarification because the question was too vague to indicate any relation to sound design (and therefor also timbre dynamics). Constructive criticism is needed if questions are not clear. – Arnoud Traa Sep 6 '14 at 17:31
  • @ArnoudTraa The question's scope is clear, regardless of the OP's intentions. If anything it needs detail in the procedures in order to obtain any meaningful answer, but it is not vague at all. – Tom Cat Sep 7 '14 at 9:23

Maybe that is because linguistics and/or semantics are cleverly designed. At least that is an interesting way to look at those. I'd be curious to see what the results are with different language (I guess you analyzed english, no?)

| improve this answer | |

Guessing severely because you gave too little information in your question, I think you are hearing the letter E.

As shown in http://en.wikipedia.org/wiki/Letter_frequency, E is the most common letter in common English by a wide margin:

Relative frequencies ordered by frequency

If you look at this chart as if it were amplitudes of tone generators at various frequencies, it is clear that "E" dominates the spectrum. If this were MPEG (or other psycho-acoustic compression) encoded in the chain, it would likely toss away harmonics below some threshold; for example, D–Z might not be carried through compression. Higher amplitude harmonics might be generating the noise.

Apart from the fundamental, there is a spike at 1Hz which is probably an abrupt transition to zero, I'd guess that that may be the point of repetition of your data.

I'd have to guess much less if you'd made the SoundCloud track downloadable, and as others have noted above, could eliminate the guessing if you gave the generating code.

| improve this answer | |
  • 1
    It understood it that the OP used the letters to generate audio samples, rather than frequencies. But that's just guessing on my part... – leftaroundabout Sep 7 '14 at 19:59
  • I agree, however, the letters would generate fundamental frequencies with amplitudes (roughly) corresponding to the chart. It may help if you think of the bars as drawbars on an organ or the whole chart as being an FFT analysis of the resultant signal. I'm just guessing too. – msw Sep 7 '14 at 20:09
  • 1
    Nah, that wouldn't really work. Those letter frequencies aren't periodic frequencies, just statistical frequencies, so each letter would rather generate a sort of wide band-filtered noise than sine oscillations. – leftaroundabout Sep 7 '14 at 20:12

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