I'm pretty sure that what I'm trying to do is impossible but I figured I'd try anyway. I'm making a MIDI file using tuxguitar I'll later export into Logic for refining, and there's rap/spoken vocals in this song. In theory the human voice is made of tones and pitches that fluctuate between each other too fast to call it a musical note...in my theory anyway. So I'd think some degree of replication of the tone could be achieved by various sliding notes, but I have no idea of where to even start, what tone to begin, end, slide to, etc.

Here's an mp3 of a quick humming of the kind of thing I want to accomplish in MIDI.

If it's not impossible to do with a virtual instrument or MIDI, could you show me where to start, or point me to a tutorial or something?



4 Answers 4


The easiest solution would be to just use apple's say text-to-speech program. Try $ say hello on the command line.

Any sound can be broken down into a series of sine waves. The easy way to do that is via the Fourier Transformation. Then you can re-synthesizer with an inverse-fourier tranformation, which essentially means playing back the sine waves with the proper coefficients.

If you want to go really crazy, you can resynthesize a sound with something other than a sine wave, like a piano:


Unfortunately your theory of how the human voice is made up is not accurate enough to make this possible - you just can't get the detail.

However, there are a wide range of examples of partial emulation of vocal tones, using combinations of effects and notes - see Steve Vai's guitar at the start of California Girls or Greasy Kid's Stuff for a couple. He uses the combination of a Wah (which allows you to create a lot of vowel sounds) and sliding guitar notes.

It still doesn't come across as speech, but if you listen carefully you can guess at the words he is trying to pick out.

  • See also formant synthesis and related techniques.
    – Warrior Bob
    Commented Jul 2, 2012 at 15:52

I've never tried it but Yamaha's Vocaloid technology was developed for this purpose. I think it's more designed for robots and anime characters, but maybe with the right processing in Logic it could work for what you need. Check out http://en.m.wikipedia.org/wiki/Vocaloid for more info. If you try this please report back results, I'd be curious to know how it goes

  • I'd be interested to know if it'd be possible to create a wavetable of the 44 phonemes btinternet.com/~ted.power/folkchart.htm in the english language and produce speech that way.
    – dwwilson66
    Commented Jul 3, 2012 at 15:29
  • 1
    I read somewhere that there was an iOS version, I can't find the info now. Just found this in the Japanease iTunes: itunes.apple.com/jp/app/ivocaloid-vy1/id396081265?mt=8 If anyone can find and post info about the American version (unless its imaginationware on my part) that would = awesome.
    – JPollock
    Commented Jul 3, 2012 at 20:41

Another possible solution is autorap: http://itunes.apple.com/us/app/autorap/id524299475&+Synthtopia And iOS and android app that you speak in to and then it rapifies it... I'm afraid to try this.

Your Answer

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