Making instruments with the human voice


The human voice is a pretty versatile instrument, and many skilled vocalists have been able to do convincing imitations of other sounds, and we've all heard "human beat box" artists work with a microphone to do great sounds.

That got me thinking, could we train a choir to work together to sound like anything, starting with violins, and perhaps even a piano or more?

The idea would be to get some vocalists to make lots of sounds, both pure tones and more complex ones, and break them apart with spectrum analysis.   Do the same for the target sound -- try to break it up into components that might be made by human vocal cords with appropriate spectrum analysis.

Then find a way to easily add the human sounds together to sound like the instrument.  Each singer might focus on one of the harmonics or other tonal qualities of the instrument.  Do it first in the computer, and then see if the people can do it together, without being distracted.  Then work on doing the attack and decay and other artifacts of the start and end of notes.

If it all worked, it would be a fun gag for a choir to suddenly sound like a piano or violin playing a popular piece.   Purer tones like a flute might be harder than complex tones.  Percussion is obviously possible though it might need some amplification.  Indeed, amplification to adjust the levels properly might help a lot but would be slightly more artificial than hearing this without any electronics.   Who knows, perhaps a choir could even sound like an orchestra playing the opening to Beethoven's 5th, something everybody knows well.


Contemporary composers have been doing basically this for about 30 years. There's a style of composition called "Spectralism", practiced mostly in France. The big name is Tristan Murail. Wikipedia's discussion is pretty good. (BTW, the comment-validating question flunked me when I used a capital T.)

Looks to be mostly in europe. Surprisingly the wikipedia article does not have links to samples of the results.
As for the comment question, it always works caseless in my tests, are you sure you did not have another typo perhaps?

Yeah. It works OK for me now. I looked at it carefully when I did it, but maybe I missed something.

Aren't we soon suppose to have ways for a computer to read our minds. There's been studies with monkeys that show monkeys moving a cursor to click on signs, simply by a computer that reads his brain signals. Also, I think NASA researchers developed a very simple way for a computer to read brain signals that apparently travel to our tongue. So here in a different way, the computer reads our mind.

Now imagine the above methods being used to compose music. We don't have to know how to play any instrument, but simply imagine the tone and rhythm, and the computer translates it into midi signals. Isn't it the way everyone will start creating their own music pieces?

It can be done. What the heck will be the long term result of it?

These techniques are extremely primitive and there are already much superior ways to get information out of the brain. Skilled musicians aren't really that aware of how their fingers or vocal chords etc. are making the notes, they just think the notes and out they come, more like walking and typing. (When you type a letter, do you think about how to move your finger to hit the T key or do you just think "t" and it happens.)

So we already have the ability to just think the music and have a computer read it.

Add new comment