sCrAmBlEd?HaCkZ! is a program that takes your voice and finds the closest matches to the sounds you're making in a database of snippets from music videos, then plays them back. The creator describes this as "vocally describing" the sounds you'd like to hear to a computer, which can then find close matches and play them back. There's a video-component, too, as the video bits that accompany the samples from the music videos the computer selects are played back for you. The author has promised to release his code under the GPL free software license as soon as he can "find time to clean up that mess / comment the code / document it and find a way to make it easily installable." In the meantime, the videos of the code running are highly amusing and inspiring:
What would it mean if a mind music machine existed which would render it unnecessary to write, play or sequence music by allowing one to just think music to make it happen. What would it mean if such a machine would construct imagined music out of samples of ones digital music library, or even out of the vast amount of music found on the internet? Of course it would be used – for no other reason than its existence.
It would have discursive power just like P2P has discursive power. The very fewest P2P users share data because they feel that information wants to be free. They share just because they can. But that also causes a notion to be formed on what unlimited copyability and digitality actually means. A clear case of massage.
As there is no appropriate technology to read minds to transform thoughts into music the human voice supercedes the thoughts as it is the original instrument – the most direct way to express imagined music.
Even if s?H! in the existing form is far from being able to really recognize the desired music expressed by a vocal description it is working well enough to let, with some phantasy and my suggestion, the hypothetical mind music machine become vivid.