using audio

I've been thinking for a while about using audio as part of the navigation experience. Lots of CD players have an intro mode that plays a few seconds from the beginning of each track on a CD but, almost by definition, this doesn't give a good indication of the flavour of the whole track. Think Bill Evan's piano etude intro to So What or the multiple sections of Bohemian Rhapsody.

So what could be done? How about a representative section of the track? This could either be manually created for a particular track or algorithmically determined. This paper describes how to determine the segment of audio that best represents the whole work by calculating some kind of similarity measure.

So imagine selecting a track on your iPod and, if you leave it highlighted for a few seconds, a summary of the track fades in and out. And we could take this further and get similar representations of a whole album or a playlist or even an entire music collection. And the processing could be done offline on the desktop, so avoiding costly processing on the iPod. It would be like an automatic mixtape/cut-up. See DJ Food's "Raiding the 20th Century" for an excellent example of one, in this case representing the entire history of cut-up music. Obviously the research, compilation and mixing of this was a highly skilled process but we don't have to aim for this quality.