UIST'09: Relaxed selection techniques for querying time-series graphs

Imagine changing someone's dialect in a sound editor. I have a vision of editing sound in the same powerful way we can edit photos. This research tips the scales a little bit in favour of sound editing. From a song track, you would be able to trace the shape of a vowel, and select all similar vowels across the track. Change every a to e, and you get a weird dialect. Funny.

We're not there yet. But this is one step in the right direction.