Sven König wrote software that will take a piece of music/video and automatically divide it into 16th beat samples. It will then listen to another sound-source (such as your voice), and find the best match for each sample of your voice in the original group of samples. It does this real-time, and interesting performance art can be generated with it.

To be honest, I'm not thoroughly impressed with the actual songs/videos it produces, but it's a great concept and the raw technology can no doubt be be improved and used by other technologists and artists to interesting effect. The site is a bit daunting. You may want to check out this video for a quicker understanding.

Link