MEAPsoft will analyze a sound file, break it up into pieces characterized by their acoustic features, and stitch the pieces back together based on a number of different algorithms. Listen to this sample for example of the silliness possible (I won’t spoil it by telling you what it is).
The interface is a little psychedelic. Run it a few times and you’ll see. The pic above is from the concept description of the “nearest neighbour” algorithm. The top bar represents the analyzed sound file, and the bottom bar is how the segments have been arranged to put similar ones next to each other.
The Music Engineering Art Projects (MEAP) group at Columbia
University is pleased to announce the 1.0 release of MEAPsoft:
http://labrosa.ee.columbia.edu/meapsoft
MEAPsoft is a program for automatically segmenting and rearranging music audio recordings. It is open source (GPL) and runs on all major operating systems.
MEAPsoft works by (a) segmenting original audio up into individual beats or events, (b) calculating some features for each segment, and (c) matching or rearranging segments from one or more pieces to create a new piece of music.
MEAPsoft comes with a wide and increasing range of feature extraction routines, and algorithms for matching and rearranging the segments (called “composers”). MEAPsoft is written in Java, and makes it pretty easy to write your own feature extractors and composers.
The links between the different parts of the program are via simple text files, describing the start times and durations of each segment (for the segmenter), appending a set of feature values to each line (for feature extractors), and including a target time for resynthesis
(as determined by the composer).
For examples of what MEAPsoft can do, please see our music showcase at:
http://labrosa.ee.columbia.edu/meapsoft/showcase.php
***
MEAPsoft was created by the participants in the Music Engineering Art Project at Columbia University.
Development by:
Ron Weiss, Douglas Repetto, Mike Mandel, Dan Ellis, Victor Adan, Jeff Snyder
Additional contributions by:
John Arroyo, Johanna Devaney, Dan Iglesia, Graham Poliner
Sponsors
* The Columbia Academic Quality Fund
* LabROSA
* The Computer Music Center
