Most of the programs that are equipped to get track data from the web have the ability to identify a particular "performance" from data deliberately placed in the track info (data) on a CD. For it to work someone has to insert the data in each track.

This program apparently doesn't have to have any "data." The program "listens" to a tune, and says "sounds like" another tune, compares at a detail level, and makes a list of duplicates based only on the audio signals. The second program "listens" to a tune, picks out "characteristic phrases" and makes a "sample" based on the most characteristic musical content.

Note the claim that it works with streaming audio where no pre-packaged data is transmitted.

For those numerous "can somebody identify this tune" type questions, the program could immediately pick out another tune (possibly named) that "sounds like" the one in question(?????) What's not known from the report is how "loosely" two music samples could be "the same" for the program to match them up. Maybe that's a next step yet to be handled.

John