So Celemony finally released their DNA technology to a wider audience (beta version for existing customers). It took a little longer than originally announced, but who cares for such great technology. Congrats to the Celemony team!
Just found the time to actually sit through their demo videos.
Having been active in the music tech research community for several years now, and knowing what are the possible problems with polyphonic music analysis, I was particularly interested in how they are dealing with the inevitable spurious misinterpretations of harmonics by the note detection algorithm. Have a look for yourself (part 3, start around halfway the video):
http://www.celemony.com/cms/index.php?id=editor_betavideo#
I think they did it the right way: they still allow the user to correct decisions made by the algorithm, but also provide some help in the process of doing so by using this "hollow notes" concept. A good example of where it's not purely about the algorithms alone (which in the case of automatic music analysis will never be perfect, forget about the holy grail), but about the thin balance between hiding the complexity to a user, and still allowing him enough access to details to correct imperfections of the technology.
On another note, I'm also curious to see how many users will eventually prefer tweaking their audio at this level, rather than just re-record the part when a mistake was found (if still possible of course). Or will it find more use as a creative tool, rather than as a correction tool? (*) In any case, nice technology and application!
(*) A lot more interesting thoughts and opinions about that at Peter Kirn's CDM blog, by the way.