pubs
A music search engine. Literally, in the same sense that we have text search engines. The user would be presented with a flash keyboard and attempt to input a tune he’s got rattling around in his head. He’d be allowed as many takes as he needed; a sign would advise to wait four seconds then try again, between takes. At the end, the user would select which take is closest and submit it. (There would of course be an optional advanced user interface, for people who know something about music theory.) The performance would be sent in MIDI format, and interpreted. The site would then search its database of midi transcriptions for approximate matches. It would output titles, composer names, years, and albums (if applicable). The user would click on the responses to hear the MIDI files (or perhaps associated digital samples). If the user finds what he’s looking for, there would be unobtrusive yet obvious links to iTunes or Amazon or whatever.
A “name that tune” engine. It’s completely feasible, in theory. The largest hurdle that I can see would be the RIAA.
Somebody else had this idea, actually: http://www.songtapper.com/
Well, to clarify, not quite the same idea. But the same sort of direction. You take it further.
Sorry to keep it coming like this, but after I dredged up the first link on del.icio.us I also stumbled across this: http://www.midomi.com/
Uh… fuck, I remember seeing something like this. Can’t remember the URL or anything though.
Rest assured it did not work very well though.
Yeah, it would be a little simpler than the various picture-search systems that people are working on right now, it would still be more difficult that your standard text search because there’s an interpretation process going on even on the most basic level, with the engine having to account for pitch, timing and possibly other aspects (depending on the complexity of the search engine).
It was attempted during the last internet bubble. The idea was you hum the song into a microphone and upload the wav file. I don’t know off hand what the name of it was or if it ever saw light of day.
Yeah, this is definitely feasible. I’m surprised an organization like ASCAP hasn’t done this already. (Or maybe they have, but only for internal use.)
neat idea
the RIAA would be a hurdle in the search for the cure for cancer if it could.
Current text search engines often also account for a certain amount of error — though yeah, the factors are somewhat more elaborate with music. Still, I can think offhand of a few simple ways to build in a margin of error. I’m sure someone schooled in this stuff could devise something far more sophisticated.
Yeah, this is… along the right lines, though working with just rhythm has its obvious limitations. Still, the most distinctive songs that come to mind are all pretty recognizable by rhythm alone.
Could it be that Midomi thing?
Wait, this was Song Tapper that I was referring to.
Can’t test Midomi well, as I don’t have a microphone.
Or, rather, Song Tapper?
Actually, no. As I recall it let you enter notes on a musical staff…