Mufin: An Automated Music Recommendation Engine That Actually Works

Finding songs that sound the same isn't really enough. It might make do for superfluous listening, but if you have more than a passing interest in music your requirements will be different.

Emotions and lyrics are much more important than the "sound" when I choose new music, and it will be a while until a computer program can categorize these. Remember that it isn't enough to just produce something that matches a pattern. This is a creative field we're talking about, music and movie tastes contain more human factors than just about any other problem tackled by computer scientists.

As a case in point, Ayreon may do an excellent job of copying the darkly brooding and reflecting atmosphere of certain prog rock bands...but it's still a copy, and just under the surface it stinks of hypocrisy.

Media recommendations is a huge problem (it bothers me on a personal level every day, I would pay at least 200$ a year for a service that works), but nothing I have seen so far has impressed me. I won't be sold until I get at least two recommendations that made me cry.

Some of the recommendations were on target, but there were about half that were far off. I like being able to browse through the similar tracks and it's cool that it keeps the history of them so you can go back.

Do any of the music recommendation places have an api? I've been thinking about creating a site that recommends music for groups of people that might have different tastes in music. Not many of my friends like the same stuff, but I know there is some overlap there.

Try out http://mydj.musicip.com or the MusicIP Mixer, http://musicip.com/mixer/index.jsp

This technology combines acoustic analysis and comparison with aggregated acoustic profile info and a really deep metadata and social data database for interesting and TUNABLE mixes and discovery.

Had mixed results with this, was getting Elton John tracks on songs that sound nothing like Elton John (and yes I listened to the Elton John track to compare)