Posts
I thought I'd talk a little bit today about song recommendation engines, as far as where we are currently with the technology online, and how it can get any better.
First of all, we've all heard about Pandora's Music Genome Project. They actually have a very dedicated staff that goes through each song for about 10-15 minutes and reports on a list of many different musical variables. The results get fed into the algorithm and BAM, there's your playlist.
Last.fm does something very similar, but they use different variables, and many sites, like thesixtyone.com, use a "similar" function that introduces you to music with similar variables.
So how do you tell which one works the best? You'd almost have to go through and look at the variables they use to tie music together. It would be almost impossible to tell if the site could have played you a better song than the one it did. Streaming music seems to be the way to go, but right now I believe the biggest factor in people's choices between these different websites may be the design appeal and ease of use. That, and the lack of anything better.
There are several problems I see with this picture. Number one is that it never seems like a good idea to use a tool that has no clear distinction between it's competitors. There's gotta be one that's better, but in this case it's too hard to tell, or would take too much effort. This most likely has a lot to do with the fact that these concepts are no new, and no one has really settled on one ultimate solution, yet people do have their favorites of the moment.
That brings me to the second problem with this picture, which is a fundamental one. The current song recommendation engines all use the song's variables to tie the songs together, and then tie you to the songs by entering a song or artist you like. This a pretty cool, but your control over your music ends after you enter your favorite artist or song.
Music is such a social thing. It seems to me that our playlists shouldn't be controlled by similarities between songs, but similarities between people.
There needs to be a system where I can follow people that I share a taste in music with(my friends, favorite band members, ect). Then anything in those people's playlists will get sent to my radio player, at random, or at my control. This not only ensures that you'll hear only the best music, but also it will automatically update you when new songs are out, and it doesn't bind you to one genre, or one sound.
If you're like me you could listen to 4 or 5 different genres, back to back. This system would also allow for filters on things like genres, moods, tags, ect, and could create a much more custom listening experience.
Also for new bands, this would almost take the place of promotion, because it is basically automated word of mouth and is the epitome of viral. With this model, who knows, you could be the one to discover a band for your whole generation.
I don't know about you, but I think that sounds a lot better than trusting variables and algorithms. This model will actually be out soon. It will be included in my website, Beat-Play, coming out in beta this April, 2010. It will be undergoing many changes early on in the beta process, but we hope to get it all fleshed out by June.
When it comes to the internet and all of the crazy, complex, and really cool tools out there, it's best to keep this thought in the back of your mind: "Is this the experience I want?" If the answer is "I don't know", then there's usually a problem somewhere, and also a void waiting to be filled.
For more info about the Beat-Play beta check out the BeatPlay Beta Overview
And to sign up to beta test, visit: http://MusicWithoutLabels.com
Written by: Dante Cullari Founder & President, Beat-Play, LLC
First of all, we've all heard about Pandora's Music Genome Project. They actually have a very dedicated staff that goes through each song for about 10-15 minutes and reports on a list of many different musical variables. The results get fed into the algorithm and BAM, there's your playlist.
Last.fm does something very similar, but they use different variables, and many sites, like thesixtyone.com, use a "similar" function that introduces you to music with similar variables.
So how do you tell which one works the best? You'd almost have to go through and look at the variables they use to tie music together. It would be almost impossible to tell if the site could have played you a better song than the one it did. Streaming music seems to be the way to go, but right now I believe the biggest factor in people's choices between these different websites may be the design appeal and ease of use. That, and the lack of anything better.
There are several problems I see with this picture. Number one is that it never seems like a good idea to use a tool that has no clear distinction between it's competitors. There's gotta be one that's better, but in this case it's too hard to tell, or would take too much effort. This most likely has a lot to do with the fact that these concepts are no new, and no one has really settled on one ultimate solution, yet people do have their favorites of the moment.
That brings me to the second problem with this picture, which is a fundamental one. The current song recommendation engines all use the song's variables to tie the songs together, and then tie you to the songs by entering a song or artist you like. This a pretty cool, but your control over your music ends after you enter your favorite artist or song.
Music is such a social thing. It seems to me that our playlists shouldn't be controlled by similarities between songs, but similarities between people.
There needs to be a system where I can follow people that I share a taste in music with(my friends, favorite band members, ect). Then anything in those people's playlists will get sent to my radio player, at random, or at my control. This not only ensures that you'll hear only the best music, but also it will automatically update you when new songs are out, and it doesn't bind you to one genre, or one sound.
If you're like me you could listen to 4 or 5 different genres, back to back. This system would also allow for filters on things like genres, moods, tags, ect, and could create a much more custom listening experience.
Also for new bands, this would almost take the place of promotion, because it is basically automated word of mouth and is the epitome of viral. With this model, who knows, you could be the one to discover a band for your whole generation.
I don't know about you, but I think that sounds a lot better than trusting variables and algorithms. This model will actually be out soon. It will be included in my website, Beat-Play, coming out in beta this April, 2010. It will be undergoing many changes early on in the beta process, but we hope to get it all fleshed out by June.
When it comes to the internet and all of the crazy, complex, and really cool tools out there, it's best to keep this thought in the back of your mind: "Is this the experience I want?" If the answer is "I don't know", then there's usually a problem somewhere, and also a void waiting to be filled.
For more info about the Beat-Play beta check out the BeatPlay Beta Overview
And to sign up to beta test, visit: http://MusicWithoutLabels.com
Written by: Dante Cullari Founder & President, Beat-Play, LLC
Comments (0)
Post a Comment