In our mission to improve music experience, we ‘ve been working for months on our Music recommendation API. Here is an example of the kind of audiobased-playlist you will be able to generate from a seed track.
Using the following track by Nicolas Jaar:
Here are the 10 first results we got :
We’ve been training our deep learning system on lots of information derived from audio features extracted directly from the signal.
We analysed the thirty-second samples from about 100,000 tracks of eletronic music without using any external metadata. The main purpose of this approach is to be able to recommend new music right when it is released.
We hope you will soon be able to test and enjoy our full-featured Music Recommendation API. If you have any suggestion, let us know by tweeting @nilandmusic. We’ll be quick to respond because we believe that ideas from the community are key to the service we’re creating.