WOWHack, Last.fm Recommendations and Music Identification

Right now I'm in Gotherburg, Sweden, at WOWHack, a hackathon organized by EMI, Spotify and the Way out West Festival. I'm here with music hacker Jens Nockert and are working on one of the most requested features since the beginning of the Youtify project: Last.fm scrobbling and recommendations.

Jens Nockert and Per Thulin doing the live demo - photo by possan

1. To get started, connect your Last.fm account in Settings.

2. If you have scrobbled before, you should be seeing recommendations on the new home screen. Otherwise you have to wait until you have scrobbled a few tracks.

On artist pages you can discover similar artists:

When we scrobble tracks, we have to provide the correct song and artist names. This is tricky for us as YouTube and SoundCloud lacks a structured way of identifying tracks. That's why we go through a series of steps to identify the song being played:

• First we look at the track/video title to see if we can extract the artist and song names from there, e.g. "Robyn - Call your Girlfiend" is a common way to name tracks.
• If the initial check fails, we send the song link over to the Echo Nest API for an analyze of the waveform and identification of the song (does not work for YouTube videos).
• If we can't extract the artist/song names from the track title, and we can't get an answer from Echo Nest, we use the uploader name as the artist name and hope for the best.

This method is not bulletproof but works surprisingly well. However, a better way for identifying YouTube videos and SoundCloud tracks would be great for what we are doing.
That's all! Hope to release this baby soon!