The last.fm playlist generator is a nice place to start but it's fundamentally flawed - playing the tracks suggested by last.fm will lead to those plays being scrobbled creating a feedback loop that reduces the chances of finding anything else in your collection (likewise, obscure artists will never get plays) and it pays no attention to what the song is actually like so the odds of jarring transitions go up the more diverse the artists in your collection are.
The solution to these problems is to use a playlist generator that analyses the music and makes suggestions based on that data. I can't even pretend to understand the maths but applications like MusicMagicDJ (Windows, now un-maintained but still functional even though it uses an external database of attributes) and bpmdj (Linux, standalone) use this approach and work particularly well. bpmdj is a bit of a nightmare to use so gjay attempts to expose only the playlist functionality (works to a limited extent).
I'm current running gjay over my collection - it's taking a very long time and so far the suggestions are a bit mediocre (it has a tendency to be lazy and just pull more tracks off the album of the seed song). Plus the colour system requires more input than I really want to give (there is no way I could assign colour data for my entire collection).
Perhaps the best example of what I think guayadeque could really use is the way Banshee integrates Mirage (http://hop.at/mirage/) - adding the ability to intelligently populate playlists with the criteria and selection system that guayadeque already has would be fantastic.
I guess an alternative would be to extract the Mirage functionality into a CLI app and then feed playlists back into guayadeque but that would lose the filtering that guayadeque does on previous tracks/deny list/etc.
1 post • Page 1 of 1