Hi Quentin,

have you considered generating "smart" playlist based on similarities found in music files? When having
a large collection of music labeling each song becomes unfeasible and time-demanding.
There are many algorithms for classification of instruments, measuring mood, and genre
similarity etc. which could automatically put a label for each song or help with filling up queue.

Hi deric,
if you're using the artistinfo-plugin (or at least have it installed and enabled) you have that functionality. Simply go the the queue and choose "similar artists" from the available queue-modes.
Be aware though that it is limited by last.fm's definition/usage of artist-similarity.

You can easily tweak the plugin and its settings by looking at the "similar artists"-list inside the artistinfo-plugin and changing the similarity-limits in the plugin's settings.

Let me know how it works for you!

#2 November 21, 2011, 17:36:53 Last Edit: November 21, 2011, 17:39:05 by Satoshi
There used to be an online music service (akin to last.fm) that would classify tons of music and show similarities with other music. It was really good.

EDIT: Oh, here it is: http://www.pandora.com/mgp.shtml

Seems it still exists, but only for the US.

There is different way to do it, the simpler way, do the comparison by a website, is already done in ochosi's artistinfo plugin using last.fm, of course there may be other website, using different way to measure similarity.
Other methods require adding some information in the tag, that could then be read by gmb (that part should be easy) and used to generate playlist/fill the queue/used in a random rule (that part may be hard)

You may be interested by trashdal's albuminfo plugin that can import allmusic.com's genres/styles/moods, though it is currently limited to the current album, and of course doesn't distinguish between different songs from the album as allmusic.com only attributes these to an album, not to individual tracks.

I must say I'm not myself particularly interested by generating playlists based on similarity, I don't mind going from one genre to another. And I think, though I admit I haven't experimented much, that these similarities measure are very far from perfect.
So I probably won't add this by myself for now. But if someone has a specific method in mind, I could tell how difficult it would be to add, and maybe help a volunteer do it.