Something like IMMS for XMMS. Measure the BPM and frequency (like in GJAY), and have recommendations mixed in with audioscrobbler\'s, or as an option opposed to such. Or just have a GJAY or IMMS-similiar adaptation built-in somehow, if that\'s even possible.

I wouldn\'t know the technicalities of it myself, but I just thought it would be a great feature to think about for developers.

Yes, we all think this would be the epitomy of party mode/auto append suggestions.

However, problems lie in calculating the bpm/frequency of a song.

Do we scan every track for bpm on a collection scan? No, this would be ridiculously expensive.

How about putting another constant load on the application whilst we are playing the track? Again, I don\'t like the idea.

If somebody can plan a devious scheme for grabbing bpm out of a song in a very efficient manner, I\'m sure we would integrate it.

Do we scan every track for bpm on a collection scan? No, this would be ridiculously expensive.

How about putting another constant load on the application whilst we are playing the track? Again,
I don\'t like the idea.

I see. Hm.

I wonder what it would take to have one central server process the BPMs of incoming songs, and have a database from which to call on the information. That would probably be more up Audioscrobbler\'s alley, though. Or at least, a service similiar to it and last.fm but with a different philosophy of music recommendation. Not that I have any idea what I\'m talking about.

Have I mentioned that I love amaroK? I love amaroK. Thank you and kudos to you developers.

You could make it a separate background process--low priority. It wouldn\'t slow down the collection scanning, but would be put into the database when done.

As I remember, IMMS had two ways of doing it. You could call a separate program to analyse tracks (as specified on the command line). The results were stored in the SQLite database. This meant the user called the system load when it was acceptable. Otherwise, IMMS scanned the track as it was played. In either case, the track only needed to be done once. I have to say, I never noticed much slowdown when using IMMS on the fly.

Would this be a good way of approaching it? A menu option (with appropriate options) for whole collection scans (perhaps with a mathcing function to skip done songs), and on the fly otherwise?

Just my (small) input

I got another idea.

MusicBrainz tagging is partially based on TRM ID, or acoustic fingerprinting/audio signature. It measures \"...the duration of the audio file, the frequency distribution, \"beat\" detection, and so forth.\"

So the data has already been collected into a universal database, and one that amarok already uses. One possible hurdle would be somehow isolating the data that would be useful for suggestions, such as discounting the durations of audio files. Otherwise, the songs listed in the MusicBrainz Results window could actually make useful suggestions in a seperate suggestion feature.