Is there a plugin that fetches "music genome project" data from Pandora or other sites (I don't know whether other sites use it) and adds it to a tag? This would be great for a function like "play similar" or something like that...

Well... i don't think that the genome database is available from outside pandora... what you can do is use the last fm plugin buried somewhere in the forums to create playlists with similar artist from an artist.

I don't like Last.fm's "similar artist" function, it's not nearly the same. :-\

i smell similar artist foo plugin...
can that be done?

I agree. But as I said. I don't think there is any API for the Music Genome Project...
So no luck there, and besides last.fm I don't know any similar sites.

does foosic have any of these features? i am actually sorry to post this without reading into it myself, i just heard about it... but just wondering if there can be a feature implimented within foosic?

You can use Music IP Mixer and use the "send to" function setting Foobar2000 as default player.

http://musicip.com/mixer/index.jsp

YS,
Rozzo

I agree that we need this functionality today. Many of todays technologies could help build a database with this. There are a few free/open initiaives that comes close to this.

One of the first I tried was MoodLogic. It generates a playlist based on your mood. It allows anyone to submit characteristics from songs to it's database. However I personally don't think it's music fingerprinting is very good and many data are incorrect. Also it's userbase is too small proberbly because it's not free and not open software (you need to use their software to generate a playlist).

Winamp has a similar feature, but I think it's relying on a pure technical analysis of a track, which was fairly reliable, but still often lead to misplaced tracks.

MusicBrainz offers a service using fingerprints to match tracks with their database of metadata. However my personal experience was that it wasn't able to match many of my tracks, so I'm assuming they are not using real fingerprints.

Gracenote (and Sony) has a very nice fingerprint service called TrackID. With many SE Walkman cellphones you are able to record a small sample, which will be matched against the Gracenote database. I know that Philips a few years back invented a similar music recognizion system, but have no idea if it's available to the public.

I think that we need a powerful database with detailed informations that users could provide similar to the analysis that takes place with the Music Genome Project and MoodLogic, and detailed information on every song/album like AMG, Discogs, Wikipedia provide. Such a database should be Wiki-like to avoid misinformation, and it could store a powerful fingerprint system similar to Gracecnote TrackID to avoid duplicate information on a song.

The main problem with such a system is that it need attention from the public to become reality. I know that foosic has some kind of fingerprint system, but I have no details on it, and as far as I'm concerned, foosic has been abandoned. It would be nice if anyone has additional information or ideas for such a system. Also I would like to know more about good and open music fingerprinting if anyone knows about that.

Sidenote: There's a guide to Electronic Dance Music, which categorizes most (if not all) dance-genres in a graphical interface. It would be nice if such a music-map could be generated out of such proposed database.

http://en.wikipedia.org/wiki/List_of_Music...ject_attributes

Interesting.
One could set up a website or plugin for popular players so that people could tag their music using this collection of tags (unless there is a legal penalty to pay for doing that ).

If popular, that could duplicate the music genome from Pandora, and release it in an open-source way, so music players could tag with it and one could search one's music in a nifty way.

Thoughts?

If that could be set up, I think it'd be an amazingly useful and powerful resource. Only problem would be standardizing everything and keeping everything regulated. The problem I see with having a wiki-based source for these tags is that, while being open for improvements, it's also open to abuse. Not that I'm sure many people would want to abuse it, but the wider spread it becomes, the more idiots we find who want to mess things up. I think the Music Genome Project's (mgp) way of having musicians analyze things is reliable, but time consuming and costly.
Anyway: I'm all for this idea, if it could be set up reliably, and indeed, someone should find a way of using mgp's tags if at all legally / practically possible.

Yeah. I think pingue said it all.
A wiki would mean problems (I mean, just look at artist pages at wikipedia, they change genre every day...)
A database like GMP is a great solution, but would require time and money... and The GMP is closed ATM (but they have said that they are going to open up an API if they find a way to get revenues without people having to go to their page and look at the ads, but that that API will contain, noone knows).

Certainly these are good ideas to throw around, and I am all for it... however I am a believer in the collective good/wiki proposal because after a while, with enough people working on it... things get regulated properly. The genre may change everyday on wikipedia but there are enough people making sure that there isnt any true abuse.

But I do understand your complaint about wiki style organization.. which makes me wonder what are the alternatives? This should be open source and free, and should not cost anyone money but I feel succeptible to the fact that someone will end up having to pay...

Why is music genome project and freedb not able to be utilized in a way that can benefit the entire community? I personally dont know, but something tells me that the people who built it, think they can make money off it, or at least protect the database by keeping it under control - these are both bad things and I hope that all of us who continue the project ideas do not fall to those levels of thinking, - we need a free and reliable way to make this happen.

my2$

That's the biggest nonsense I've read today (and everybody knows that the internet is full of nonsense).

Firstly, while both the server software and the database content of freedb are available for download, freedb is completely irrelevant for this discussion due to its nature.

Secondly, Pandora Media, Inc. invested money into the Music Genome Project (MGP). They pay for the servers, they - as far as I know - have to pay license fees for the content, and they pay trained professionals to analyze music. All this adds up to a not-so-insignificant amount of money. Why should they give away the data from the MGP just like that? It's their most valuable business asset. And why shouldn't they try to make some profit with that? Trying to earn money in a completely legal way is not a bad thing. Not even on the internet. It would be quite refreshing for a change to see a company that does not pour money down the drain while waiting to be bought by a big player like Google.

i dont know enough about this but i just think that there are so many music databases around that it should be easier to mine the collections to get a new one together that can do what the original poster wanted, if this still makes no sense or is off base, ignore it and i have nothing further to add to the topic except my personal opinion of a free and non commercial database.

Yes there are many music databases around - however many of them do not have the quality of data that MGP has. Think of it as the difference between Wikipedia and Encyclopedia Britanica. One is community maintained, and should never be relied upon as a reference, the other is fact-checked by professionals.

Rather than build an entirely new engine - I'd suggest latching onto a database that is free, and open. If you got the guys at Last.FM on side, it'd give you access to a solid server and a large userbase.

word

You mean using last.fm tags? I think they are public domain, and there are even RSS feeds that you can use. Although, to tell you the truth, they are unmaintained; I read somewhere in the last.fm forums that there was noone in charger of the RSS feeds. And they certainly lack some features that'd be important in this context.

You can always mine the site by doing http queries; if you do it in a civilized way (not pestering the servers) you may well get a good bunch of tags that way.

However, I wonder how good that tagging is. Many people use "seen live" as a tag, which is so idiosyncratic that is useless for your library. I personally rarely tag stuff in last.fm. And the tags I have seen are pretty general, mostly genres stuff.

Worth some more thought though.

The point here would be to use a common set of tags. For that, I think Pandora's MGP 'd be hard to beat. Imagine having say a foobar2000 component can be called with a shortcut that presents you with the cloud of MGP tags for the currently playing song, and you can click the ones that apply. You can type them too, using tag completion. Pretty much like in del.icio.us but for music. Then those tags are uploaded to a server. Since we need to know that we are talking about the same song to upload tags, maybe we need fingerprinting. For that I guess foosic is the right way to go. (Is foosic really abandoned?).

Thinking about it, this component (which required dependencies: foosic) would connect to foosic and suggest what other people have tagged that song with.

Actually, maybe we could also upload song ratings. Sometimes they are too personal, but if we have say more than 10 users ratings for a song they must converge.


Anyone here interested in such a component?

it sounds good, perhaps the first thing to do is find out if foosic is abandoned and if so, try to restart the project under new management? then we can work on the database and functionality

What difference does it make? The foosic library libFooID is free to use. As I see it, the only way to do it right is creating a new database.

I disagree that wiki's are unreliable. It's possible that people will change them everyday, but you can track changes, and for example genres should be multiple, since a song often spread across different genres.

Until such database is estabilshed, I have hard to see what a component would accomplish. Also you need to consider the different user bases to get widespread use: Winamp, WMP, iTunes, MM etc...

The primary need are currently which details such a database needs and guides to how they are determined for users.