Any one know a good music database?

[takingover]

Hi,

All my songs are in a reall mess and I'm tring to edit each of them but wondered if any one knew of a good online database where you can search for a song and it will tell you what year it was made, what type of song it is and other random stuff.

I've already done all 293 albums but it's just the random indivdual songs left to do.

Jay

[mzemina]

If you could define "...and other random stuff." a developer or someone who is well versed in SQL code plus XML (would be needed to create a new screen) could then take this on as a project. Something which woul dhelp us, would be for you to create a dummy screen image(s) to help us understand what you would like the screen to look like. You don't need to know photoshop, even Microsoft Paint will work to help in getting your idea across.

Mike

[takingover]

It's nothing like that, I don't want anything developed. I'm just sorting out my music files so they display correctly in My Music. I'm looking for a website that I can search for a song, and find out when it was made, what catagory it is. So that MP displays all the songs correctly at the moment i'm using amazon but it's not always got all the songs.

Thanks

Jay

[jawbroken]

Any tagging program that supports CDDB access should be able to do things like this, I believe.

[Smirnuff]

Give musicbrainz.org a try, last.fm may be helpful too.

[takingover]

Thanks Smirnuff that's what I was looking for.


Jay

[knutinh]

Logically, CDDB + ID3-tag + filname/path should mean that a program could fix tags as well as humans.

However, the ones that I have tried (quite a few) dont.

Reasons:
-Bad and confusing UI
-Slow, buggy implementation
-They can work along "one axis", but I still havent seen anyone doing clever combination of all available info (filename/path, existing ID3, FREEDB/CDDB and user-supplied naming convention)

Also, there are incredibly many ways to present the data. I think that the user should be able to leave the PC while it is processing. When returning, you should get a sorted list with the worst hits first. IE, a file named c:\music\singles\unknown.mp3 with no tags should be on the top of the list, as there is no way other than analysing the actual music to identify it. When the user clicks "play" and recognize it as "yesterday", he/she should write that info into a textbox and be presented with suggested name/group combos. If that wasnt enough, it could be necessary to write "Beatles" if the user knows what band etc.

If a non-tagged file is named 12.mp3 inside a folder of 14 recognised songs from the same album, the program should suggest that this song is perhaps from the same album.

regards
Knut

[knutinh]

IF ID3-tag contains all necessary fields
fill fields into temporary ID structure
flag as recognised
ELSEIF ID3-tag contains one or more necessary fields
fill recognised fields into temporary ID structure
set the rest as "unknown"
END

IF filename follows user-supplied convention (for instance "x - x - x.mp3")
fill recognised fields into temporary ID structure
Else
keep filename and folder for later parsing
END

IF there are less than 30 music files in this folder
flag all songs as possible album
check if other files have been recognised earlier
END

In the end, you may have a "matrix" containing #Artist, #Song, #Album etc where each field may have zero, one or several (potentially different) suggestions. Like:

#Artist : "Madonna" OR "madonna" OR "80s hits VA"
#Title : "unknown"
#Album : "True Blue" OR "Greatest hits of the 80s"
#filename : "trueblue"
#path : "madonna"

Do an elimination for fields with multiple hits. Some text parsing should be able to calculate the textual distance between "Madonna", "madonna" and "80s hits VA" and recognise that madonna/Madonna are most similar and should be trusted?

The most trusted sources of #Artist (in this case) can be used to give "True Blue" somewhat higher score than "Greatest hits of the 80s".

Do a FREEDB/CDDB search for "Artist=madonna, Album=True Blue", and get a list of perhaps 12 songs in that album.

Do a search between fields like #path and #filename matched against song titles from FREEDB.

Hopefully, get a correct match.

Store some "trustworthyness" parameter inside the struct, alongside the 2nd best and 3rd best hits etc.

After done for every song, sort after trustworthyness and present the user with the worst hits first, and a dropdown menu of alternate tags.

-k

[JulianBuss]

try this tool: http://www.tagrunner.com/website/v2/en/home.php

I did a short test two days ago and was quite satisfied.

[Zheele]

Also. Do not forget the length of the song as a search/matching parameter. Most programs I've used to tag stuff just messes everything up so I have resorted to hand-tagging all my music and that is a royal pain.

If you can nail a function logic that eliminates most of the user input and stop to ask for user input only if the tagging validity falls outside the predefined set of parameters it might be a great thing.

Some kind of "smartness" to make the program realize that this is something I can't tag and I need help of the user is something that is blatantly missing in many automation programs.

Also, preview before accepting tagging results would be wonderful.