Log in

No account? Create an account

Previous Entry | Next Entry

Creative paranoia

I suffer from a strange paranoia. I am constantly coming up with ideas for projects that I'd like to work on and release to the world. They are often ideas that I think a lot of other people would find useful or interesting. However, once I think one up, I am generally loathe to discuss it with anyone in a public forum for fear that they might steal my idea and implement it before I get the chance to, and then gather all the laud and honor for doing so.

This is stupid, and I know it. Everyone else has their own ideas and projects to fill their time, and aren't going to be so incredibly interested in mine that they will immediately latch onto them and "steal" them from me. And the ideas aren't really so great that anyone will actually get any laud and honor through their creation, so I don't actually have that much to lose. In fact, if someone else did actually implement them, I'd probably benefit from that much more, in that I'd be able to actually make use of the things without having to worry about maintaining and administrating them.

Moreover, I generally think up these ideas, let them stew in my brain for a few months, and then move on and rarely think of them except in a "man, I wish I could find the time to do X" sort of way, so they never actually get done and nobody benefits.

So, without further ado, I present to you some of my latest big ideas:

  • Music metadata database

  • It's relatively easy to find music based on standard metadata (artist, title, album, release year, etc.). That stuff is commonly collected in many locations. However, a friend of mine recently said, "The song The Rainbow Connection asks ``Why are there so many songs about rainbows, and what's on the other side?'' I can think of two songs about rainbows: that one, and Somewhere Over the Rainbow." That got me thinking: wouldn't it be handy to have a resource that would let me search for songs with the subject "rainbow". In the past, I have wanted to search for songs with the genre "electronica" which had digeridoos in. What would be useful for queries like this is a database of songs with any arbitrary metadata associated with them.

    For instance, for the Elton John song "Goodbye Yellow Brick Road," the key/value pairs in the database would have the usual ARTIST → "Elton John" and ALBUM → "Goodbye Yellow Brick Road", but also might contain things like REFERENCE → "The Wizard of Oz" and INSTRUMENT → "piano". Any song could have any key with any value, and it would have to rely on public review to maintain integrity, which is iffy.

    It would have to allow any user to add, remove, or change any key/value pair on any song to be truly useful, which would introduce complications (a malevolent user could delete the entire database, so backups would have to be made frequently). Constructing search queries would be tricky, too: users would have to learn a complicated syntax to perform advanced searches ("'YEAR' >= 1960 AND 'YEAR' <= 1969 AND any LIKE '*love*'" for love songs written in the 60s, for instance).

    Tools could be created to parse a user's music collection's ID3/info tags and submit that information to the archive in bulk (but users should, of course, be careful to verify that the information is correct first.) Plugins for popular music players could be created that would let a user submit information for whatever song is currently playing, so when they notice something interesting about a song, they can immediately let the archive know.

  • Public searchable archive of royalty-free, public-domain high-quality sound samples

  • The creative community on the Internet is very concerned about allowing people to be creative in a low-cost, mutually beneficial way. They encourage people to create and release works under minimally restrictive licenses, and to build upon others' work to create new, better things. However, when it comes to music and audio creation, there seems to be very little truly free content for potential creators to work with, and what does exist is scattered and of questionable quality.

    I propose the creation of a massive, centralised, digital archive of high-quality, public domain, royalty-free sound samples for musicians and sound engineers to use for whatever works they choose.

    I recently decided to attempt to learn to use some music creation software and see if I could put together music I was happy with. I downloaded a few packages, but the instrument samples I could find were very limited and poor quality. Searching for commercially-available collections of samples leads me to sites that sell CDs full of audio for several hundred dollars apiece. I just want to tinker a bit, not shell out hundreds upon hundreds of dollars before I even know if I can create something halfway decent!

    The archive itself would be relatively simple to create: a file upload system, a database to keep track of all the samples, a rating system so users can weed out poor-quality samples and recommend high-quality ones, maybe a forum to discuss various issues related to music and audio. The big difficulty with this is bandwidth management. A useful collection of audio will easily use many gigabytes of bandwidth each month. Offering small collections of audio samples for download via BitTorrent would help, as would finding several mirrors for the archive.

    The main stumbling block for both of these projects is less the creation and setup of the actual systems (I could throw either of them together in a week or two of my spare time) but the gathering of a critical mass of actual useful data to fill the archives. While the song-database would be very useful once it was well underway, it would be less so before it had enough songs to help anybody out, and while the library of samples would be fantastic, it would need a number of contributors willing to release their samples into the public domain to begin with. If anyone knows the best way to solve these problems, please let me know.


    ( 5 comments — Leave a comment )
    Mar. 5th, 2004 11:12 am (UTC)
    If you started the song database, I'd gladly fill in every scrap of obscure reference information I know. Also needed: a field separate from the usual references, I think, for when songs have referred directly to other songs. Sweet Home Alabama and Southern Man, for example.
    Mar. 5th, 2004 06:56 pm (UTC)
    Well, it's heartening to hear that you'd be interested in it, at least. Hopefully you're not the only one!

    The idea is that anyone could any any key/value pair that they want. If someone wanted to add "HIPPO → birdie" as a piece of metadata for a song, they could, although it might not be meaningful to anybody else, and might get deleted by another user. So adding "REFERENCE → Southern Man" to Sweet Home Alabama would be easy.

    However, once I start making "special purpose" metadata fields, suddenly everyone will say "why isn't there a special field for X and Y? You have one for Z!" and I'll have to constantly be reevaluating and changing things. When I mentioned this on the MUD, someone said "You do realise you'll never, ever have enough predefined metadata keys, right?" and I responded by saying "Okay, then, I won't predefine any of them. I'll let the users define them as they see fit." It's not a perfect solution, but I don't know if there really is such a thing in this case.
    Mar. 20th, 2004 09:01 am (UTC)
    I once built a photo database where users could add keywords to images. The GUI looked like this:



    Top Level keywords:

    [ ] people
    [X] animals
    [X] outdoors
    [ ] indoors

    Other keywords:

    [X] puppy
    [X] rainbow

    Add keywords: [textfield]


    The top level keywords all showed, with the ability to check them on. The idea was to put things here that were very common in pictures.

    The "other keywords" were all user-created. Only the keywords matching the image were displayed, since there were so many of these.

    When adding keywords, the user would just type in a comma-separated list. So in my above example of a picture of a puppy and a rainbow, I could type "turtle, photographer:joe smith, camera:sony dcp-100, clouds".

    Keywords in the list that already existed would be associated with the image. Keywords that didn't exist would be created and associated on the fly. The colon separated name/value pairs was something I came up with after the keyword system that allowed me to have metadata without building a metadata system. It actually worked really well.

    Any keyword could be toggled "top level," even those that had been created on the fly. Top level really just meant that the keyword would always show up for someone to toggle on.

    The problem you have is that people will put "car" and "automobile" to mean the same thing on two different pictures, so I built an alias system. Anyone could pull up any keyword and add a list of aliases. If someone typed in one of those aliases, the "master" keyword would be associated with the image. This was very useful for common misspellings as well.

    I never did this, but I could see the ability to add aliases for names and values in pairs individually. You could alias "date" and "date released" so that the latter would be translated on the fly to the former before storing in the database. That would help to keep your metadata consistent.

    So here's a quick adaptation of that system:


    Predefined fields:

    Song Name: Atomic Garden
    Artist: Bad Religion
    Album: Generator
    [and other standard id3 tags]

    Lyrics: [complete lyrics to song here, for searching inside] (Likely violates copyright law.)


    [X] nuclear proliferation
    [X] war
    [X] russia
    [X] cold war
    [X] bombs
    [X] irony
    [X] gorbachev
    [X] lyrics:greg graffin
    [X] music:greg graffin


    Whew. Sorry for rambling.
    Mar. 5th, 2004 11:28 am (UTC)
    The song database: Great idea. Perhaps it coule be like IMDB's, with keywords: RAINBOW, DREAM, WISH, STAR, SIREN SONG.

    I once remember seeing a reference book of popular song in film, that tried to be a complete listing of songs used in movies. Very useful for those old films where you hear some studio song and wonder what other film you've heard it in. And really points up how in those days when studios owned a good song, boy would they use it.
    Mar. 5th, 2004 02:12 pm (UTC)
    ( 5 comments — Leave a comment )

    Latest Month

    February 2011
    S M T W T F S
    Powered by LiveJournal.com