Idea: a "super single" CD format

[mdlbear]

Inspired by a recent article in Electronic Musician on remote collaboration (unfortunately not on their website yet), and an even more recent encounter with PTXdist, I've been thinking about ways of shipping Audacity projects around on CD.

It struck me that this could be an interesting album format, which one might call a "super single" -- a CD-ROM containing one or two songs (an Audacity project can easily hit 100MB, since it comes to about 10MB per mono track-minute) plus the software required to play with it: Audacity for Mac, PC, and Linux, and a bootable Linux distro (such as a stripped-down DeMuDi live). Maybe even the CD audio of a sample mix, as a concession to those Luddites who feel that a "single" should have an A side and a B side that you can actually play.

Of course, I immediately realized that for real collaboration, it would be better to just set up a subversion or cogito server with a fast upload connection: since Audacity stores audio in 1MB blocks, a collaborator would only have to upload the new tracks and the new XML project file. On the gripping hand, a CD would be a good way of bootstrapping the process, since even on DSL a couple of hundred MB takes a long time to transfer.

Any interest in this?

Comments

[aerowolf.livejournal.com]

If Audacity could be convinced to handle audio in 128k blocks, then it would work really well -- subversion could be used to handle changes to the project file itself, and everyone could have their own local copy of it for whatever branches/changes they wanted to work with.

Of course, it'd be REALLY nice if Audacity could work with the revision control system directly, and save its changes to there instead of out to just the generic disk files.

(An additive project to this could be a MIDI editing system that would save in the same repository -- and a way to associate one or more waveforms with that MIDI file revision, for the rendered versions. Since there are multiple renderers, it might make sense to have the ability to associate the name/ID of the renderer with the waveform it rendered -- and a 'standard' soundbank could be the one that Java has as part of its optional components.)

God, I hope that's understandable.

This way, anyway, someone who's good with synthesis could put his raw materials in the repository or on the CD itself, and let the people who work with it later figure out how to render it. It would also allow for fairly precise timing of a played track with a layered vocal or analog instrument -- listen to the rendered version of the track while singing or playing an analog instrument.

[mdlbear]

It's understandable, but Audacity's block size is fixed as far as I know. And svn doesn't really care. git cares even less -- files are treated as indivisible blobs.

Unfortunately Audacity doesn't do MIDI; you'd have to go to something like Ardour for that, and I don't know how it stores its data. Don't think it's compatible.

In my case, I'm an acoustic guitar player, and Audacity's what I'm working in, but I need to play with Ardour some.

The virtues of open source

[aerowolf.livejournal.com]

Hack Audacity so that its block size is much smaller (this would make it possible to migrate the engine, at the least, to smaller devices as well). Then...

I'm envisioning a storage system where a MIDI file is stored in the repository, and then "renditions" of that MIDI file (on different synthesizers) could be attached to that revision of the MIDI file. Change the MIDI, and the new revision doesn't have any renditions until one's created and attached.

I mean, if you're going to go for the collaborative sharing of music, you might as well make it possible for everyone to play along. Midi files and 'tracked' files (basically MIDI files that include their own samples for the player to render it with, initially started with the MOD format from the Amiga) are perfect for source-editing -- you don't have as many options with a waveform as you do with a sequence of commands, which means that any changes will have to be done manually on the waveform. Yes, Audacity makes that fairly easy, but you can't exactly change the pitch of a horn from middle C to one octave down without it sounding really strange.

So, why not make it possible? Why not make a back-end that's compatible with both concepts? (Storage size of waveforms is, as you pointed out, an issue -- if necessary, waveform renditions of MIDI files could be dropped from the repository to save space.)

Vocals always have to be waveforms. Analog guitar and its specific tonalities do, as well. But, think of it this way: how would you be able to differentiate two renderings from their waveforms alone? If an author changes four bars in the middle of the song, would you be able to figure out from the waveforms alone which version is the one that's meant to go into the final mixdown? (Renderings are pitifully easy to make. Revision control systems are pitifully easy to clutter up with revisions that sound almost alike, and it's easy to mix up which revision you want to go into the final mix.)

(I'm not trying to be argumentative... I just know that my own instrumental capabilities are best with synthesis. I can't count the number of times I've screwed up because I deleted the wrong revision of the rendition.)

Re: The virtues of open source

[mdlbear]

Audacity isn't my program, and there are good reasons why 1MB is a better block size for audio (like, fewer file opens means less OS overhead and fewer disk seeks).

As I said, there are programs that let you work with MIDI, but as far as I know none have been ported to all three popular platforms. Feel free to research the field; you can start at http://linux-sound.org/

[aerowolf.livejournal.com]

Also: your idea of sending audacity stuff around is good, don't get me wrong. It's like the idea of sending individual tracks off of the master reels to remixers, which is good in and of itself. (Creative Commons sampling license, anyone?)

My own thoughts are on an entirely different tangent: it would be nice to be able to merge these two ideas, allowing anyone to work in whatever medium they want in the most efficient and effective way possible. (There's a website out there for musicians and singers to get together and form a virtual band -- I think it's myvirtualband.com or something like that -- that this kind of project would be absolutely PERFECT for.)