DIY Zombie Karaoke

Dearest Rachel –

Since I’m meeting with Lars today, and I just got confirmation that everything is a ‘go’ as far as the church books are concerned (which is to say, everything from last month has finally been entered, and I can run the numbers and assemble the reports for the board members), I have plenty of actual work to do today at the ‘office.’ What this means is, that I have to get this letter off to you that much sooner and quicker, because I won’t have the time to put anything together once I’m there, and after I leave, I won’t even be near a computer to finish assembling my thoughts (yes, I know I can use my phone, but it’s not an ideal solution).

I had a couple of weird dreams last night, one in particular involving myself in the military (which had me thinking at the time that “it’s just like being in the marching band in high school, except with rifles,” they have an issue with you calling your issued weapon a ‘gun,’ I should mention, “instead of instruments,” until I remembered that I could have joined the band rather than an actual fighting platoon, which would have been so much safer, especially given the C.O. my imagination had assigned me. Let’s just say that Uncle Milty would have been preferable), but they rather evaporated a bit as I began to pull myself together this morning. And in fairness, you might be able to tell from the tail end of that parenthetical comment that they had a certain… political… bent to them that I really wish I could avoid in these letters. I know you and I would discuss politics a lot back in the day – and, since we were more or less on the same page, we could do so civilly – but as that no longer matters to you where you are, I don’t see the point in paving my path with eggshells when this is available to a wider (and apparently, thinner-skinned) audience.

I even have a couple of drafts in process – well, technically, there’s something like seventy such letters that I’ve started, and not finished for one reason or another, but these are about ongoing issues that I’m dealing with, so I keep adding material as I go along – but since the activity I’m describing in them is an ongoing thing, I can’t exactly wrap them up at the moment and send them for you, simply because I’m strapped for material and time, now, can I? Besides, I have plenty of material on the topics; it’s just that more is constantly coming in for me to add and assemble.

And so, with that last phrase mentioned, I thought I might tell you about something that I’ve just learned about, which might have enormous potential, if I can only figure out how to break everything down to its component pieces (and, for that matter, determine what pieces I actually want), and rebuild it into whatever it is I have in mind.

You might remember, before you had to go, I was trying to determine what exactly I wanted to do with my theoretical channel, and realized that one of the things I needed for any reasonable content was background music. Of course, I wasn’t (and still am not) a good enough musician to assemble professional quality stuff together on my own. But at the time, there were some nascent concepts floating about the internet, where AI was believed to be able to – not yet, but soon, given enough data to study – mix artists and styles and come up with something… not completely new, but different enough that any resultant re-combination would be sufficiently unique as to be effectively original. I was studying OpenAI’s Jukebox program, but at the time, most of what you may have seen over my shoulder was either purely theoretical, or just not available to the public (and certainly not available in open source format, for one to use as one pleases).

Well, if the future isn’t here yet, it’s as near as by now. The bad news is that there are other creators out there who have beaten me to the punch, and are issuing musical creations by dead celebrities – such as Frank Sinatra crooning a version of “Gangster’s Paradise,” or Johnny Cash singing “Barbie Girl” to the tune of one of his Folsom County prison songs (I forget which one), complete with crowd noises from said concert, and a bemused interjection from Cash to the effect of “this is what they’re calling music these days, huh?” – the good news is that this sort of thing is now apparently available to the interested public.

Thus far, I’ve only been pointed to a program that will take a recording, and separate the vocal and instrumental tracks using AI (thus creating a do-it-yourself karaoke track – what we could have done with this sort of technology, back when we were in college!). From there, there are collaborative programs that can be run (and presumably, with enough oomph from one’s own computer, they can be set up on one’s own machine with a reasonably user-friendly interface) that will take the vocal track and use it to generate artificial vocals from any one of a number of voices in their data banks, allowing one to put the new voice up against the original instrumental. The drawback is that this doesn’t change the style of the song; it only re-renders it as if sung by a specific dead (or otherwise; I haven’t personally checked out the library of potential voices), but it’s pretty impressive for less than an hour’s work, if the tutorial I’ve seen is to be believed.

Obviously, there’s a lot of things I’d like to do with this kind of technology. I could imagine our church’s music as performed by Michael Been of the Call (especially since, like you, he’s no longer around to make music) or Michael Roe of the 77s. Maybe the ‘real’ Art Garfunkel singing a duet with one of the girls of Garfunkel and Oates (they have several songs that read like dialogue between a boy and a girl – I hesitate to say they have the maturity of a man and woman, but whatever)? But those require me to determine how to create a voice-generation file on my own, which I haven’t figured out how to do just yet.

And besides, I think that, while you were no professional singer, I’d want to recreate your voice first of all. I’m sure you’d understand why. Granted, there’s a lot less material to work with, but it might be easier if I’m not trying to get you to sing.

For all this, however, there’s probably nothing that AI can do that real life can’t do better… yet. You’ve already heard me complain that what I’ve discovered thus far can’t change the style of the original song (even the ‘Sinatra’ renditions are built from jazz versions of certain songs that already exist for consumption on the internet); here’s another piece of music I discovered only yesterday, a version of Friday I’m In Love with Natalie Imbruglia, giving it a country & western (wait? isn’t she Australian?) flavor, of all things:

We’ve got a ways to go yet, honey. In the meantime, keep an eye on me, and wish me luck. I’m going to need it.