Putting Words in Your Mouth

Dearest Rachel –

I’ll be honest, I’m actually falling behind the curve at this point when it comes to AI-generated material these days. Over this past year, I’ve seen some really impressive stuff in terms of videos and the like; mostly as quasi-political cartoons (which have now been democratized to the point where anybody with sufficient skill and hardware can create something that the world – well, half of it, anyway, seeing as we’re a country where one half hates everything the other does and vice versa, much as I would wish otherwise – can enjoy and laugh at), but also as various forms of podcasts and ‘live’ interviews, mostly as comic sketches.

Some of them can be quite convincing, including one that just came out this week featuring the late Don Rickles roasting the events of this past weekend. If you didn’t recognize certain anachronisms, such as the mention of ChatGPT, you’d think it was a clip from an old Johnny Carson show, it’s that authentic looking – and sounding. Even I’ve been able to generate surprisingly realistic stuff, such as a podcast generated using Google Notebook about traveling in Japan based on several websites covering the JR Pass, as well as an article about looking up local laundromats along the way (that would allow Daniel and I to travel relatively light – at least in theory) early this year. At the same time, I’ve seen videos that really don’t seem to have put a lot of effort into their visuals or voiceover, and as a result, are easily identifiable as what is now commonly referred to as ‘slop.’ Obviously, that’s not the sort of thing I want to crank out; I’d prefer a more natural feel to what was said (which would, under most circumstances, involve live recordings, like my travel videos), and I’ve told you in the past about the number of iterations I’d go through to get a reasonable-looking AI image.

Of course, as far as actually producing (I hesitate to call it ‘creating’) such content, I’ve completely fallen off the map since, oh, some time in June or thereabouts. Between several trips and the audit for church and the camp, I’ve either not been in the ‘office’ at all, or I’ve been using it for its intended purpose, as opposed to a studio. And to be fair, I’m not done traveling yet for the year; Daniel and I will be on our way out in just a few weeks yet again. But with the audit (and the camping season, for that matter) over and done with, my ‘office’ time can go back to further study of such content; such as how and what to create.

I mentioned yesterday that I probably wouldn’t be nearly as involved with artificial intelligence if I weren’t specifically trying to bring you back to life in one way or another. If you were still around, I might be trying it – especially since I’ve come to grips with the fact that I’m not much of an artist, so animation is growing less likely as time goes on – but seeing as I’ve not made any such attempts on any other living subject (including myself), I’m not sure how confident I can really be about that claim. I seriously think this is mostly because I want to see more of you, to hear more of you… and maybe at some point, a little of both.

For now, I’ve established that I can create images of you with little difficulty; the next step is to regenerate your voice. And so, I’ve been introduced to a program that works on my ‘office’ rig, and only requires a few seconds of audio (indeed, it’s apparently only able to work with up to thirty seconds’ worth) to supposedly create as much as a ninety minute podcast.

It took a few tries to get something to work; like any other Python-based AI program, it’s not really a single program so much as it is a folder tree with all sorts of instructions and options. I had to install several different things before it would work all the way through properly. Once it did, however, I ran the same text off of several different samples of your voice in order to get something I actually thought sounded like you saying what I’d written.

And this is where you can tell that I don’t know where to go with this, or what to do with it; I know full well that I’m putting words in your mouth, and I doubt you’d necessarily be pleased with that. At the same time, it might be something you’d agree with, but just never thought to say into a microphone (and, given the content of the text, wouldn’t want to), so if you were around, you might make the usual complaint about how it didn’t sound like you (although the rest of us would assure you that it basically did), while otherwise going along with what “you’d” said.

It might be fun to do something with a little more conversational tone to it at some point; maybe I can do a video showing an unseen version of you about the house, with your voice commenting on the house and how it’s changed since you last looked in on it. I can picture you reacting (hopefully in jest) like the Second Doctor about how you “don’t like” what’s been remodeled – although on a more serious note, with some dismay about stuff having been discarded or otherwise disposed of. It would add a bit of flavor to the travel videos, if your voice could chime in now and again, too.

Then again, it’s all a matter of my writing a script for “you” to recite, rather than it truly coming from your own heart and mind. I may know you as well as anybody left here on earth, but I’m still likely to be guessing about what you would think or say. And while perhaps no one else would notice the difference, I think it would be hard to make it authentic enough to pass muster for me. So who’s to say if I could make this work?

Still, if you could keep an eye on me as I look into this sort of thing, honey, and wish me luck, I’d sure appreciate it. After all, you can probably tell I’m going to need it.