So hi my name's Lynn, I'm a long standing interdisciplinary process artist and musician from North America. And welcome to Stoned Storytime. This is where I get stupendously high and talk to you about how I use AI the hard way to make art or music.
Part of why I do these the way I do them, is because a lot of people seem to be under the mistaken idea that AI Art is easy, and only involves a couple of seconds of work. However, as someone who's deep into it, and uses it in their work, I have a different perspective on the topic.
And I hope you might provide me the opportunity to share it with you here.
My promoter requested that I move Stoned Storytime to the blog, rather than post about it on the descriptions for the youtube videos. And since that made sense, I'm going along with it.
Some of my songs are part of exercises I give myself to improve and work with my skills as a songwriter. When I'm doing it honestly (sometimes I cheat), it's an opportunity to stretch my wings and do something interesting with my time, while growing as a songwriter. If you've read my process breakdowns before, you'll know that I'm a conventional songwriter that uses a process called "neural bending" to produce studio tracks with AI. Frequently, I employ another process called "Noise Painting" to subvert or repurpose the AI's I'm rendering with, by essentially getting them to go on a trip that bends their sense of reality. It's the closest thing to feeding a computer acid that anyone's been able to do so far.
Neural bending with a noise painting can be done to any ai in any medium. And it involves feeding the machine inputs it doesn't expect to see, that will cause whatever type of system it is to go off the rails, and balance the math in such a way that new information is produced. As it relates to noise painting in particular it's all either about It's also proof that the "ai's can't produce new things BY DEFINITION!" guys have always been full of shit. But you didn't need me to to be the one to tell you that.
Anyway, that's a stupidly verbose way of telling you that I gave myself a writing assignment.
And that assignment was... Take a song from the back catalog that fits the following criteria:
- You have to hate it. Sometimes, I write songs that I think are terrible.
- It has to be a year old or older, so from a different era of Lynn Cole songwriting.
- It has to be live on soundcloud, and not so bad that you took it down completely.
Then, do what you have to as a songwriter and producer to MAKE IT AWESOME! But, don't substantially change the core of the emotional song, keep the verse count the same, and preserve as much of the original piece as possible, while still making it awesome.
As I learned during this assignment, that is harder to do than it sounds. Also, these are challenging for a reason. I don't go easy on myself when I write.
There are several songs I've written that fit this criteria. But this one had bothered me since the first day I wrote it. The song I decided to dredge of from the depths of my Soundcloud? "I Just Wish You Were Here. "
I wish you were here is #138 in my catalog, which means I wrote it after Poison Pill, but before Somebody Else and Burnt Out.
Another problem? Let's just say, I wasn't quite right in the head when I wrote it.
Oh boy. Let me give you some background on this.
When you've been married for a long time, and it's just not working anymore, and you can't get away, so the relationship goes toxic, you end up in this really weird headspace. It's a combination of sunk cost fallacy, resistance to change, and the realization that as much as you HATE the person you're with, you still love, and probably always will. Even if living with them, and being intimate isn't really an option anymore.
It's this really weird state of indecision, where, on the one hand... you know damned good and well that things aren't working, and the relationship isn't healthy and never could be. But, y'know, you wish things were different.
And that, I think is the emotional core of the feeling of both versions of this song.
In all about four days of songwriting, so 60ish hours, which includes testing and prototyping at every step of the way.
Let's break down that process.
First I write a draft of the song, the way it comes out of my head. This was a good song to start with because that's what it was. Last year, I was just learning how to do all of this, and I didn't have as much of a process, so what you're hearing in the suno version of this song is about what my process looks like, just getting it all out of my head.
Then, I work with suno, to get something nice and burned sounding, to use as a melody prototype. Suno's actually got some bugs right now in the sound quality, and they're actually amazing to feed to udio. You see, when you feed it a distorted or robotic vocal, it melts it in this really pretty way that sounds full and wide, but without sounding robotic in the final render. And it gives udio a lot of really nice spectral sound texture to move around. That is after all, what diffusion processes do.
The cool thing about using Suno as a prototyping machine is because of how fast and cheap it is relative to Udio. Also, it has more models to choose from, and some are a lot more fried than others. Also, I think the melodies it comes up with are more interesting, more haunting, prettier. This song had about nine major iterations where I went back and forth, the Process playlist has all of them, and it's interesting to listen to the different versions of the song, and see how the lyrics change from take to take.
I did a lot of Udio
remixing, at one point, I broke an early version of the song into stems and manually changed the gender of the vocalist. It's pretty neat.
You can find the playlist on Udio