Participants 2024
About the TEAM
Cabe is a music production graduate who is constantly looking for new, creative applications of audio AI. He also resents AI for how good it's becoming, because soon people will think all his music is AI, which it isn't. :(
About the SONG
"Whatever" portrays the stream-of-consciousness thoughts of a writer alone in their apartment, obstensibly to work on writing, yet their solitude induces a mental decline, leading to substance misuse. The lyrics are therefore all writing-based euphamisims to substance misuse; the "lines" they are writing aren't words, and the "paper and pen" they constantly "grab" are not for the purposes of writing. The song also portrays their manic movements around the apratment while inebriated, causing chaos and destruction in their wake.
About the HUMAN-AI PROCESS
The key feature of my workflow is somewhere between 12-18 hours of editing in-DAW (Ableton), and 1-2 days prompting and collecting the outputs. This is something I suspect not many do, because of how 'granular' a workflow it is. Editing hundreds of outputs also leads to "paralysis by analysis". This approach can be likened to a gambling addict in a casino: 1-2 days turbo-prompting such that the AI forgets the context, hallucinates, but hopefully gets to know you enough to take some risks. AI is constantly trying to gauge user intent, so for example, your first few chats with an LLM will give largely the same tone of response, but that won't be the case many hours later, once it's built up a more nuanced picture of who you are. Applied to audio, this means that for best results, dedicate a few hours to prompting. The problem with all of this, and what makes this workflow so granular, is the AI starts to drift over time, losing sight of one aspect of the prompt in favor of another; outputs begin to differ in length, timbre, language (for some reason) but most importantly tempo. The editor therefore has to zoom in so far to each clip where each sample (as in the "sample rate" of an Audio file i.e. the audio equivalent to "resolution" in images") is represented as a 'dot', and match the exact corresponding sample of say, a kick or snare, to the corresponding one on another track, and fade them together to make a seamless transition. For this reason, when people say they "edit their songs in DAW", I'm not entirely sure what they're referring to, because I can't imagine many have the patience for that.
Lyrics
Write a few lines
Don’t know what they mean
I love these little white things
Break a mug
Crack a smile
Bastardize my favourite style
Slip and slide
Up again
Blood check
Grab paper and pen
Write a few lines
Don’t know what they mean
Don’t know how I feel
Break a mug
Crack a smile
Bastardise my favourite style
Flashing light
[Light] Up again
Blood check
(Cracked again
Grab paper and pen
Sterilized
Don’t know what they mean
I love these little white things)
Break a mug
Crack a smile
Fetishize it for a while
Slip and Slide
Up again
Check
Grab paper and pen
Break a mug
Crack a smile
Bastardise my favorite style
Try to drink
It takes a beast to tame, It’s insane.
Raise the dead
Born again
Crack the mirror with my head
Break a mug
Crack a smile
I’m fed another day
Voices in my head, they all like:
“Lock the door”
“Hit the floor”
“No one loves you any more”
Oh well.
Okay.
Okay.
Okay.
Taste of blood on my skin
Keep the Savior from a win
Break a mug
Crack a smile
Bastardise my favourite style
Catching flight
Up again
Blood check
I’m at home
I know I’m not alone
Uh
Oh well
Okay
It’s whatever, so they say
Well
Okay oh yeah okay
It’s whatever so they say
Okay oh yeah okay
It’s whatever so they say
Okay oh yeah okay
It’s whatever so they say
Slip and Slide
[Got the style
Does it matter if I die
Taste of blood
Up again
Didn’t taste the love within
Raise the dead to life
Turn to stone
Break a mirror with my fear
I’d like to testify or shut this down…]