Dream Deficit Zone #22

A track finally made it to mixing stage this past week! I struggle to describe the lightness it has left me with; it all now feels like real progress after an endless period of tinkering and tweaking and very slowly assembling music out of a million ideas. I’m happy with how Async sounds. It still needs a deliberate haircut before we can let it step out.

The focus has now shifted onto the almost-done-but-just-shy batch of tracks. I remember Doomscroll from when I first wrote it many years ago, and I still see bits of that spirit in there - a deep solemnity caught in the abrasion and cacophony of the arrangement. How do you get a song about the tediousness of being online to not feel...tedious?

More progress on Afraid To Try as well: I’m pleased with the phat drums for the ravey part of that song. After all, this song did grow out of two years of performing it at algoraves.

I played two shows recently. One was an acoustic set. Matthew and I decided on it when we found out that Young Ethel’s had an upright piano. We played acoustic renditions of some of the WIP songs and some old “hits”. I also cobbled together a completely new track that’s currently called “g minor song”.

Here’s a video recording of that show from a friend. We sound caught off guard:

A few weeks ago, I went down the rabbit hole of fabricating instruments with a 3-d printer that I happen to have unfettered access to (inadvisable). It started with a pocket whistle and a recorder, and ended with a transverse flute. I may have enabled a few friends to go down this path as well. The room is now filled with okarinas and slide whistles.

Getting the embouchure right on the flute is a struggle. Blowing too hard for too long (as I did with bad technique) can actually deplete carbon dioxide in the system and make you light headed. A giddy feeling (not of joy) when that happens. With all that, I’ve managed to crawl around just enough notes to play vaguely intelligible melodies. Tomorrow Never Knows by The Beatles, anyone? How about Trapdoor by King Gizzard?

Last month, I also got briefly engrossed with the RAVE model from IRCAM, and took it upon myself to train one. I liked interrupting the training process getting the poorly trained versions of model to spit out noisy, incomplete and unintelligible bits of sound.

This whole thing was spurred on by a Joel Simon essay, which talks about the artistic excitement of discovering emergent properties in works-in-progress. Modern AI systems (transformers and diffusion models neatly manicured with an RLHF) rob that from us, presenting only with a little prompt interface and then a fully finished output.

I’m enjoying this process preventing these models from acquiring a clear idea of what music should be, and forcing them to make sounds anyway. I’m all for the uninhibited rawness in art that ensues when you enable the right circumstance. A recurring theme in my current work is this idea of constriction (of sound signals, ideas, emotions) and it tickles me to constrict an AI model like this.