For the last two years I’ve been picking at this concept of AI improvisation. I wanted to jump on this quickly as the AI models learn really fast and wanted to take advantage of it’s dumbness while it still made interesting sound. Think of it like the AI is my studio musician’s band or a new synth. The concept is to do 15 sec pieces because that’s the average attention span of “these kids these days” according to the data.
I found an audio generator that made interesting sound. I don’t want to tell you which one because I don’t want to promote them as they were really difficult to work with. They kept changing pricing and policy while I was creating. BUT – the real fun thing here is that the text prompts are all Psychotronic Film titles like “A Bucket of Blood” or “Dracula Has Risen from the Grave”
So it’s sort of a jazz improvisation between me and the AI but challenging it to interpret a surreal or avant garde prompt. I would do this improv until I had enough material to mix something spatial that was interesting to me. The results were consistently amazing until they “updated” the model and it became boring. Luckily I was still logged into the first model and just put my head down to generate as much content as possible given the time frame and amount of credits I had. Then, the dreaded Windows update caused me to lose contact with the first model. Luckily, I was pretty much done and could struggle with the version 2 model to finish the pieces.
You’ve heard of a sound bath? This is like a sound drowning!
Then, I used the same prompt to create an image, then using an animator model that allowed 15 sec lengths, animated the image, plopped on the spatial mix and DONE!
I’m releasing this on Oct 6th when the moon is full. I don’t believe anyone has done anything like this before and I don’t think I will be able to repeat it as the model has been alive for too long now and is probably churning out Kenny G clones right now.
This release is 120 pieces to be consumed 1 at a time for 120 days.