I first started dabbling with AI for art and music years ago.
It was fun, but it was also frustrating. I could get interesting artifacts, but getting the tools to produce what I actually had in my head was hard. Most of the time, the gap between the idea and the output was the whole project.
That has changed in the last year.
The models are better, the tools are better, and the loop is tighter. I can stay with the intent longer instead of spending the entire session negotiating with the machinery. That does not make the work less human. It makes the human part more visible: taste, memory, references, judgment, editing, and knowing when the thing finally feels like the thing.
Below is the first end-to-end song and music video that really ended up where I wanted it to be. Everything you see and hear is generated, though everything it's built upon is 100% human.
This is the rw:fwd:tapes project.
Keep reading