Part B of my shower notes of the day, which are really two distinct thoughts and part B came first.
music loom
I was thinking about how I want to make a music loom
once the API drops for UDO or once some local model that is sufficiently powerful becomes available
and actually those already are out and you could adapt this to those models pretty simple, like pretty easily
What I’m about to say is
I realized that I could make the loom right now in the same way that Janus was manually creating and managing large branches, very deep, dense branches, like networks of branches.
We can make a loom client that does that, that doesn’t have an API, but that accepts fields and facilitates expanded work, connective work, relational work.
as in parent-child, as in sibling relationships, queuing of multiple generations where you are the API, as the human with the interface.
loom interface
So what I mean by that is, it’s a client that you paste the URL for a song, right?
You paste the lyrics that you used, and this is if you’re using retroactive, if you’re storing retroactively, but otherwise, use this interface for everything.
maybe everything beyond an initial idea, right?
So like, okay, cool. I had something I want to extend it. Okay. Paste it in because you’re going to, you’re going to get two things back anyway.
Right. And you’re going to want to make more than just two probably unless they’re great.
So paste in your extension, paste in your prompt, paste in your lyrics and store the results in the loom interface.
And on the client side, this loom interface should be able to accept an mp3 of the track.
It should automatically segment out because it knows, based on the parent-child relationships, which part of the audio file that is downloaded is the segment.
and what you get is like you you can build this like very interesting player where you can like kind of like Ableton or something but not anything like Ableton
and this is where actually I want to refute whatever Conrad said that’s like this is a bad interface I don’t think it’s a bad interface in the same way that like a camera is not a bad interface as opposed to Photoshop.
They’re just completely different tools.
This is not the wrong interface for generative music stuff.
And we will have and we will need more musical interfaces.
But Ableton actually is a perfect example of why you don’t want to just copy Ableton.
Ableton did a lot of things in ways that are not very inclusive in terms of HCI.
It doesn’t have the best HCI.
unless you’re like a very particular of a particular demographic
I want the sections to mouse over or press space after moving up and down in the sibling tree.
You just hear the differences every time you move.
For example, I want that to be very responsive.
and so forth.
And you know, one more thing.
I want it to feel like loomsidian.
I want it to feel like… I want branching, reading, listening, whatever the parallels might be.
I want it to feel as natural and to be expressed in a way that could be read by a language model if you were to export it to Markdown.
conclusion
And finally, here’s how I thought of the idea, if I can remember.
I can’t right now so I’m going to record a new note once I remember the seed that was also of interest of like this.