Review

In a design process, what we actually make is not always identical with the intent and processes which guided us in making it. Otherwise put, a good artifact is subject to multiple interpretations. So what did I actually make?

• A tool for musical ear training, through the playing for known melodies and variations of them. The ear-to-finger process is the “[real meat of music] https://www.youtube.com/watch?v=NU9jx8vRW_E&t=612)”
• A tool for learning complex melodies progressively, based on simpler ones.
• A new demo for Google Magenta, which develops from e.g. Beat Blender by insisting on the user being able to play on similar terms to the computer.
• A tool for automating a generative version of A Dozen A Day.
• An experimental system for deriving new melodies and so songs and compositions from existing ones, while insisting on the musician being able to play them. A bit like how I’m told hairdressers learn to combine styles.
• A tool for musical improvisation training.
• A means of training human musicians in interpolation (i.e. internalisation of the operations done by Google Magenta AI)
• A tool for training musical sight reading with an infinite supply of examples (but an incomplete score making algorithm).
• A tool that doesn’t work unless you have a working MIDI setup on your computer
• A remix of some of my previous projects and a couple of Google Magenta demos.
• A deadline, just.

Process notes

• Well written coding abstractions usually end up being one level lower than they could be, because functionality has been achieved.
• My actual work had only a partial relation to my proposal “project milestones” (e.g. coding and learning redux took longer than expected, even though it was a good idea, so I didn’t have adequate time to test).
• I included about half of the ideas for the prototype. The essential elements though are all present, and as an “experimental system” it works.
• There is enormous merit in choosing and taking the time to set up the right tools: e.g. eleventy.js + markdown has worked for this blog; and Redux allows very easy implementation of state saving and reloaded, meaning the site remembers where you are (using local storage) between sessions.
• This project has developed my previous “Imitation lab” well. There is enough scope for another repeat of this iterative process (+analogue sources; +video; +user testing; +easier sharing of memes / melodies)
• Deadlines work. Thank you NYU!