Performing music without performing
An idea for a way to sculpt music
I am a musician. I play the piano. I don’t play as much anymore, because my wrists are not amenable to such abuse. (My technique sucks. I blame one of my teachers.) I also program computers. My wrists don’t complain as much about this. (If you spend more than a quarter of your programming time at the computer actually pressing keys or moving the mouse around, you’re doing something wrong.)
I think the root of my enjoyment, for both of these, is taking pleasure in creating complex, beautiful things. And yet, I’ve always enjoyed programming more. Why? Because while coding well takes coordination and practice, the end result is an object. And an object that can be improved, more and more, each time you work on it. Whereas with piano, the coordination and practice are an integral part of the end result. The result is a performance, and each performance has mistakes and problems. With every new performance, you get thousands of new chances to make mistakes. The only way to improve a performance is to practice it, over and over and over, until the number of mistakes goes down to something managable. (Even if you get all the notes right, you always make mistakes. Good pianists can intentionally produce around two dozen different loudness levels, and can tell the difference between a note hit at the right time, and a note hit a tenth of a second too soon or too late.) Code is perfectable, and permanent. A performance is not. Working on code, you’re always doing new things. You don’t repeat yourself more than necessary. Once you get something right, it’s right. Working on a performance, you’re slowly training an obstinate and inattentive nervous system to do something replicable.
For a while, I had a real grand piano. It was nice to practice on, because it’s a real piano. It’s the kind of piano real pianists perform on. But it was a bit unwieldly. So I switched to a digital piano. By far, the best value for the price. Much more recommended than any real used piano under $12,000, or new under $20,000. (If you have a good ear and the seller doesn’t, some used pianos can be great deals.) So, anyway, I got to start experimenting with MIDI recording and editing and playback. Quite fun. As it turns out, though, messing around with a recorded live performance is actually pretty hard. So while I was able to make some small tweaks to my performances (like fixing missed notes and deleting backtracks and tweaking the chord voicing here and there), I found that live performances were still the gold standard. You couldn’t get anywhere close to the quality of a good live performance by starting from scratch.
But, just two days ago, I realized why it was so difficult. It was difficult because I didn’t have the right interface to create it. I didn’t have an interface that was oriented towards letting me interpret a blank performance the same way I would go about it in real life practice. (Now, what I call a “blank performance” is a computerized version of a piece of music that has all the right notes at the right times, with no variations in velocity or dynamics or articulation, with no raising of the melody, with no pedaling. You’ve probably heard such music on some cheesy mid-90s style web page, when some people still thought music on web pages was cool.)
But I realized exactly what such an interface would look like. It would show the music on-screen in a normal musical format. It would understand the difference between ornaments and real notes. The main concept would be applying envelopes that modify different parameters to different groups of notes. So, you could select all of the notes in the melody for a few measures, and create a group out of them. Then, you edit the velocity envelope for that group. Initially, you see a straight line going across the page. You create handles on that line that you can drag around to raise and lower the velocity for those notes over time. For a melody, you would raise the line. You can create groups of notes that are all on the same beat, and then edit the envelopes from top to bottom. You can also edit an envelope for the length the notes are held down, and for when the note starts. (Sometimes a small delay on the top note of a chord sounds nice, and with vertical groups, you can use it to create rolls.) You can edit the tempo as an envelope, though it’s not attached to note groups. Apart from the tempo, you can insert pauses of various lengths. You can mark when the pedal goes down, and where it goes halfway up (up for about .3 seconds and back down, allowing about half the sound to clear), or a quarter up, or all the way up. Groups can overlap, and they’re combined automatically, so you can use one group to do the broad dynamic changes, another group to do smaller, subtle changes, and additional groups to tweak individual voices or to voice chords.
I think that with an interface like this, you could make very accurate, subtle, and good interpretations of any piano music, provided you have a good playback device. (Hint: there isn’t a consumer sound card out there that can play back MIDI even close to well enough.) In fact, if I do say so myself, I think performances that are “sculpted” like this could be much better than any live performance. If, as a musician, you’re able to do everything you want to with expressiveness, then you’re at a huge advantage over live performers, because you can play scales and arpeggios and cadenzas as blazingly fast and superhumanly even as you want. The speed of your fingers is no longer a limitation.
The main difficulty with making music this way, (besides writing the program in the first place,) would be getting the notes into the computer and into a readable format for every piece. Many MIDI sequencers have a basic, buggy, and ugly notation display. That might be good enough. Maybe. It would be nice to be able to write this as a plugin to one of those programs. But even then, getting the notes entered and interpreted correctly is potentially very tedious.
Also, this program would not be able to produce non-piano (and organ, and harpsichord) music rivalling the real thing, without a good physical modelling synthesizer for the desired intstruments, and those synthesizers are even more expensive than good piano synthesizers. (I’d love to have one, though.)
And, shouldn’t something like this exist already?
