Part piano, part machine

What a 1982 piece for piano and computer shows me about Savor today.

Founder journey · Part 3 · October 5, 2026

6 min read

Every time AI touches a personal memory, someone has to decide how much it should add and how visible its part should be. Composers working with computers faced a version of this question decades ago. In 2000, as a musicology student, I wrote about Traiettoria… deviata, a piece by Marco Stroppa for piano and computer, and about how the two share the stage. Stroppa found three distinct ways for a pianist and a machine to play together. Looking back at that work now, I recognize some of the choices I face in building Savor.

A piano, a computer, and a second performer

Stroppa is an Italian composer who studied cognitive psychology, computer science and artificial intelligence at MIT in the mid-1980s. He later wrote about composing with the listener's memory in mind: which ideas the ear can hold for a few seconds, and which it keeps over a whole piece.

He wrote Traiettoria… deviata in 1982, as the first part of a longer work, Traiettoria, for piano and computer-generated sounds. The electronic sounds were synthesized in advance at the University of Padua and fixed on a recording. On paper, that makes the computer the least flexible player on stage: it can't listen, wait or react.

Stroppa worked around this in two ways. First, he added a second performer, who plays the recording into the hall, shaping its dynamics as the music unfolds and choosing which speakers it comes from, in response to the pianist. The machine's material is fixed, but how it enters the room is a human decision, made live. Second, he placed a speaker under the piano, pointed at the soundboard, so the computer's sounds make the piano's own strings vibrate.

He also made a choice that matters for everything below: the computer never processes or imitates the piano's sound. The piano stays a piano. The two instruments meet in the music itself, without one absorbing the other. From there, Stroppa brings them together in three ways.

1. Playing together: the machine enriches the piano

In the first mode, the piano and the computer strike the same pitch or chord at once. In one passage both play the same E, and the computer adds two pitches the piano's E can't produce on its own: a high A and a faint low D♭. You don't hear them as separate notes. They melt into the piano's sound.

The pianist plays hard enough that you still clearly hear a piano, but the note sounds slightly strange, richer than it should, almost like a string gone a little out of tune. The piano leads, and the machine adds what the instrument can't do alone.

A photo has a similar limit. When you take one, you're focused on the moment, not on everything around it. Later, when you want to remember it, some of that is gone. The photo is only one fragment of the memory. Savor helps you recover more: the café you were sitting in, who was with you, what the photo seems to be about, and the details you add yourself, from a note to a rating of the dish. Each fragment sits next to your picture without replacing it. Your moment still leads.

2. Framing: the machine gives the gesture its shape

In the second mode, one instrument punctuates the other's gesture. In one passage the pianist plays a broken arpeggio, and brief percussive sounds from the computer accent its beginning and its end. The arpeggio itself is untouched, but you hear its shape more clearly. Stroppa also reverses the roles: at one point the piano frames a long, sliding sound from the computer. Either way, the framing instrument changes nothing in the gesture. It only makes it stand out.

Savor does something similar with photos. On its own, a photo is a single gesture. When Savor gathers several photos into one memory, they gain a beginning and an end, and they start to read as a story. A trip goes one level further: Savor frames a set of memories, each with its own place, time and meaning, and retraces your steps on a map, day by day. Nothing inside the photos changes, but the shape around them does.

3. Blending: you can't tell who's playing

The third mode is the most striking. The piano's opening section is built around a single five-note chord (C, A♭, D, G, C♯), one of the two chords the whole work grows from. For about three minutes the pianist assembles it piece by piece: isolated notes, ornaments, intervals, arpeggios, a long crescendo. At the very end the full chord finally sounds, as loud as the piano can play it, and the pianist holds the keys down and lets it ring.

Out of that resonance, through the speaker under the piano, the computer's first sounds begin, playing the same chord. They fade in so gradually that you can't hear the moment they start. Then the computer begins to take the chord somewhere else. Later, once the two instruments play together, a trill passes from the computer to the piano and back. For a moment you hear a single instrument that doesn't exist: part piano, part machine. In music, that ambiguity is the goal, and it's where Traiettoria is most beautiful.

That's the moment I want Savor to create. The computer starts from the piano's own chord; Savor starts from your own photos. Most of them sit untouched, and the moments behind them aren't lost, only hidden. When Savor brings back the place, the people and the time of day, sometimes a memory you'd forgotten comes back with them, and you can't quite say whether you remembered it or the app did. Like the speaker under the piano, Savor works through what's already yours: it makes your own strings vibrate.

What the piece teaches

Near the start of the section where the piano and computer finally play together, Stroppa introduces the three techniques one after another, in about ten seconds: playing together, framing, blending. It's almost a lesson for the listener. Only then does he use them freely, and he keeps the blend rare. For most of the piece, the two instruments stay distinct, which is why the moments when they merge feel like magic.

Savor can learn from that. Playing together, framing and blending give the machine three roles: adding to a moment, shaping a group of moments, and bringing back what you'd forgotten. They can be introduced in the same order, starting with the roles people can easily check, like the place added to a photo, before the one that asks for more trust. And like Stroppa's blend, the most magical role works best when it's rare: a forgotten memory that comes back at the right moment feels like a gift, while one that comes back every day becomes noise.

One more lesson comes from the second performer. Traiettoria is written for two performers of equal importance: the pianist, and a musician who performs the electronic part, often Stroppa himself. That part never changes. Every sound is fixed on the recording before the concert begins. But the performer decides how it lives in the room, shaping its dynamics, choosing where each sound comes from, and following the pianist. A little like a conductor, they can change how the piece turns out. The result is a true duet: piano and machine, played by two musicians listening to each other.

Savor works the same way. It builds the memories, the trips and the collections, and you make them your own. It's a duet too: the memory is what you and the machine play together.

More stories