mitchell vitez

toggle dark mode

blog about music art media

resume email github

Speed and Velocity

The informational difference between a piano piece that moves you to tears and a piece that sounds like someone’s just trying to get to the end is actually fairly small. Let’s explore two variables that go into this kind of difference: speed and velocity.

By speed, I mean varying the tempo purposefully in small ways throughout the piece to build or release tension, or even accidentally, which might make a rendition sound more human. Speeding up while entering an exciting phrase can make it feel more powerful, and slowing down can give listeners room to take it all in. Subtle changes here make a difference. In my playing, they often make me sound like I’m just too slow at getting to the next note in time (which is generally true).

Velocity refers to changes in tone depending on how “aggressively” a note is played. It’s typically tied to loudness, but there’s certainly a difference in the sound of a soft piece played at \(x\) decibels compared to a loud piece played at the same volume. On a piano, velocity is essentially tied directly to how hard a key was pressed down, or how much force was applied.

Here’s one way an amateur (yours truly) might play a selected passage from “River Flows in You” by Yiruma.

Of course, one of the biggest variables in tone and sound quality (which both go into emotionality) is just the choice of which piano to use. Unfortunately though, that choice is often not up to the pianist. Here’s a different virtual piano, which will form our basis of comparison for varying speed and velocity. Also, from here on out, I’ll play without pedaling, to eliminate as many variables as possible in the later comparisons.

Generally my changes in timing are probably just due to a lack of touch, but some of them genuinely provide more interest. Specifically, I think a light speedup into the main upwards run of notes makes it sound better.

One operation most DAWs (Digital Audio Workstations) have built-in is temporal quantization, or making notes more exactly match the amount of time they should take up. We can easily make all notes the exact same length as notated.

Listen to what happens when we quantize to exact sixteenth notes:

Notice that there’s a stronger sense of pulse. The beat is now exact, never off by even a millisecond. However, to my hearing, there is also a slight roboticism here that’s not present in the real human playing.

What if instead we just change the velocity of each note, to make every note match in tone and volume?

I think this sounds the most amateur of all. There’s still a fair bit of tempo being all-over-the-place, but that’s now also accompanied by a lack of any differentiation between how each note should be played. Especially evident to me here is that the left hand bassline now sounds too loud; generally we like to emphasize the melody in any given piece, which here is solely in the right hand.

Finally, let’s both quantize note lengths and standardize their velocities:

We’ve only been fiddling with two variables, but this sounds inhuman compared to the first sample. It sounds like a bad demo, exact, but nonemotive.

The most surprising part of all of this, to me, is just how much of the emotional content is captured by such small changes. Essentially, individual note velocity is what stands behind a piece sounding amazing vs. just okay.

For a final comparison, here’s a professional (Yiruma himself) playing the same passage. He does have the slight advantage of being able to use the pedal, and the reasonably-sized advantage of using a better-sounding (read: actually recording a real) piano. But those are eclipsed by just having better touch, which (in my opinion rather surprisingly) seems to mostly come down to controlling note velocity.