The instrument you play without touching it

Léon Theremin was a Russian physicist working on proximity sensors when he noticed that the pitch of a tone changed as his hand approached the circuit. He built an instrument out of the accident. It has two antennas and you touch neither: your right hand near the vertical one sets the pitch, your left hand near the horizontal loop sets the volume, and the sound is continuous between them.

That last part is the whole difficulty. A piano has keys and a guitar has frets, and both of them are a promise that if you put your finger in the right place the right note comes out. A theremin has nothing. There is no fret, no key, no edge, no place. Every pitch in between two notes is available at every moment, which is why a theremin in the hands of somebody who has not practised sounds like a ghost falling downstairs, and why in the hands of Clara Rockmore it sounds like a cello.

So the instrument that asked least of your body asked most of your ear. It became a curiosity, then a horror soundtrack, then a thing in the corner of a studio. Moog built them for decades. Their Etherwave is currently unlisted at most European retailers and sold out at the American ones, and the Claravox that replaced their flagship has gone from Moog’s own site entirely.

Not the idea. The sensing. A theremin’s antenna knows one thing about you, which is roughly how far away your hand is, and it knows it as a single number. That is why the instrument only ever had two controls.

A camera watching a hand reads twenty-one joints on it: every knuckle, every fingertip, the wrist. It can tell an open hand from a closed one, a hand turning from a hand rising, one hand from two. The same machinery reads a face, or a shoulder, or the rise and fall of a breath. None of that is new science any more — it ships inside macOS — and it is the difference between an instrument with two continuous controls and an instrument with as many as you can learn to move.

It also means the hard part of a theremin can be made optional. If the software knows which key you are in, it can decide that the space between two notes simply is not there. You keep the freedom to swoop when you want it, and lose the obligation to be perfect when you do not.

Honestly, and including the ones that cost nothing:

Browser toys. Several free pages will turn your webcam into a simple instrument in about ten seconds, with no install. They are genuinely fun and they are where most people’s search for this ends. None of them makes a sound you would keep, and none sends MIDI anywhere.

Camera-to-MIDI apps. AirBending and KAOS both watch your hands through a Mac’s camera and send MIDI to whatever you already run. They are around nine to thirteen pounds. Neither makes its own sound, so what you hear is whatever synth you already own and already know how to load — which is either exactly what you want or the reason it sits unused.

Hardware that watches you. ROLI’s Airwave puts infrared cameras on your desk and reads your hands above a MIDI keyboard, at around £299 plus the keyboard it sits behind. It is the most serious version of this idea anyone has shipped. At the time of writing ROLI’s own website is offline.

Instruments built for disabled musicians. AUMI is free, camera-based, published by a university, and has nineteen years of practice behind it. If you are looking for this because somebody cannot use a conventional instrument, start there rather than here, and start by talking to Drake Music.

An actual theremin. Still the real thing, still the hardest, and still the only one of these that will be working in fifty years without a software update.

Sonari is a Mac application in the fourth category and the second one at once: it watches you through the camera your Mac already has, and it makes the sound itself. There is nothing to plug in, nothing to wear, and no synth to own first — it ships with 10,000+ of them.

It reads hands, but also the face, the head and breath. Of its 17 ways to play, four read no pair of hands: One Hand reads one and never looks for the other, and Face, Head and Breath read none at all. It holds you inside a scale if you want it to, across 490+ of them from 8 living traditions, so the swoop is a choice rather than a hazard. It will also just send MIDI, if what you actually wanted was a controller for the rig you already have.

It is not a theremin and does not pretend to be one. What it takes from the theremin is the only part that was ever the point: that the instrument is the space in front of you, and you do not touch it.

If you want the detail rather than the story, the gesture MIDI controller page compares these properly, and what Sonari sends over MPE is written out message by message.