Every gesture MIDI controller worth knowing about

A MIDI controller sends messages and produces silence. That is what the words mean, and for a keyboard it is completely fine: nobody buys a keyboard controller without already owning something for it to play. But a camera controller is usually bought by somebody who watched a video of a person moving their hands and heard music, and what arrives is a window with numbers moving in it.

So the first thing to ask about anything on this page is what happens when you move your hand and you have opened nothing else. For most of them the answer is nothing, and that is not a criticism. It is the design. If you already have a synth you love and know your way around, a small thing that does nothing but drive it is exactly the right purchase and it costs about ten pounds. If you do not, you have bought half of something, and the other half is an afternoon you had not planned on.

The second thing to ask is what it needs on your body or on your desk. A camera needs neither. A ring needs a charged ring. A glove needs gloves, and hands they fit.

Prices and details are what they were at the time of writing, and this is a category that moves.

Free browser instruments. Several pages will turn your webcam into a simple instrument in about ten seconds with nothing to install. Start here. Ten seconds costs you nothing and settles whether you like the sensation at all, which is the only question worth answering before spending money. They make simple sounds, they send MIDI nowhere, and none of them is a purchase you have to justify later.

KAOS. A Mac application that watches your hands and passes what they are doing to whatever you already run. There is a free tier and the paid one is around ten pounds. Its own line is that there are no gloves, no sensors and no pedals, which is a fair summary of the appeal. It makes no sound of its own.

AirBending. Around nine to thirteen pounds, and the only inexpensive one that runs on an iPhone and an iPad as well as a Mac. Since August 2026 it can hold an instrument inside itself, so unlike the rest of the cheap end it can now make a noise on its own. There is a fourteen day trial and a limited free tier after it. If the machine you want to play on is not a Mac, this is the answer on this page.

GECO MIDI. Around fifteen dollars, and by every account a competent piece of work. It needs a Leap Motion controller, and Leap Motion was discontinued. If you have one in a drawer already it is worth the fifteen dollars. If you do not, buying a discontinued sensor in order to run software that depends on it is the definition of a stranded purchase, and this is the one entry here to start from last.

ROLI Airwave. Around £299 for a desk unit with infrared cameras in it that reads your hands above a MIDI keyboard you supply. It is the most serious version of this idea anybody has shipped, and keeping a real keyboard under your hands answers the hardest problem in the whole category in the most direct way available. At the time of writing ROLI’s own website returns an error page, which is worth knowing before buying hardware from them. There is more on that, and on what to do with one you already own, on the ROLI page.

Genki Wave. Around $349 for a ring worn on a finger, sending its motion over Bluetooth. Not a camera at all, and that is the point of it: it works on a dark stage, it works with your back to the room, and it works while your hands are busy holding an instrument you already play. If you are a guitarist who wants one hand free to shape a sound without putting the guitar down, this is the answer and nothing else here is close.

MiMU Gloves. $1,499 for one and $2,899 for a pair, with their Glover software at $149 on top. This is the top of the field, made by people who perform with them, and Glover already takes an iPhone, a Leap Motion and a Wave as inputs, so it is where several of the things above eventually point anyway. If playing this way is going to be your practice for the next ten years rather than your next fortnight, this is the serious answer and everything else is a way of finding out whether you meant it.

Worth knowing before you spend anything, because it is true of every camera option here including this one.

It has to see you. A camera needs light on your hands and it needs them in shot, and a bright window behind you is worse than a dim room. Sonari asks for a hand held open about half a metre away, roughly at chest height, palm towards the screen, and draws an outline over it the moment it has you. When the light is against you it says what to do about it rather than naming the fault: turn on a light, turn to face the window, move back a little so the whole hand is in view.

It cannot feel anything. There is no key travelling under your finger and nothing pushing back, so how hard you meant a note has to be read off how you moved. Sonari’s answer differs by instrument: in Drum Pad how fast you swing is how hard it hits, and elsewhere a note lands as firmly as your hand was already closing when it landed. Its iPhone companion exists to supply the tilt, the breath and the thumb that a camera cannot see. None of that is a key under your hand, and if you play keys for a living you will notice inside a minute.

It only looks at you thirty times a second. That is what a Mac’s built-in camera does and it cannot be asked to hurry, so there is one look every thirty three milliseconds and nothing downstream can be quicker than the thing it is waiting for. It is the floor under every camera instrument on this page. What matters more than the figure is that the wait never wanders: a delay that is always the same is something you play around inside a minute, and a shorter one that varies is not.

Reading the hand has to be quick too, and that is a hardware question. It takes Sonari about three milliseconds on a Mac with Apple silicon and about twenty five on a machine without it, and the slower figure alone is more than an instrument can spend before a note starts. That gap is the entire reason there is no version for Intel Macs: it would install and be unplayable, which is worse than not offering one.

And your arms are holding themselves up. No keyboard has ever asked that of anybody. Most people meet it on the second afternoon rather than the first.

Sonari is a Mac application that does both halves. It watches you through the camera the Mac already has, and it carries its own synthesizer, 64 voices and 10,000+ instruments, so moving your hand makes a sound before you have opened anything else. Nothing is worn, nothing is plugged in, and there is no synth to own first.

It is a controller as well when you want one. It turns up in Logic Pro, Ableton Live, GarageBand and anything else that takes MIDI, with nothing to install and no driver. One behaviour surprises people and is deliberate: when another music application is leading, Sonari goes quiet rather than playing over it, and it takes one press to have both at once.

What the camera buys you here is range rather than one more control. 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. Two of them do not need the camera in the first place, so a dark room is not the end of the evening: Key Deck is played on the Mac’s own keys, and Breath is hummed into an iPhone. And one of them, Expression, plays no notes by design and does nothing but colour something that is already sounding, which is the exact job the cheap end of this page is bought for.

It can also hold you inside a key while you do it, across 490+ scales from 8 living traditions. No controller on this page does that, because a controller has no idea what you are playing.

Plainly, because a comparison that never says this is not a comparison.

If you already have a rig you love. You own the sounds, you know where everything is, and what you actually want is a pair of hands over it. KAOS or AirBending does that for about ten pounds. Sonari costs a great deal more, and most of what you would be paying for is the half you already have.

If you want it on a track in your session. Sonari is not a plugin. It stands beside your session as its own instrument and plays into it, which is a different shape of thing from something you insert on a channel and forget about. If that was what you had in mind, this is not it.

If the machine is an iPad or a phone. Sonari Remote is half an instrument: the sound is made on the Mac, and the phone needs one on the same network to play at all. AirBending runs on iOS by itself.

If your Mac has an Intel processor. It will not run. Sonari needs macOS 26 and Apple silicon, and the reason is the three milliseconds above rather than a policy.

If you play on a dark stage, or with your hands full. A camera is the wrong sensor for both and a ring is the right one, so buy the Wave.

If this is going to be your life’s work. Buy the gloves. They cost what a second-hand car costs and they are worth it to the people who wear them.

If you came to this from the wider question rather than from a shopping list, the instruments you play without touching page starts further back and includes the theremin. If you want the messages rather than the argument, what Sonari sends over MPE is written out one at a time, and Sonari with your DAW covers what actually happens in Logic Pro, Ableton Live and the rest.