Theremin
Play music with thin air
Theremin turns the front camera into an instrument you never touch. Vision tracks your fingertip, height sets the pitch, sideways sets the volume, and a synth built from scratch turns that into sound in real time. It is a toy, on purpose.
- Coming to the App Store
- Front camera, no touch
- 10
The one idea
A real theremin is played by moving your hands near two antennas, without touching anything. Every phone already has the antenna: a camera that can see your hand. So the app is that, and nothing else. Open it, hold up a finger, and you are playing.
Everything past that is there to make the noise more fun: four instruments, from the classic eerie wail to an 8-bit square wave; Auto-Tune that snaps you to a scale in any key, Hard for the robot effect or Smooth for a glide; reverb and a rhythmic pulse; and Duet Mode, where each hand gets its own sound and settings.
How it works
The camera feeds `VNDetectHumanHandPoseRequest`, one or two hands per frame. Frames reach Vision unmirrored, so it can tell a left hand from a right one, and are flipped afterwards to match the selfie preview. Muting is a curl: index finger length over palm length, smoothed, with a dead zone and a two-frame confirmation so a wobble does not cut a note.
The sound is an `AVAudioEngine` source node per hand, rendering every sample itself: glides toward each target, a sine stack, detuned PolyBLEP saws through a low-pass, FM, and a PolyBLEP square, all soft-limited. Settings are read fresh every frame, and the live overlay is the only view that watches the tracker, so the settings sheet never rebuilds sixty times a second.
Where it stands
Finished and on its way to the App Store, in ten languages, on iPhone and iPad and on Apple silicon Macs. Hand tracking happens entirely on the device. Nothing is recorded, uploaded or collected, and there is no third-party code in the app.
