DJ Hands

Play chords in the air, in front of your webcam. Your left hand picks the chord, your right hand shapes it.

Left hand
  • 1–5 chord
  • silence
Right hand
  • volume
  • filter
  • 480+chords12 roots, 40 qualities, every inversion
  • 25on tap5 sections × 5 fingers, switched mid-song
  • ADSRreal synth3 oscillators, 4 waveforms, 3 filters, 6 effects
  • 0to installno signup, no MIDI gear, no upload

Not a toy — a synth you actually build:

  • Shape the voice. 3 oscillators over 4 waveforms, detuned against each other, under an ADSR envelope you draw by hand.
  • Play the filter. Turning your palm sweeps a lowpass, highpass or bandpass, live.
  • Drench it. 6 effects — bitcrusher, chorus, tremolo, phaser, delay, reverb — reorderable, and lockable to a tempo.
  • Make it a pattern. An arpeggiator walks any chord you hold — 5 patterns, up to 3 octaves, locked to the tempo or free.
  • Land it on the beat. A metronome clicks and blinks the tempo, and a chord change can wait for the next quarter, half or whole bar.
  • Rewrite every chord. Inversions, slash bass, an octave shift per finger.
  • Keep it — and send it. Everything saves as you go, and what you are playing can be named, kept as a song and copied into somebody else's browser.

Needs webcam permission. Video is processed entirely on your device — never recorded, never uploaded.

Inspired by gesture-synth — respect to the original.

Built by Yusif Aliyev · Music as Joe in the Studio · Source on GitHub

What it is, how it works, questions

A musical instrument you play with your hands, in the air

DJ Hands is a free online synthesizer you play with hand gestures in front of your webcam. There is nothing to download, nothing to install, no MIDI keyboard and no account — open the page, allow the camera, and hold up your hands.

Your left hand plays the harmony: one to five fingers picks one of five chords, held for as long as you hold the shape — or arpeggiated, one note at a time, if you have that switched on — and a closed fist is silence. Your right hand shapes it — raise it to get louder, turn your palm to sweep the filter open or shut, and hold up fingers to jump between song sections. It is closer to conducting than to typing.

Everything that makes sound runs on your own machine. The hand tracking and the synth are both browser code, and no frame of video is ever recorded, uploaded or sent anywhere — each one goes from the camera straight to the tracker and is discarded.

How hand tracking becomes music

  1. The webcam watches your hands

    A hand-tracking model finds 21 landmarks on each hand, sixty times a second, entirely inside the browser tab.

  2. Landmarks become gestures

    Those points are read as a finger count and a palm angle — how many fingers you are holding up, and which way your hand is turned.

  3. Gestures play a synth

    The count picks a chord, the height of your hand sets its volume, and the angle sweeps the filter. A Web Audio synth turns all of it into sound with no perceptible delay — or holds the chord back to the next beat, if you have asked it to land on the grid.

Computer vision by MediaPipe, sound by Tone.js over the Web Audio API. The tracking model and its runtime are served from this domain rather than a CDN, so nothing the instrument needs is fetched from a third party while you play.

Not a toy — a synth you actually build

Chords

480 of them — 12 roots across 40 qualities, from plain triads to thirteenths and altered dominants. Any slot takes an inversion, a slash bass and an octave shift of its own.

Song sections

5 named sections, each holding its own 5 chords. Your right hand switches between them mid-performance, so a verse and a chorus are a gesture apart rather than a menu apart.

The voice

3 oscillators, each on any of 4 waveforms — sine, triangle, square, sawtooth — balanced against each other and detuned by as much or as little as you want, under an ADSR envelope you draw by dragging its corners. It is a real synth voice, not a sample.

An arpeggiator

Hold a chord and it is played one note at a time instead of all at once — 5 patterns, up to 3 octaves of it, and a gate that runs the notes from staccato to legato. The rate is free or locked to the tempo, and every new chord starts its pattern from the beginning, so the rhythm follows your hand.

A clock to play to

The instrument keeps its own beat from the moment you start — 40 to 240 BPM, clicked by a metronome and counted out by 4 lamps along the bottom of the screen. A chord change can be made to wait for the next quarter, half or whole bar, so a progression lands in time however loose the gesture was.

A filter you perform

3 filter types — lowpass, highpass and bandpass — swept by the rotation of your right palm. Turning your hand opens and closes the sound in real time, the way a knob would.

An effects rack

6 effects — bitcrusher, chorus, tremolo, phaser, delay and reverb — each with its own depth, reorderable along the signal chain, and the timed ones lockable to a tempo grid.

It remembers

Every chord, envelope, filter range and effect setting saves to your browser as you go. Close the tab and the instrument you built is the one waiting next time.

Songs you can send

Name what you are playing and it is kept as a song — up to 24 of them — holding the sections, the voice, the arpeggiator, the filter, the effects and the tempo. Copy one to your clipboard and it will open in anybody else's browser, because a song is text and carries nothing about your camera.

Questions

Do I need a MIDI keyboard or any music gear?
No. A webcam is the only input device. There is nothing to plug in, nothing to install, and no account to make — open the page, allow the camera, and your hands are the controller.
Is my webcam video uploaded anywhere?
Never. Hand tracking and sound both run inside your browser, on your own machine. Frames go from the camera to the tracker and are discarded — there is nowhere for video to be sent, nothing is recorded, and no image ever leaves your device.
Does the site collect any data about me?
Only anonymous usage analytics — Google Analytics records that the page was opened and which features were reached for, such as a chord played or an effect turned up. It never sees your camera or your audio, and the instrument you build is not part of it: chords, envelope, filter and effect settings, and any songs you save, are kept in your own browser and are not sent anywhere unless you copy a song out yourself. There are no accounts and no advertising.
Is DJ Hands free?
Yes, completely. No ads, no accounts, no paid tier, and no trial. It is open source under the Apache 2.0 licence.
Do I need to know music theory to play it?
No. Every chord slot is already filled with a chord that sounds good with the others, so holding up fingers plays music straight away. The theory is there if you want it — inversions, slash bass and forty chord qualities are all editable — but you can ignore all of it and still play.
Which browsers does it work in?
Current Chrome, Safari and Firefox on desktop. It needs WebGL, a webcam, and a secure origin, since browsers refuse camera access over plain HTTP.
Does it work on a phone or tablet?
It runs, but it is built for a desktop or laptop webcam. Hold the device in landscape and stand back far enough for both hands to be in frame — in a small frame the difference between four and five fingers gets unreliable.
Why is my hand not being detected?
Almost always light or framing. Put a light in front of you rather than behind you, keep your whole hand inside the frame with fingers spread, and check that another app is not already holding the camera. If the tracker never starts at all, WebGL is usually disabled in the browser.
My left and right hands are reversed. Can I swap them?
Yes. Some cameras and virtual-camera drivers hand the browser an already-mirrored image, which flips the handedness the tracker reports. Turn on Swap hands under Tracking in the settings panel and the two hands change places.
Can it play arpeggios on its own?
Yes. Turn the arpeggiator on and a chord you hold is played one note at a time instead of all at once, for as long as you hold the shape. There are five patterns — up, down, up and down, down and up, and random — a span of up to three octaves, and a gate that runs the notes from staccato to legato. The rate is either free or locked to the tempo, and each new chord starts its pattern from the beginning, so the rhythm follows your hand rather than the other way round.
Does it have a metronome, or a way to keep me in time?
Both. The instrument keeps a beat of its own from the moment you press Start, anywhere from 40 to 240 BPM, counted out by four lamps on the bar at the bottom of the screen — and you can switch the metronome on to hear it as well as see it, a click on every beat with the downbeat a little higher. You can also hand the timing to the machine: set the grid to a quarter, a half or a whole bar and a chord change waits for the next mark instead of playing the instant your hand moves, so a progression lands in time however loose the gesture was. The click and the grid both start switched off, so nothing keeps time for you until you ask it to.
Can I save what I make and share it with someone?
Yes. Name what you are playing and it is kept as a song — up to twenty-four of them — holding the sections and their chords, the voice, the arpeggiator, the filter, the effects and the tempo. Copying a song puts it on your clipboard as a block of text you can paste into a message, and whoever pastes it into their own Songs panel hears exactly what you were playing. A song carries nothing about your camera or your tracking, so it arrives set up for their hands rather than yours.
Can I record what I play?
Not from inside the app. There is no recorder built in, so capture it the way you would any browser audio — a screen recorder, or routing the tab into a DAW with a virtual audio device.
How is this different from a regular software synth?
Under the sound it is a regular software synth: three oscillators over four waveforms, detuned against each other, an ADSR envelope you draw by hand, a resonant filter, an arpeggiator, a clock with a metronome and a quantize grid, and a rack of six effects. What changes is the controller. Instead of a keyboard and knobs, chords come from how many fingers you hold up, volume from how high your hand is, and the filter sweep from the angle of your palm — so the gestures a performer already makes are the ones that shape the sound.