Field manual
Six voices on a step grid, each with a place in the space around your head. This page covers every control on the machine, top to bottom. It runs entirely in your browser — nothing is uploaded.
A pattern is already loaded. Everything else you do is done with the loop running.
Each tap cycles a step through normal, accent and ghost. Tap a track name to mute it.
Headphones on, press ◎ Spatial, and the hats start circling your head.
Orientation
One column, stacked in the order you tend to use it: transport at the top, then the pattern bank, the grid, the per-track mix, the loop station, and the ways out at the bottom.
Top row
Starts and stops the sequencer. Browsers will not make a sound until you click something, so this is also what brings the audio engine up.
Everything else on the page is live while it runs. There is no need to stop to edit.
The clock for the whole page, the loop station included.
Delays every second sixteenth, pushing the pattern from straight toward a shuffle. Around 15% is a lilt; past 40% it is properly triplet-ish.
Swing is part of the beat, not an effect on top — the loop station quantises to the same swung lap, so recorded loops stay locked to it.
Steps per lap. 16 is one bar of sixteenths, 32 is two. Switching to 16 does not erase the second half — go back to 32 and it is still there.
Starting patterns. They overwrite the current pattern, so copy it to a spare slot first if you want to keep it.
Empties the current pattern. Tempo, swing, levels and positions stay as they are.
The master switch for placement. Off, the six voices are a plain stereo kit. On, each one runs through the binaural engine at its own angle, the per-track spatial controls appear, and the radar opens beside the mixer.
Second row
Four patterns, and a four-bar arrangement built from them. This is how a loop becomes something with a verse and a fill.
Four independent patterns. Tap to switch — it takes effect immediately, mid-lap.Long-press or right-click to copy the current pattern into that slot, which is the fast way to make a variation: copy A to B, then break B.
The slots hold steps and probabilities. Levels, samples, positions, tempo and swing are shared across all four.
Turns the four slots underneath into an arrangement — one lap each, in order, looping round. Tap a slot to cycle which pattern that bar plays.
A A A B is the whole trick: three bars of the groove and a fill on the fourth.
The chain drives the export too, so what you hear is what the WAV contains.
Lets a closed hat cut the open hat’s ring short, the way one physical hi-hat cannot be both at once. On by default; turn it off if you want the open hat to ring through.
The grid
One row per voice, one cell per step, with every fourth column marked so you can see the beat. The lit column is the playhead.
What a tap cycles through
Silent.
The ordinary hit.
Full velocity — louder and, on the synth voices, brighter.
A quiet hit at about half level. What makes a pattern breathe instead of march.
Probability is the one worth learning
Long-press a cell and it shows a number: how often that step actually fires. A hat at 75% plays most bars and leaves a hole in the rest; a ghost snare at 25% turns up just often enough to sound played. It is the difference between a loop and a performance, and it costs one long-press per step.
Everything is reachable by tap
Nothing here needs a right-click or a keyboard. Long-press does everything right-click does, so the whole machine works on a phone.
Below the grid
Six synthesised voices — no sample files to download, and no wait on load. Each row is one voice: level, an optional sample of your own, and where it sits.
Pitched sine with a fast drop. Front and centre by default — the anchor everything else is placed around.
Tone plus filtered noise. Also front and centre.
Noise burst with the classic stutter. Sits slightly right.
Short filtered noise. Orbits by default, which is the quickest way to hear the whole idea.
The same voice, left to ring. Orbits from behind you, and gets cut short by the closed hat when Choke is on.
Pitched sine, slower drop than the kick. Off to the rear right.
Replaces that voice with your own one-shot. It keeps the row’s steps, level and position — only the sound changes. The ✕ beside it goes back to the synth voice.
Samples stay on your machine. They are not part of a share link, and they are not stored in a project — reload and the row is back on its synth voice.
Per track, with Spatial on
The part no other drum machine has: each voice gets its own place in the space around your head, or its own orbit.
The voice sits at one angle and stays there. The slider beside it sweeps from −180° to +180°, with 0° dead ahead and ±180° directly behind you.
Kick and snare start dead ahead on purpose. Moving the anchor of a beat off-centre is disorientating in a way that moving a hat is not.
The voice circles you continuously. The slider sets seconds per lap, 0.5 to 8. Slow laps read as the room moving; fast ones as an effect.
Two voices on slightly different lap times drift in and out of phase with each other, which never quite repeats.
The round display beside the mixer, with you in the middle facing up. Each voice is a labelled dot that flashes when it fires. It is drawn from the same maths the scheduler uses, so what you see is where you are hearing it.
Why hits land exactly on the angle
A drum hit is a transient — a few milliseconds where a continuously moving panner would be approximately somewhere. Instead, each voice’s position is set at the hit’s scheduled time, so a hat on an orbit lands at precisely the right angle every time, live and in the export alike.
Wear headphones
Placement is binaural — it works by recreating the timing and tonal differences between your two ears. Over speakers most of it collapses. On headphones a hat genuinely passes behind your head.
Below the tracks
Four slots that capture whole laps in time with the machine — the drum bus, or your microphone.
What the slots record. Drums captures the machine’s own output; Mic switches to your microphone and asks permission the first time.
Mic capture is latency-compensated: what the microphone hears lags the beat by the round trip out to your speakers and back, and the capture window is shifted to match.
Tap to arm. Recording starts at the top of the next lap, not the moment you tapped, so you have a whole bar to get ready. Tap again while armed to cancel.
Arming does nothing useful while stopped — the machine tells you so and waits for Play.
The ring shows the loop’s position in its lap. Tap to mute or unmute.Long-press or right-click to arm it again and overdub another layer on top.
The slider underneath sets its level; ✕ empties it.
Loops stay locked to their lap
Every slot is quantised to the lap it recorded against, so layers stay in phase with each other and with the pattern indefinitely — no drift to nudge back into line.
The three buttons
Renders the loop offline and downloads drum-loop.wav. Two laps of the current pattern normally; with Chain on, one lap per chain slot, so the file is the whole arrangement.
It renders extra tail for the decays and wraps it back onto the start, so the file is exactly the musical length and loops with no gap and no clipped cymbal.
The bassline and every unmuted loop-station slot go in with it, so the file is what you were hearing rather than the grid on its own. Mute a slot to leave it out.
Renders the same bounce and drops it straight onto a deck in the DJ studio — no file to save and re-open. The loop travels as WAV bytes in your browser's own storage; nothing is uploaded.
It lands like any other track: waveform, tempo and key detection all run on it, so you can beatmatch your own loop against a record.
Packs the patterns, chain, mix and spatial settings into a URL and copies it. Whoever opens it gets your groove, on the synth voices.
Samples and recorded loops are not in the link. Opening one overrides that browser’s saved session.
Renders the loop and carries it straight to the8D converter, already loaded. There the whole groove can ride an orbit of its own — placement inside the kit, then placement of the kit.
Bottom of the page
Named snapshots, kept in this browser. Your current work autosaves separately and comes back on your next visit, so projects are for keeping more than one idea rather than for not losing this one.
Downloads the project as a file, and loads one back. This is the only version that survives a cleared browser or moves to another machine.
Worth knowing
Things that will surprise you if nobody says them first.
Samples are not in the share link. A link carries the patterns, the mix and the spatial settings — everything but audio. Anyone opening it gets your groove on the synth voices. Send the exported WAV if you need them to hear your samples.
Loops live only in the tab. The four loop-station slots are held in memory and are not part of a project, a share link or the autosave. Reloading clears them.
Export is two laps, or the whole chain. With Chain off you get two passes of the active pattern. With Chain on you get one lap per chain slot, so the file is the arrangement rather than the pattern.
Every render rolls the dice again. Probability is re-rolled per export, so two exports of the same pattern are two different takes. That is deliberate — export twice and keep the better one.
Spatial needs headphones. Positions are binaural. Over speakers the angles largely collapse and you are left with a slightly odd stereo mix. The radar only appears when Spatial is on, for the same reason.
Mic recording depends on your device. Latency compensation uses the numbers the browser reports. They are usually right; on some Bluetooth and USB setups they are not, and loops can land slightly early or late.
Nothing is uploaded, ever. Samples are decoded in this browser, and the render happens on your own processor. Projects live in this browser’s storage — clearing site data clears them.