Syntho UI Kit

Drag a handle for frequency and gain, wheel over one for Q, or tab to it and use the arrows — Alt+Left/Right (or [ and ]) for Q, Shift for a fine pass. In bars mode the columns stand at their own frequencies, so the curve behind them is the curve they make.

Tap a cell to toggle it, drag across a row to paint a stroke, drag up or down on a lit cell to ride its velocity (Shift is fine). Velocity is fill height and brightness both; gate is the fill's width. The playhead is a prop — this demo owns the clock, the grid just draws it — and the page follows it.

E(3,8) — gaps 3-3-2. Euclid spreads the pulses as evenly as whole steps allow, which is why the gaps only ever take two values one step apart, and why the defaults here are E(3,8): the 3-3-2 tresillo, a rhythm that already existed long before Bjorklund wrote the algorithm down. Rotate and the polygon spins without changing shape — the pattern is a necklace. The playhead is a prop: this demo owns the clock and advances a fractional step position at 2 steps per second, so the hand sweeps smoothly between nodes and the highlight snaps to the nearest one.

The demo owns the clock — an interval integrating beats, which is why dragging the tempo speeds the transport up from where it is instead of rewriting where it has been. The component only ever receives that position as a prop and pulses its light from it, the same rule lfo-visualizer and step-grid follow. Drag the number to set the tempo (Shift for tenths, double-click to reset), or tap the key four times: tap tempo here is pure maths over a list of timestamps, median-averaged so one fumbled tap does not move it. Swing lives in the same lib, built and tested with no face passing it yet — the sequencer that wants it is still coming-soon, the same call step-pattern made about rotateSteps.

Press a key: the velocity comes from how far down the key you land, front loudest, with a floor so an edge tap is quiet rather than silent. Drag along the keys to glissando — the previous note is released as the next is struck — or play from the QWERTY row (space panics, − and = shift the octave). This demo is a scale-locked instrument: strikes outside the scale are declined, and a declined key draws struck-but-silent rather than dark, because a key that produces nothing is otherwise indistinguishable from a broken one. No audio here: the kit emits notes, hosts make the sound.

The shaded strip is the knee — the soft in soft-knee, drawn as a width rather than implied by a number. The meter is fed a level sweeping through the threshold, read through this same curve.

The dashed diagonal is the signal untouched; the gap to the curve is the distortion, at true scale. The bright stretch is the part of the curve a swelling tone is exercising, measured from real samples.

11 audible repeats above -60 dB, a 2.89 s tail. Height is level and so is opacity: by the fifth repeat a pip is two units tall, and one cue is not enough to read a tail by. The hollow pip at zero is the dry hit, whose own height is 1 - mix, because the stage crossfades rather than adding a send.

Four chips, four different claims. Audition and Host sync toggle between lit and off — the facility is there either way. Output tap is the one that matters: when the host publishes no tap it goes dim, not off, because "this host cannot tell me" is a different statement from "measurement is switched off", and a face that draws them the same has told you something false about your session. Source sits on warn: engaged, but not the source you asked for. Only live pulses.

Untick the tap and the first readout becomes an em dash with a note saying why — not 0.0 dB, which would be a measurement nobody made. A real zero still prints as zero; that negative case is pinned in the test, because confusing a measured 0.0 with "no reading" is the same bug wearing the other hat. Push the reduction past the budget and the value alone takes the warning accent, while the placeholder readout on the right never lights up at all. Figures are tabular, so the number does not reflow as its digits change.

The two x columns are the point: the beat grid's xForBeat and the piano roll's are the same numbers, because they are the same function — the grid builds on createLaneLayout rather than growing a second time axis, and the test asserts the agreement at every subdivision. The dB column runs the other way: 0 dB at the top of the plot, reduction downward, and the axis floor labelled −∞ because it is the bottom of what the plot can show rather than a value anyone measured.

Free-running at 120 BPM — a plausible illustration, not the host's position. Tick Host transport and the phase starts coming from a stand-in DAW's ppqPosition; untick Transport playing and it falls back to free-running, because a face frozen on the host's last position looks broken — but note it keeps the host's tempo, since the tempo is real even when the position is not. The source field exists so a caption can state which of those is true instead of implying a sync it does not have.