Steps

Steps makes no sound. Like the arpeggiator it is a note filter: it sits between your hands and whatever voice is mounted, holds the notes you press, and fires them on a grid. The difference is what the grid says. The arpeggiator fires every step; Steps consults a sixteen-slot pattern first, and each slot answers with a velocity, 0.9 is an accent, 0.45 is a ghost note, zero is a rest. The note that fires is still the walk's choice (up, down, up-down, order played, or the seeded random), so the two halves of the name are literal: the sequencer decides WHEN and HOW HARD, the arpeggiator decides WHICH.

The step's velocity multiplies the held note's, so playing softly makes the whole pattern softer without flattening the accents, the pattern is a shape imposed on your dynamics, not a replacement for them.

Swing is the module's headline knob and it is the kit's MPC convention: 50% is straight, every second slot is delayed toward its pair partner, and 66.7% lands the pair exactly on triplet positions. The default pattern deliberately puts half its hits on the odd slots so the knob is audible the moment the page loads.

The whole pattern is ordinary params, steps.v1 through steps.v16, which is not an implementation detail but the feature: the fallback panel can edit every slot, and a share link or preset carries the groove with no bespoke serialization at all.

This is a preview module. The scheduling is real, every note is committed to the AudioContext clock as an absolute time, so timer jitter cannot move it, but the voice it drives is the website's stand-in engine, not a compiled kernel, and no parity claim applies.