Phrase Step
User manual

Touch layout

The touch layout gives the lanes the full screen width and keeps the piano at the bottom. Open Channel editor beside Settings to edit Sounds, Notes, Step Actions, Timing or CC. The panel stays open while you play the on-screen piano or an external MIDI keyboard. Close it with × to see the full sequence again.

Channel tools above the steps holds Copy, Paste, Clear, Rotate, Double and step-selection tools for the selected channel. Randomize stays on each channel’s dice; the duplicate bottom action row is removed.

Settings → Layout switches between Touch, Desktop and Auto. Touch is the initial layout; Desktop restores the docked editor. MIDI connections are in Settings. Channel sublanes also open in a separate panel in Touch layout, keeping all eight channel rows in place. This is the first touch preview; physical iPad testing and further control refinements are still planned.

1. Start making sound

Phrase Step is an eight-lane step sequencer with built-in FieldSynth sounds and optional MIDI input/output. You can make a complete arrangement in the browser without connecting another instrument. To turn those patterns into a playable field, follow Grow and play a sound canvas.

1. New users start with Sparrow’s Dream, a saved factory song. Existing autosaves are preserved. Presets also includes Me-ander (latch 2+ notes) and Wonderland. You can restore the starter from Load demo → Load Sparrow’s Dream · starter song. Or open Load demo and choose Sunbeam for a bright two-lane example, or Daybreak for three interlocking parts.
2. Press Play. A user gesture starts the browser’s audio system; if sound was explicitly turned off, use Enable sound first.
3. Find the moving playheads. Each lane advances independently: they do not have to reach their first steps together.
4. Click a lane’s Voice button to browse sounds. Use its ♪ preview button to hear a candidate before choosing it.
5. Open Preset Menu, give the session a name, and choose Save song.

Sound comes from your browser’s current audio output. The Volume control changes the built-in synth’s output; it does not change MIDI note velocity or an external synthesizer’s volume.

Loading a demo replaces the working session, but Undo restores it. The five demos are Sunbeam, Daybreak, Afterglow, Glasshouse and Low Orbit.

2. Build your first phrase

A step is one position in a lane. It can contain a note, a chord, a rest, effects and MIDI controller values. A lane repeats its steps at its own speed. A phrase holds all eight lanes.

1. Select a lane by clicking its name, number, controls or empty track area. Use its Clear action if you want to remove its notes; this clears the selected lane’s steps, so use Undo if it contained material you wanted.
2. Click steps 1, 5, 9 and 13. Clicking an empty step adds the lane’s default note. Clicking an occupied step selects it.
3. In Notes, change the selected note’s pitch and velocity. You can also click the on-screen keyboard in Single note entry mode.
4. Choose a pluck or bass sound from Voice, then press Play.
5. Set a second lane to 12 steps and add a few notes. Its restart will drift against the first lane’s 16-step phrase.
6. Open the second lane’s speed button and choose 1/8. Each step now lasts twice as long as a 1/16 step.

Make one change at a time. Short phrases with two or three contrasting sounds often reveal the timing features more clearly than filling all eight lanes.

3. Find your way around

Choose Init at the top of Presets for an empty starting point: eight empty 16-step lanes with Sawtooth, Sine, Triangle, Square, Pad, Kick and two IDM percussion voices; 80% lane volume, 120 BPM and C Major. The pad starts with a longer note gate, and the kick uses a low root note. Loading Init leaves your named song library intact; Sparrow’s Dream is the factory startup song.

The right-hand editor heading shows the channel number in its lane color followed by the voice name, for example 01 Soft Chime. The Presets dropdown shows the active factory or custom preset name, without a duplicate name field beside it. Name a custom preset in Preset Menu. The top of the screen has two rows: transport and sound controls above, and Keyboard Sequence, Phrases, MIDI and Settings below. The note-input menu sits beside Record. The Phrases card has icon buttons for new, duplicate, copy, paste and rename. Hover a control for a short hint; hints are on by default and can be turned off in Settings → Parameter tooltips. Settings also shows the build identifier.

Preset Menu is immediately right of Presets. The theme switch is immediately right of Randomize in the top bar. The dark-theme playhead is filled with its channel color, matching the light theme. The default theme is light. The moon button in the top bar switches to dark and the sun switches back; the choice is remembered on this device and does not change the song. Sliders, numeric values and envelope nodes respond while you drag, and releasing the drag saves one Undo action. The Sounds panel shows the amp Attack, Decay, Sustain and Release controls next to the Amp Envelope graph: drag its nodes, or click ↗ for a larger graph. Mod Envelope opens in its own popup. Voice → Lane options also contains the amp graph.

The synth has eight channels. Channels 6–8 start folded, enlarging the buttons, controls and step tiles on channels 1–5. Click the thick chevron marked 6–8 below the lanes to show or hide them. Folded channels keep playing, and their notes and settings stay intact. The main grid shows 16 positions at a time, with shared pages for steps 1–16, 17–32, 33–48 and 49–64.

• Top bar: Presets, Preset Menu, save, Undo/Redo, Deterministic, keyboard visibility, demos, JSON Export/Import, global Randomize and the Manual link at the far right.
• Transport: Play/Stop, Record, entry arrows, Panic, master tempo, swing and key/scale.
• Utility groups: Keyboard Sequence, Phrases, MIDI and Settings. Sound sits with playback. Humanize sits with tempo/key.
• Keyboard Sequence: On/Off, Quantize, Multi Division and Latch stay on the left; Ch Seq Order and Transpose are stacked beside the far-right Phrase Play button. Channel Stack and Voice Stack are stacked in this tile and also available in Phrase Play → Canvas Settings.
• Lane controls: name, Voice, direction, MIDI channel, mute, solo, dice, hold-to-reset ×, VOL, OCT, step count and playback speed.
• Pink strip: opens velocity, octave, note-chance and arp sublanes beneath that lane. Click it again to restore the step editor.

The selected lane and selected step determine where editing, keyboard entry and recording go. A playhead shows playback position; a selected step shows your editing target. They can be different.

4. Click, drag and reset

Double-clicking notes is deletion, so treat a note chip differently from an ordinary setting. A continuous value drag normally becomes one Undo action when released. Tempo previews while dragging so you can hear the change immediately.

Use explicit buttons for Copy, Paste, Delete and confirmations. These are actions rather than numeric settings and do not share the value-reset gesture.

5. Lane lengths, pages and direction

Each main lane has 1–64 active steps. Change the count with its number field, up/down controls or the draggable boundary after its last active step. Only active steps play; shortening a lane keeps notes stored beyond the new boundary, so increasing its length can reveal them again.

Page buttons change which 16 positions you see across all lanes. They do not change lengths, stop other lanes, or make the whole song share one length. A lane shorter than the current page has no active steps on that page.

Follow selected lane changes pages as that lane advances. Manual page navigation switches follow off. The position readout beneath each lane still shows its current position even when its playhead is on another page.

Mute silences a lane without deleting it. If any lane is Solo, only soloed, unmuted lanes sound. The direction control is available beside the lane name and in its Timing settings.

Independent lengths are often called polymeter. For example, a 5-step lane and an 8-step lane at the same speed align their first steps every 40 ordinary steps. Weighted speed changes, arps, repeat/hold effects and jumps can change that relationship.

Every lane has visible VOL and OCT controls. Drag vertically or use arrow keys. VOL is the built-in sound level from 0–100%; 0 silences that lane’s synth without deleting notes or muting outgoing MIDI. OCT shifts the entire lane by −2 to +2 octaves, including its arps and MIDI notes. It adds to the semitone Transpose setting and octave sublane; written notes stay unchanged. Double-click OCT to return to 0. Copy/Paste lane includes the octave setting.

6. Notes, chords, rests and probability

Click an occupied step to inspect it. The right-hand editor opens on Sounds, its leftmost tab. Choose Notes to edit pitches. In Notes, select one of its note chips to edit that note. A step can store up to 16 notes; choose Chord entry · add notes beside Record, use + Note, or open Draw chord to build one. Keyboard entry defaults to Single note entry: with Record armed, each played key replaces the selected step’s notes, including any default note inserted by clicking an empty step. Chord entry deliberately adds tones instead. Chord steps have a pink border and a note-count badge. The built-in synth has an eight-note voice pool per lane, so very large or overlapping chords can steal older voices. External MIDI instruments have their own limits.

• Pitch · MIDI: the stored note number, 0–127. MIDI 60 is displayed as C4. Key, scale and transpose can change the pitch you hear.
• Velocity: how strongly the note is struck. It affects the onboard synth and emitted MIDI.
• Gate · steps: note duration relative to this lane’s current step interval. At 1/16 and 120 BPM, a step is 125 ms: gate 0.5 is about 62.5 ms; gate 2 spans about 250 ms.
• Nudge · step fraction: an early or late offset from −0.5 to +0.5 of a step. Reduce lane Quantization % to hear it.
• Audition: hold to hear the selected note. Its label shows the mapped output pitch.

Rest removes a step’s notes while keeping its other data. Double-clicking a main-grid step also removes its notes. Clear step removes the full step content, including effects and controller events. To delete one chord tone, double-click its chip or use Remove note.

Draw and edit a whole chord

Use Draw chord · all … notes in the Notes editor, tap the pink note-count badge on a step, or hold any step for about half a second to open its chord popup. Right-click also opens it. Holding an empty step opens the editor without inserting a note.

The popup lists every stored note, including notes outside its current three-octave grid. Tap a note chip’s × to remove just that tone. The pitch grid shows the sounding pitches after key, scale and lane transpose: tap an empty pitch to add it, or an active pitch to erase it. Drag from an empty pitch to paint notes; drag from an active pitch to erase them. Only the cells you touch change. Pitches outside the chosen scale are unavailable, and octave buttons move the visible range.

A step holds up to 16 notes. New notes use the lane’s default velocity and gate. Existing tones keep their velocity, gate and nudge; Step Actions, CC events and probability are preserved. Each drawing stroke is one Undo action. Edits save immediately. Use Hear chord to audition, Clear notes to remove all tones, and Done, Escape or a tap outside to close.

Two layers of note chance

Step probability controls whether the step’s notes fire. The Note chance sublane multiplies it. A step at 75% meeting a sublane value of 50% has a combined 37.5% note chance. This gate is shared by the step’s chord; it is not a separate roll for every chord tone. Position changes and CC events can still happen when the note gate fails.

7. Keyboard entry and recording

Playback & Note Input contains Record and the note-input menu. The separate Keyboard Sequence tile at the far left of the next row controls keys that launch prepared patterns. Its On/Off button does the same mode switch as choosing Keyboard Sequence in the note-input menu. See Keyboard-controlled phrases for Channel Stack, Voice Stack, Quantize, Multi Division and Latch.

The keyboard has four modes, selected beside Record:

• Single note entry: the default. With Record on, playing a key writes exactly one note into the entry step. With Record off, keys only play sound. This also replaces the default note inserted by clicking an empty step.
• Chord entry · add notes: with Record on, held pitches are added to the entry step. With Record off, keys only play sound. Use this deliberately to build a chord. Keyboard and MIDI notes never change the pattern unless Record is armed. The Record button turns red when armed. Recording follows the chosen single/chord entry mode.
• Play notes: play the selected lane without writing notes.
• Keyboard Sequence: keys launch and transpose the entire sequencer. See Keyboard-controlled phrases.

On the computer keyboard, Z lowers the playing octave and X raises it. These shortcuts are inactive while typing into text or number fields.

Record while stopped captures notes into successive entry slots. Play and release a note, then the cursor advances. The arrows beside Record let you skip a slot or return to a previous one without adding a note. They wrap within the selected lane’s length and move the visible page when needed.

Record while running writes into the selected lane at its current position. Note length comes from how long you hold the key; the recorded offset preserves where your input fell relative to that step. Lane quantization determines how much of that timing you hear.

Choose the lane before playing. MIDI input follows the selected lane for entry/recording. Record is unavailable in Keyboard Sequence mode because those keys control the phrase rather than enter notes.

Use the chevron in the keyboard’s top-right corner to collapse or reopen it. Its compact toolbar shows the note range and octave arrows. Drag the 2 oct value up or down to show one to four octaves; double-click it to restore two octaves. The −/+ arrows shift the visible range. Computer keys start from the lane’s New note pitch, which defaults to MIDI 60; those are a separate reference from the piano’s displayed octave.

Phrase Play · Mosaic and Spiral

Clusters work on Mosaic and Spiral. Open the ▦ builder, choose slot 1, 2 or 3 and click small squares to draw a shape (up to 16). Single, Line, Cross and Square are shortcuts. The small dot marks the placement anchor. Open the cluster’s Tempo grid for fixed or weighted divisions, straight/triplet/dotted feels and the change interval. The grid sets the interval between individual tile notes: at 108 BPM, 2/1 waits about 4.44 seconds between notes, regardless of shape size. Each of the three clusters has independent timing; only the master BPM is shared. Spread sets the distance between squares. Choose an order, then click Place and click the canvas. A preview highlights the tiles underneath. Several small squares over one large tile play that tile once per cycle. Empty space stays silent. The dice order picks randomly without replacement: every distinct sampled tile plays once before the next shuffled cycle starts. Each cluster has its own Loop (on by default) and Mute. Use the numbered toolbar buttons to select a cluster, then drag or press arrow keys to move it. Remove clears that placement. Shapes and settings are remembered on this device; placed clusters last for the canvas session.

Space or the canvas Stop button stops all clusters, gesture loops and pending playback. Positions are preserved. Play resumes placed clusters and a recorded mouse loop; the separate + button starts a new canvas. The Mouse loop bank records and replays canvas gestures.

Prepare your song in the editor, then click Phrase Play ↗ beside Keyboard Sequence. It opens a full-screen performance surface, turns on Keyboard Sequence with five-key playback, keeps the selected stack mode, starts sound and turns note-entry Record off. Your notes and sounds are not changed. Play a computer or MIDI keyboard: up to five root notes drive the prepared lanes. By default these are lanes 1–5. The editor’s Sequence list draws one box around the lanes in use (labelled “boxed lanes = Phrase Play”), and Ch Seq Order changes which five they are.

The single-row top bar shows the canvas name and a digital display matching the sequencer: BPM, key/scale, Degree and Transpose. These controls share the song settings. Play/Stop sits beside the display; once the canvas is sealed, + starts a new canvas. Drag the BPM to change tempo immediately, or click it and type a value from 40–300. It is the same saved tempo as the editor. If X/Y tempo modulation is active, a small XY readout shows the override. Transpose sits inside the digital display. It is the same −36…+36 semitone setting as Keyboard Sequence and accepts typing, vertical dragging, arrows and a 1/3/5/7/12-semitone step choice. Choosing a step size does not transpose anything; it sets how far the next arrow click moves. Transpose changes upcoming sequence notes and sound-canvas playback, and leaves the captured field as it is. Replay applies Transpose and Degree relative to each segment’s captured settings, so shifts are never doubled or accumulated. Pitches stay in the current key and scale.

Canvas Settings at the top holds the two canvas choices, detail quality, Channel Stack / Voice Stack, Multi Division, keyboard quantization, Latch and the replay-grid controls. Rec, Loop and clear × buttons sit in the top bar on Spiral and Mosaic and turn on once the field is playable. Master FX and Back to sequencer sit beside them. There is no lower toolbar and no separate Sound button; audio starts when you enter Phrase Play.

Spiral draws a colored Fibonacci mosaic on white. In Channel Stack, each finger uses its assigned channel’s color; Voice Stack uses five finger colors. Each finger starts with a pinhead-sized origin, then varied rectangles spread across the art area in roughly 15 seconds of active playback. When the last lines finish, or you stop playback early (including by releasing the last key with Latch off), a brief golden flash marks completion, playback stops, and Latch switches off. An early stop keeps only the shapes already drawn; it does not fill the empty space. The drawing becomes a Sound canvas.

Mosaic fills the field with graphic stamps on warm paper, or on a dark background in the dark theme. Held notes fill the field in about 15 seconds, with up to five fingers adding their channel colors. In Canvas Settings → Icon family, choose Paper Orchestra (fans, rings, folded paper, combs, blooms and stairs), Botanical or Graphic Rhythm. Changing the family redraws the same saved field without changing its sounds, and the choice is remembered on this device. Like Spiral, Mosaic ends with a golden flash when full or stopped early, then supports segment replay, Segment grid, X/Y mods and gesture loops. Switching into or out of Mosaic clears the field and any gesture loop.

Stir sits beside BPM in Mosaic and Spiral. Turn it on and drag through the field. Mosaic grabs a 3×3 group around the pointer; occasionally a corner piece slips free and stays behind as its neighbours flow around it. Spiral grabs one whole box. Whole pieces push and exchange places with neighbours, which fill the spaces they leave. Release to settle into the canvas grid: a full field stays full, without overlaps or new gaps. Spiral boxes stay whole; larger boxes move only where a compatible group can exchange places. Their sounds, pitches, colors and captured settings travel with them. Turn Stir off to play the rearranged pieces. Stirring triggers no notes or X/Y changes; existing gesture loops can continue.

Eraser, next to Stir, removes every whole piece your stroke crosses, even a large Spiral box. It leaves deliberate holes, which stirring preserves and further growth does not refill. Erased pieces no longer sound from clicks, pending grid hits or recorded loops. Stir and Eraser are separate, mutually exclusive modes. Turn either off to return to playing. Changes last for the current canvas session; New canvas clears the field.

Each segment remembers the finger’s most recent sounded note, velocity, duration and synth settings, including FM, envelopes, modulation, lane effects, tempo and master FX. A simultaneous chord from that voice stays together, up to the two-note limit per replay group. This stores settings and MIDI pitches, not recorded audio. Press or drag across rectangles to synthesize those saved sounds. Each played segment fills temporarily with the dominant color of its bordering lines while its saved note sounds. It also gets briefly bolder and gives a tiny shake (bold only with reduced motion). Isolated segments use their own line color. Canvas Settings → Segment grid offers Off, 1/2, 1/4, 1/8, 1/16, 1/32 and 1/64. It is separate from keyboard-input Quantize. Canvas replay uses the current master BPM for synced modulation and delay, even when pieces were created at another tempo. When several segments are crossed before the next grid beat, the latest one plays on that beat.

The replay voices are separate from the editable lanes. Up to four groups can overlap, with up to two notes per group (8 voices total); saved note gates are bounded to 80 milliseconds–2 seconds for drag playback. Replay has its own master FX bus; the most recently played snapshot sets that bus’s effects. Sound-canvas playback does not restart the sequencer. Panic/Escape stops its notes and pending grid triggers while preserving the drawing. New canvas clears it and unlocks keyboard playback for another run. Returning to the editor releases replay voices. Canvases currently live only for the open Phrase Play session; saving a song does not save the canvas. No audio is recorded, and there is no microphone access.

Light detail is the default; Rich detail increases drawing resolution. Reduced-motion preferences lower the drawing frame rate. Hide controls inside Canvas Settings leaves the canvas unobstructed; Show controls at the top right brings them back. Escape closes an open settings popup first; otherwise it stops playback. The small 0 button to the left of either transpose value returns it to zero semitones. Back to sequencer restores the editing view and releases canvas-replay voices and loops. Held or latched sequencer playback can continue. The choice of visual style is remembered on this device. Phrase Play uses computer or MIDI keys; the on-screen piano stays in the editor.

Phrase Play has two canvases. Channel Stack sends each finger to its assigned channel (1–5 by default); Voice Stack plays the full arrangement under each root. The visuals show discrete notes and arps; pitch bends and LFO curves are not drawn.

Heard a pop at the bottom right saves the preceding 30 seconds of audio diagnostics as a JSON download and keeps a backup on this device. Click it soon after a pop, crackle or clipping sound. The log includes sampled stereo levels before/after the limiter, sound changes, voice reuse, timing delays and current canvas settings. It contains no audio and is not uploaded. Turn it off with Canvas Settings → Audio logging. Brief transients can escape the sampled meters; the report helps investigate the noise rather than declaring a cause.

X/Y performance controls

Spiral follows the synth theme: light mode uses white paper and channel-colored lines; dark mode uses a black canvas with white-centered lines edged in each channel’s color. Played cells keep their channel-color fill in both themes.

Dragging between sound-canvas segments uses a brief fade between saved sounds to soften switching clicks. Very sharp attacks and releases receive a tiny amount of smoothing in canvas playback; the editor’s envelopes stay unchanged. Switching segments preserves the existing reverb and delay tails. The existing voice limits are unchanged.

Gesture loops: once Spiral or Mosaic becomes a playable sound canvas, click ● Rec in the top bar, or quickly tap R twice. Click and drag across segments. Choose ■ Stop in the same toolbar or double-tap R again to loop the gesture performance. Recording stores the segments you actually sounded and their timing, after the Segment grid; it never records audio. The loop length rounds up to that grid (Off keeps free timing). You can play computer/MIDI keys or more canvas segments over the loop; these extra notes are not added to the recording. ■ Loop pauses it, ▶ Loop restarts it, and × clears it. The seconds readout shows recording time or loop length. Recorded gesture loops retain their captured timing when you edit master BPM. Recording again replaces the previous loop. Escape/Panic stops it; New canvas, changing visual style, or leaving Phrase Play clears it. Loops are session-only and are not included in saved songs.

X-axis mods at the bottom center and Y-axis mods midway down the left edge of Phrase Play are clickable labels that open their assignment popups. Each shows its selected targets, or none selected when empty. They map pointer or finger position to sound. X increases left to right; Y increases bottom to top. Move over the canvas to modulate; drag over a completed sound canvas to play its stored notes at the same time. Choose up to three targets per axis. Each assignment has independent 0–100% Low/High limits and an Invert button. Drag an end handle to resize the range; drag the colored middle sideways or up/down to move both limits together. Low/High numbers are also editable. Click a selected target again, use its ×, or Clear axis to remove it. Click outside the popup or press Escape to close it. If both axes target the same control, Y takes priority.

Mappings save with the song, including autosave and export. Performance movement changes live values, not the saved notes, voice patches, tempo or master levels. Removing a mapping restores that control’s underlying value; leaving Phrase Play removes all live X/Y overrides. Canvas gesture recording remembers the X/Y position at each sounded segment hit and reapplies those positions during loop playback; it does not record raw audio or the motion between hits. The canvas remains capped at 8 voices. Canvas notes keep the amp envelope they had at note-on: moving Attack or Release changes the next triggered note, while a note already sounding keeps its envelope. This also applies when the mouse moves after you release a clicked segment.

8. Key, scale and pitch mapping

Key and scale are global musical constraints. Phrase Step stores written notes relative to C, then applies the selected key and lane transpose before snapping to the nearest allowed scale pitch. A tie chooses the lower pitch. Changing the scale changes playback; it does not rewrite the note numbers stored in the editor.

Degree: progress within your key and scale

The orange Degree button sits immediately left of Phrase Play. It opens the full progression controls. Degree shifts upcoming notes by the selected scale degree, then maps them into your selected key and scale. The key and scale stay fixed: C Major remains C Major throughout the progression, including IV and V. Notes written in the grid stay unchanged. Upcoming synth and MIDI notes, Keyboard Sequence, arps and pitch modifiers stay in the selected key and scale. Existing held notes finish at their original pitch. Playheads keep their place.

• On / Off: enable degree movement or return to unshifted playback. Choosing a progression or applying a custom one enables Degree.
• Timer: advance through the progression every 1–64 measures. The slider and numeric readout update during a drag. The initial setting is 4 measures, with Degree and Timer off. Each measure is four beats at the master tempo. The timer also runs while transport is stopped.
• Previous / Next: move one position through the progression. Option+, / Option+. (Alt on Windows) do the same without opening the popup. Manual stepping restarts the timer.
• Custom: append Roman-numeral degrees in order, including repeats, up to 16 entries. Undo last degree removes one entry; Clear empties the draft. Apply custom starts it at the first entry. Applying an empty draft gives I.
• Scale changes: degrees unavailable in the current scale are skipped, exactly as in FieldSynth. If none remain, Degree uses I. The current progression position and root are shown in the popup.
• Saving: settings, custom entries and the current progression position save with your preset and export. Automatic timer advances do not add Undo entries.

The 18 progressions are Classic Open, Pop Lift, Doo-wop Cycle, Minor-pop Gravity, Folk Cadence, Modal Lift, Mixolydian Drift, Mediant Color, Cinematic Turn, Suspended Tension, Scale Climb, Descending Pull, Modern Open, Modal Widescreen, Turnaround, Circular Turnaround, Long Pop Canon, and Full Cadence. Hover a degree or progression for its character. Degree uses the orange channel color; Ch Seq Order has a separate blue accent from the neutral Transpose controls.

Available scales are Major, Natural Minor, Dorian, Phrygian, Lydian, Mixolydian, Locrian, Harmonic Minor, Melodic Minor, Pentatonic Major, Pentatonic Minor, Blues, Whole Tone, Diminished and Chromatic. All 12 keys are available, with sharp/flat equivalents shown together.

Arpeggios, octave changes, step transposition, random pitch and scale ramps all map their output into the active key/scale. A custom arp’s chromatic offset buttons are inputs to that mapping, so two selected offsets can resolve to the same allowed pitch. Use a wider selection or a different scale if the arp has fewer distinct sounding notes than expected.

Pitch bend and Glide pitch move through allowed scale steps rather than unrestricted semitone sweeps. MIDI bends are channel-wide, so a chord uses only offsets that keep all of its notes in scale. An external synth’s own tuning, pitch effects or mismatched bend range can still change what you hear.

9. Tempo and fixed playback speed

Tempo is the shared master BPM, from 40 to 300. Drag upward/right to increase it or downward/left to decrease it; Shift gives finer adjustment. Each lane’s speed button opens its Tempo grid. Changing a lane division does not change master BPM.

A division is the duration of one lane step. At 120 BPM, one quarter-note beat lasts 500 ms:

Fixed repeats the selected division and fixed feel. A dotted feel multiplies duration by 1½; a triplet feel multiplies it by ⅔. For example, dotted 1/8 is 375 ms at 120 BPM, while triplet 1/8 is about 166.7 ms.

A lane’s approximate loop duration is active steps × step duration when speed stays fixed and there are no arps, holds, repeats or jumps. A 12-step lane at 1/8 lasts three seconds at 120 BPM.

10. Weighted timing and feel

Weighted lets a lane choose changing speeds from the Tempo grid. A weight is a relative chance, not a percentage. Only rows above zero participate. A weight of 3 next to 1/16 and 1 next to 1/8 means 75% versus 25% on each fresh draw.

When does it choose again?

Hold each draw keeps the chosen division and feel for 1–64 master beats or lane steps. Master beats are quarter notes at the master BPM. Lane steps include rests. A hold of four master beats keeps a draw for approximately one bar of 4/4; a hold of four lane steps keeps it for four lane advances, whose real duration depends on the chosen speed.

A new choice waits for the current step to finish. A very slow step is not cut short when the hold time expires. The same division can be drawn again, so a 25% choice is not guaranteed to appear exactly once in four draws.

Feel is a separate draw

The Straight / Triplet / Dotted section has its own weights. Division and feel are drawn independently and then held together. Division weights 3:1 and feel weights 4:1 create combinations of both distributions; they are not competing rows in one list.

If all division weights are zero, the Fallback division plays. If all feel weights are zero, the fallback feel applies. Clear grid zeros both sets of weights while keeping fallback and hold settings. Editing a weight enables Weighted mode. To stop variation entirely, select Fixed.

11. Velocity, octave, note chance and arp sublanes

Click the hot-pink strip between a lane’s speed control and its first step. Four sublanes (Velocity, Octave, Note chance and Arp) expand across the grid width; all eight main lanes remain visible. Each has its own 1–16-step length, independent of the main lane’s 1–64 steps and independent of the other sublanes.

Velocity sublane: the top sublane sets MIDI velocity from 0–127, with its own 1–16-step length, On/Off switch and tempo grid. Zero skips the note. Editing a value turns this lane on. It also affects triggered arp notes; a Set velocity Step Action takes precedence. Older songs keep their stored note velocities until you enable or edit this sublane.

• Octave: values from −2 to +2 shift the current written notes by octaves before final scale mapping. Default 0 leaves their octave unchanged.
• Note chance: values from 0–100% multiply the main step’s note probability. Default 100% leaves it unchanged; 0% suppresses the notes.
• Arp: values from 0–100% decide whether that main step becomes an arpeggio. Default 0% plays the written step normally.

Drag a cell up/down to edit its value. Change each row’s length with its step-count value or last-step boundary. For example, main length 16, octave length 3, note-chance length 5 and arp length 7 will keep combining differently on successive passes.

Each row has an explicit On / Off switch. Off bypasses that sublane without deleting its values; it does not switch the other rows off. An inactive octave row behaves as 0; an inactive note-chance row behaves as 100%; an inactive arp row plays the written notes.

The highlighted sublane position shows which cell was used for the current main step. A triggered arp temporarily plays its own phrase before the main lane continues.

Paint a row: press a step and sweep left or right while holding. Near the top sets higher values; dip down for lower ones. This works for velocity, octave, note chance and arp chance. Release to apply the stroke; one Undo restores all its values. A straight up/down drag still adjusts one step.

12. Arpeggios inside a step

An arp turns one main step into a short sequence of pitches. When it triggers, the main lane stays on that step while the phrase plays. Each arp slot takes a full interval at the triggering step’s timing; the arp is not squeezed into the original single-step duration.

1. Put a note or chord on a main step.
2. Open the lane’s pink strip and find the corresponding arp sublane position.
3. Tap its arp value to open the editor; tapping a zero-chance cell arms it to 100%. Long-press also opens the editor.
4. Choose Triad, Seventh, Sus2 or Sus4, or select individual pitches from the 12 pitch-offset buttons.
5. Choose the lane’s arp style, then set the step’s trigger chance.

Clicking a pitch button changes only that pitch. It does not select the range between it and the previous choice. Selected pitches are mapped into the active key/scale; repeated mapped pitches are combined. If all custom pitches are off, the custom pool is empty.

Arp timing

Every arp slot uses the triggering main step’s chosen duration. If that step is playing 1/8 and the phrase has four slots, the arp takes four 1/8 intervals. There is no separate timing mode in the arp popup. Older saved arp timing selections follow this behavior too.

Arp chance is conditional on the main step’s note gate firing. A failed arp roll plays the written notes normally; it is not a rest. If you want silence on failed rolls, use note probability instead. The hot-pink pulse identifies the originating step and the active arp sublane cell; SUB also lights hot pink during arp triggers. The editor highlights sounding arp pitches in the same hot pink.

New arps default to scale triads. A per-step chord preset or custom pitch selection takes priority; choose those directly in the arp popup.

13. Arp style reference

Style controls pitch order. Each trigger plays one complete cycle of that style through all selected pitches, then returns to the lane. For 12 distinct pitches, Stutter plays 24 slots; Up – Down plays 22. Saved Notes limits from older songs no longer truncate the cycle. There is no separate Notes limit. The lane’s style is shared by that lane’s arp steps; each step keeps its own chance and pitches.

Random styles become repeatable in Deterministic mode. Sound timbre can still contain FM sidebands or reverb modulation; an in-scale note sequence is not a promise that every frequency in a complex synth sound is a scale note.

14. Step Actions reference

Step Lock: a different sound on one step

The purple Step Lock tile is inside Playback & Note Input. Select a step, switch it on, then change that channel’s sound normally. Double-tapping Shift toggles the mode too; holding Shift for fine dragging does not. Edits go only to that step while the mode is on. Turn it off to edit the channel’s normal sound. Entering Step Lock stops playback. Keyboard notes audition this step’s sound directly, without running Keyboard Sequence or recording notes. Up/down arrows select the same step in the channel above/below, or its last step if shorter. Use left/right arrows to move between steps: each loads its own locks, or the channel defaults when unlocked. The tile shows plain ON or OFF and its color gets stronger when the selected step has locks. A ◆ also marks steps with locks in the grid. Clear, which appears only when the selected step has locks, removes them without deleting notes.

A locked sound captures oscillator/FM, amp and filter envelopes, LFO, Mod Envelope, level/pan and lane effects. Channel pitch, timing and MIDI sound settings edited in this mode are step-specific too. Each playback step resolves from the channel defaults, then its own overrides. The next unlocked step, even a rest, returns to the defaults; it does not inherit the prior lock. Whole-voice changes fade briefly and are prepared ahead when possible; very short gates may play a brief delayed hit through that transition. Tone/effects are shared by overlapping voices on the same channel. Locks copy and save with steps and phrases.

Step Lock does not capture globals or Keyboard Sequence. Channel names, MIDI routing, lane length, note editing and sublane patterns stay ordinary pattern edits. Global Randomize is unavailable in this mode; use the dice in Sounds to randomize just the locked sound. To give several steps their own random sounds at once, use the dice on the tile (see Sound randomization).

In the side editor, open Step Actions and enable the actions you want on the selected step. Each action has a direct 0–100% chance, with a draggable value and up/down arrows. Enable it first, then adjust its value and any mode selector below the grid.

Each enabled action gets its own chance check on visits to that step. A value describes what the action does; its chance describes how often it happens. For example, Repeat notes value 3 with chance 40% means “three triggers on roughly 40% of visits,” not “play the whole step 40% of the time.”

There are fourteen action categories: Pitch rise and Pitch fall are separate entries. Actions can be combined. Hold step, Repeat step, Jump step and triggered arps change when a lane returns to its beginning, so count actual lane behavior rather than assuming every pass lasts length × division.

15. Swing, nudge and lane settings

The Timing editor contains lane-level settings, even though it appears beside a selected step. Changes there affect the lane, not just that step.

Swing and nudge are measured relative to each lane’s current interval, so the same amount has a different duration at 1/4 and 1/64. Early offsets are bounded by the playback scheduler’s lookahead; very slow lanes do not allow unlimited anticipation.

Use fixed timing to establish a groove, then add a little Humanize or weighted feel. Using large swing, nudge and humanize together can obscure the underlying pulse.

16. Humanize

Humanize adds small timing and velocity variations without rewriting notes or changing the master clock. Open the global Humanize popup to enable it, then choose timing variation up to ±50 ms and velocity variation up to ±30.

Each lane’s Tempo grid also has a Humanize mode:

• Follow global: use the global enable switch and amounts.
• Lane settings: use independent timing/velocity amounts.
• Off for this lane: keep that lane exact even if the global setting is on.

Timing variation is shared by a step’s chord so its notes stay together. Velocity variation is per note. Timing is limited to a fraction of the current step so fast divisions do not receive disproportionately large offsets. The same variation applies to built-in playback and MIDI; it is not written into exported note data.

When Deterministic is on, the chosen humanize results repeat with the other locked playback decisions.

17. Deterministic: lock the variation

The hot-pink Deterministic ON/OFF control captures recent random playback decisions and uses stable choices for positions that have not been visited yet. It makes a changing idea repeatable without manually removing every probability, weighted speed, random pitch or humanize setting.

1. Play a phrase with the variations you like.
2. Turn Deterministic ON. The switch is seamless: running lanes keep their positions, current step timing and pending notes. Switching it off also continues in place.
3. Replay it and listen to the repeated choices. Save the song to keep the lock data.
4. Turn it OFF when you want fresh draws again.

Captured speed/feel, probability, pitch choices, modulation phases and random playback behavior become repeatable. Random direction uses a repeatable order. Lane randomization, paired ensembles and Sound / Mutate are disabled while the lock is on. You can still choose voices manually or shuffle choices in the voice browser; the playback lock is not a patch-editing lock.

For the most straightforward test, restart transport at the same place and compare complete phrases. A short lane and a long lane need not return to their first steps together on every pass.

18. Choose and preview voices

Your voice bank

Save voice in Sounds opens Your voices. You can also open it from Voice → User bank. Enter a name and save the current sound. Use ♪ to preview a saved voice, or click its name to load it and close the bank. Replace saves the current sound over that entry. To rename an entry, type the new name above, then click Rename on that entry. Delete needs a second confirmation click.

The bank stays on this browser/device and survives reloads; it is separate from the factory presets and your saved songs. While Step Lock is on, saved voices can be loaded into just the selected step. Saving a song includes the actual channel and locked sound settings, so playback does not depend on having its original bank entry.

Every lane’s Voice button opens the FieldSynth browser. The current library contains 791 factory voices, including 218 FM voices. These are synthesis settings rather than audio samples.

• Search: find voices by name or tag. Clear the search when you want to browse a full category again.
• Categories: Plucks, Keys, Bells, Percussive, Bass, Pads, Leads, Mallets, Drones, Stabs, Strings, IDM / FX, Organs, Wind / Breath, Plucked Strings, Ambient Textures, and Sub / Layers, plus Atonal Percussion.
• Waveforms and tags: narrow the pool by engine or character. Tags describe qualities such as warm, glassy, dub, IDM, evolving, bass, accent or foundation.
• Pair voice to channel: choose a destination for a complementary voice, matched using FieldSynth pairing tags. The source channel is greyed out. The destination defaults to the next channel (2 → 3, with 8 → 1). Undo and Redo beside Pair voice let you step through changes without closing the window. Pairing replaces the destination’s voice, clears its Step Locks and previews the new sound. The result shows ♪, the voice name and its destination channel. Written notes, pitches and sublane patterns stay unchanged. Default channel labels follow the active voice name; custom channel names are kept.
• Home / Back: move between browsing views without closing the menu.
• ♪ Preview: audition a candidate without applying it to the lane. Auto-preview can audition hovered candidates.
• Shuffle presets: mix up the whole category and show fresh choices on the page you’re browsing. Searches, waveform filters, tags and similar-sound results stay in their own pool. This rearranges the choices; it doesn’t change or audition your current sound. The button appears only when preset choices are visible; category home, tag lists and lane options have nothing to shuffle. With fewer than two choices, the button is disabled.
• Choose a voice: click once to use it and keep browsing. Double-click to use it and close the popup.
• Undo / Redo: buttons beside Swap similar voice let you move through voice changes without closing the popup. The pairing panel has its own Undo / Redo buttons linked to the same history.
• Swap similar voice: replace the lane’s sound with a related factory voice. Repeated clicks explore unheard voices from the starting sound, widening the search every few clicks instead of bouncing between a pair. While the voice browser stays open, no voice repeats until the library is exhausted. Manually choosing a different voice starts a new search. Preview auditions swaps when enabled. It does not add a second layer.

Choosing a voice changes its oscillator, carrier envelope and FM settings together. Lane gain, pan, routing, pattern and the separate lane effects remain available. Some long-attack voices need more time to bloom; a short note gate can make a pad seem quiet.

The browser’s Lane options page exposes the full sound parameters. Previewing a voice uses a separate audition path and does not insert notes or send those audition notes to external MIDI.

19. Shape a sound: subtractive and FM

Use Voice → Lane options for the full engine, including FM controls. The side editor’s Sounds tab offers a compact set of common lane controls.

Master note probability

Master FX → Notes → Note probability sets the chance each internal synth or canvas note will sound, from 0–100%. The default is 100%, which plays every note allowed by your lane and step settings. At 0%, new notes are silent; notes already sounding finish normally. It combines with X/Y note probability and stays repeatable in Deterministic mode. It does not change written notes, voice-browser previews or external MIDI output. The value saves with the preset; Reset master FX restores 100%.

Filter envelopes

Every lane filter and the master filter have a dedicated Filter Env popup with draggable attack, decay, sustain and release nodes. Drag Env Amount from −100% to +100%: positive opens the filter, negative closes it, and zero keeps the cutoff unchanged. Each new note retriggers the curve. These controls are separate from the amp envelope and the destination-based modulation envelope. They are saved with your song; old songs start with zero amount. Open Filter Env beside lane cutoff/resonance in Sounds or Voice → Lane options, or in Master FX → Filter. Each filter has one shared envelope: a new note retriggers it for that lane’s held voices; the master retriggers on notes in its bus. Sustain holds until that bus/lane has no held notes, then Release returns toward the base cutoff. At full amount the envelope spans up to six octaves, bounded by the filter range.

Oscillator, filter and envelope

Sine is smooth, triangle adds gentle harmonics, sawtooth is rich and bright, and square has a hollow character. A low-pass cutoff removes upper frequencies; resonance emphasizes the area around that cutoff. Level and pan set the lane’s place in the mix.

The carrier envelope controls loudness: Attack is the rise time, Decay falls toward the sustain level, Sustain is the held level, and Release is the fade after note-off. Attack/decay/release are times; sustain is a level, not a duration. Long envelopes can overlap later steps.

One LFO and one ADSR mod per lane

Open Sounds → LFO or Sounds → Mod Envelope (also available in Voice → Lane options). Each has its own On/Off switch, destination and signed amount. They are separate from the carrier and FM envelopes. Choose filter cutoff, resonance, level, pan, FM index/harmonicity or a lane effect. FM destinations need the FM engine. Both modulators add together if they share a destination; output values stay within safe ranges.

LFO: select sine, triangle, saw up, saw down, square, stepped random, smooth random or random drift. Stepped random changes once per cycle; smooth random glides between successive targets with a cosine curve; random drift gradually eases toward each new target. Random shapes hold their current value while Deterministic is on, and resume when it is off. Free Hz sets cycles per second from 0.01 to 20; Tempo division sets one full cycle per division, from 1/64 to 4/1. Division timing follows master BPM; it does not inherit a lane’s weighted tempo draws. In Free restart mode the LFO runs continuously while audio is enabled; Note trigger restarts it on each new lane note.

Restart (Free / Note trigger): Each modulator has its own Restart selector. Free · continuous keeps its clock running through new notes. Note trigger · restart resets the LFO to its starting phase on every new lane note; it then cycles at the chosen Hz or tempo division. In note-triggered Mod Envelope mode, every note restarts attack from zero, sustain holds while notes in that lane remain held, and releasing the last note starts release. Attack, decay and release remain in seconds; the free-running pulse-rate controls are hidden in this mode. These are shared lane modulators: a new note also restarts modulation affecting other held notes in that lane. Older songs default to Free.

ADSR mod, Free mode: a repeating pulse triggers the envelope, either at a free Hz rate or once per selected tempo division. Each pulse holds the gate high for half the cycle (attack, decay, sustain), then low for the other half (release). A/D/R are seconds, sustain is a level. Longer stages may be interrupted by the next gate transition. It modulates sound already playing; it does not create extra notes.

Positive/negative amount moves the destination above/below its base value; the LFO swings both ways. Cutoff uses a musical exponential sweep, level scales the existing volume so a zero-volume lane stays silent, and other destinations use their parameter range. Drag numeric values vertically; double-click resets a value. Click outside or press Escape to close. Settings are saved, copied and exported with the lane; changing factory voices preserves these modulators. MIDI notes and scale quantization remain unchanged.

Live modulation display: an active destination gets a thin colored strip under its slider. Its length and hue follow the actual combined, clamped audio-engine value, including Smooth Random, ADSR and Deterministic hold. Hover for the live numeric value. The slider handle and its number remain your saved base setting, so you can edit them normally. Lane volume and Voice → Lane options show the same feedback. Updates are capped at 12 per second and monitoring stops when the control is offscreen or the browser tab is hidden; this display never changes or autosaves the song.

FM engine

FM uses a modulator to change a carrier’s frequency. Harmonicity is the modulator-to-carrier ratio: simple ratios often sound more harmonic; non-integer ratios can produce bell-like or metallic color. Modulation index sets modulation depth. Higher values usually add stronger sidebands and a more complex tone.

The modulator has its own sine/triangle/saw/square waveform and its own attack, decay, sustain and release. This envelope changes brightness and harmonic complexity over time, while the carrier envelope changes loudness. A fast modulator decay with a longer carrier decay is a useful starting point for FM plucks and mallets.

When the FM engine is selected, the Sounds panel shows FM ratio (modulator frequency relative to the sine carrier), FM amount (modulation index), and modulator waveform. FM envelope opens its draggable envelope and attack, decay, sustain and release controls. Amp, FM and modulation envelopes use the channel’s color in both the panel and their popups. The carrier is a sine wave; Amp Envelope controls its volume. These FM controls also remain available in Voice → Lane options.

Lane effects are Reverb, Chorus, Delay, Distortion, Lofi and Phaser. Each channel’s Reverb amount sends into the shared Master FX room, with no separate reverb processor per channel. The channel send still works with master Reverb mix at zero; room size, low cut and quality remain shared. Returning the next step’s send to zero leaves the existing tail ringing. Sound changes do not rewrite note pitches; FM can add inharmonic overtones even when the played note belongs to the selected scale.

20. Master effects and mixing

Master FX is one shared stereo effects bus for all eight built-in lane sounds and auditions. These effects do not process an external MIDI instrument’s audio.

Built-in output protection helps with stacked voices, but leave room in the mix: eight loud lanes, long releases, feedback and large chords can still sound crowded. Set bass near the center, use lower levels for pads/chords, and add space after the basic balance works.

Delay follows BPM when a synced division is selected; Free uses milliseconds. A quarter-note delay is 60000 ÷ BPM milliseconds; at 120 BPM it is 500 ms. A dotted eighth is three quarters of that, or 375 ms.

21. Generate a lane pattern

A lane’s dice button opens its dedicated pattern generator. This changes musical material; it is separate from randomizing sound parameters.

1. Set the lower and upper sounding-note limits.
2. Choose eligible pitch classes. The current key/scale still applies; enabling an outside-scale pitch class does not override that constraint.
3. Set Step count from 1–64 and Density from 0–100%. The occupied-step count is rounded from steps × density.
4. Choose musical preferences: nearby melodic movement, spread hits across the pattern and anchor the first step.
5. Enable only the things you want to generate: Notes & density, Velocity, Note gates, Octave sublane, Note % sublane, Arp sublane and/or Step Actions.
6. Choose Generate preview, review the displayed notes, and use Apply to lane. Generate again if needed before applying.

Lane and global starter-pattern density both default to 75%. Double-clicking either density slider resets it to 75%. The percentage and occupied-step count update while dragging. Apply to lane saves the generated pattern in one undoable edit and closes its popup. Density means the proportion of occupied main steps, not note-trigger probability and not the number of notes inside a chord or triggered arp.

Unchecked settings, the lane’s sound, MIDI route and CC events are kept. Generated material follows the selected step count; shortening does not erase hidden data beyond the new active length. Sublane generation uses at most 16 slots.

22. Build a paired ensemble

Open global Randomize → Paired voices to choose sounds that work as an ensemble. Set Voices to use from 1–8 or click the exact target lane numbers. Roles are assigned in the order of the selected lanes.

Template genre/category rules first determine which voices are eligible. Pairing tags, register, envelope and timbral contrast help rank candidates inside those pools. A voice tagged “solo-only” in FieldSynth can still be useful on its own sequencer lane; that tag is not a ban on using percussion in an ensemble.

• Create starter patterns: write new role-appropriate notes into target lanes. Turn it off to keep their notes/timing while changing sounds.
• Density: 0–100%, default 75%, directly under the starter-pattern option. It sets occupied steps per starter lane. At 16 steps, 75% gives 12 occupied steps. Double-click resets to 75%.
• Steps per starter pattern: the active length for each generated lane.
• Play only chosen lanes: mute other lanes without deleting their material.
• Balance levels and pan: set sensible role-based levels and positions.
• Use template master FX: change the shared effects as part of the proposal.

Generate paired voices stages a proposal. Use a card’s dice to reroll that sound while keeping the other proposed voices and its pattern. Some tightly defined roles have only one eligible factory voice; their individual reroll is disabled. Hear together auditions the proposed sounds as a group, not the full generated sequence. Apply ensemble commits everything as one undoable edit.

23. Sound randomization and Mutate

Randomize one channel

The dice above the right-side Sounds controls opens Randomize sound. Select sections, set depth from 8–100%, then choose Randomize or the smaller 3–7% Mutate. Sections cover oscillator/amp, FM when the current engine is FM, filter, effects, LFO, Mod Envelope, Filter envelope and level/pan. All sections selects the complete available sound. Applying closes the popup and leaves notes alone. With Step Lock on, the result goes only to that step. Deterministic mode disables randomization.

Multi-step lock randomization

The dice on the Step Lock tile (Randomize steps) opens a large popup that gives several steps of one channel their own random sound, then lets you listen to each result. It uses Step Lock, so every chosen step gets a separate variation of the channel’s normal sound; notes, timing and the channel’s own sound do not change. It also works while Step Lock mode is on, and is unavailable while Deterministic is on.

1 · Choose steps. The popup shows one page of up to 16 steps with each step’s note. Click a step to include it. Rests are greyed out and cannot be chosen. Select all notes picks every step that has notes, and Clear selection empties the choice. A channel longer than 16 steps adds page buttons (Steps 1–16, 17–32 and so on); your choices stay when you change page. A ◆ marks an existing Step Lock, not selection. You can select or deselect those steps like any other note step; ✓ means selected. Randomizing a selected locked step replaces its sound. Unselected steps keep their current settings. The same sections and depth as the dice popup are here. Selected sections have a colored fill and ✓. All sections and None show which state is active. At the bottom right, Randomize steps gives each chosen step a new sound using the depth, and Mutate steps, to its left, makes a small 3–7% change instead.

2 · Listen and adjust. Choose Selected step to target the step currently selected in the sequencer, or select several steps. Listen to steps lets you audition and edit individual sounds without a batch randomization. After Randomize steps or Mutate steps, the new sounds are applied to your song at once without playing a note. Click any step to hear it with its sound: a chord plays all of its notes. Steps marked ✓ have new sounds; click an unchanged step to compare. The step controls show its number, previous/next arrows, Mutate step, Re-randomize this step and sound Undo/Redo. Mutate makes a small change to this step’s current sound; Randomize draws from the channel’s base sound using your chosen sections and depth. Both affect only the displayed step, even if it was outside the batch selection. Use left/right arrow keys to hear the previous/next note step; rests are skipped and the lane wraps. Undo sound change and Redo sound change let you compare edits while staying in the popup. Click that step or Hear it again to audition the result. Re-randomize all redraws every selected step. Edit step selection returns to the selection. Keep these sounds closes the popup; Cancel · restore original puts every step back as it was. Each change is also an ordinary edit, so Undo works. Editing a generated sound by hand turns it into a normal Step Lock that is still selectable in this tool.

Open Randomize → Sound / Mutate to reshape sounds without generating notes. This page has independent lane targets, group selection and individual parameter buttons.

Mutate versus Randomize

Mutate · 3–7% nudges each selected continuous parameter up or down by a fresh random 3–7% of its permitted range. Time and ratio parameters use logarithmic space, which makes small and large values behave musically. Clamping at a range boundary can make the actual movement smaller. Mutate ignores the depth slider.

Randomize · depth% chooses a random target for each selected continuous parameter and moves partway toward it. The slider runs from 8–100%, default 20%. At 100% it reaches the random target; at 20% it moves one fifth of the way. A normalized parameter at 0.5 with a target of 0.9 becomes 0.58 at 20% depth.

Choose exactly what changes

• Target lanes: choose which lane sounds may change. All lanes and Playing lanes are shortcuts; Playing lanes selects unmuted lanes containing notes.
• Sound: toggle oscillator/envelope, FM, filter and effects parameters as one bucket.
• Voices: toggle whole factory-voice replacement. This defaults off so a small timbre change does not unexpectedly replace a patch.
• Select all / Select none: set every parameter selection at once.
• Expandable groups: Oscillator & envelope, FM engine, Lane filter, Lane effects, Master effects, and Random factory voices. Each has an All / None shortcut and individual buttons. Mixed selection is shown when only part of a group is enabled.

The dedicated Filter Env ADSR and LFO/Mod Envelope settings are not targets in this sound-randomization page; edit them in Sounds or Lane options. Master FX is explicitly global: selected master-effect parameters change the shared mix even if only one lane is targeted. For a single-lane timbre edit, deselect the Master effects group. FM controls affect FM voices; if the lane is subtractive, those settings become audible when FM is selected.

Whole-voice and switch behavior

Discrete choices do not blend. Oscillator/waveform and factory-voice choices use a 35% change chance for Mutate and the selected depth as a change chance for Randomize. A selected boolean such as Lush reverb uses an 18% flip chance for Mutate and 55% of the depth for Randomize. Each selected factory-voice slot is gated independently.

If both individual synth parameters and a factory voice are selected, the whole voice applies last, replacing its oscillator/envelope/FM fields together. Filter and effects selections remain separate. At low depth, a voices-only draw may make no change; the status message reports that.

Each audible change is one Undo action. Depth, targets and selections persist with the project. Notes, sequencing, MIDI routing, lane gain and pan are preserved. Hear lane sounds auditions the current target sounds through the separate preview path.

24. Phrases and variations

A phrase is a snapshot of all eight lanes: notes, main/sublane lengths, speed settings, effects, CC events, direction and lane sounds. The song also has shared settings such as master tempo, key/scale and master effects.

• Phrase menu and arrows: select an existing phrase.
• Phrase name: rename the active phrase.
• New phrase: create an empty eight-lane phrase.
• Duplicate phrase: make a separate deep copy immediately after the source and select it.
• Copy / Paste phrase: paste one phrase’s contents into another while keeping the destination phrase’s name and identity.

You can keep up to 99 phrases. Changes in a duplicate do not edit its source. Save the whole song to retain all phrases.

Hold the × beside a lane’s dice for two seconds to clear its notes, Step Actions and lane settings while keeping its sound. Release early or move away to cancel; Undo restores the lane. Keyboard ← / → perform the same step navigation as the buttons beside Record, except while editing a field or popup.

Build an A/B phrase

1. Make a small arrangement and name its phrase “A”.
2. Duplicate it and name the copy “B”.
3. Change one melodic lane, open the filter or add occasional arp triggers.
4. Switch between A and B to compare the result. Use further duplicates for an intro, breakdown or ending.

The lane Copy/Paste action copies notes, timing, direction, transpose and sublanes while preserving the destination’s name, MIDI channel and sound. Rotate shifts active steps left/right. Double lane repeats its material and doubles its length up to the 64-step limit. Multi-select exposes page/lane selection for applying step edits to several positions.

25. Keyboard-controlled phrases

Keyboard Sequence turns prepared patterns into a keyboard instrument. Switch it On in its tile, then hold computer, on-screen or MIDI keys. A root transposes existing notes; it does not replace their patterns, sound settings or the chosen key/scale. Record is disabled in this mode.

Open Phrase Play → Canvas Settings and click either stack button to select it directly; this stops current playback. The selected mode remains active when you return to the editor. In Channel Stack, releasing a key frees its channel slot; other held keys keep their channels, and the next key takes the lowest free slot. Stop clears the assignments. Ordinary transport playback without keyboard roots still plays the arrangement. Older songs with neither stack selected can retain single-root playback until you choose a stack.

• Transpose: shift Keyboard Sequence playback by −36 to +36 semitones. Use the up/down arrows, arrow keys in the number field, type a value, or drag it vertically. Changes reach the next unscheduled notes, including an arp already in progress, without restarting playheads or changing written notes. Output still maps into the selected key/scale; Chromatic gives exact semitone shifts. The interval selector sets how far each arrow moves: ±1 semitone, ±3 (minor third), ±5 (fourth) or ±7 (fifth). The numeric field still accepts any semitone value. Zero is neutral; double-click resets it. The value is saved with the song and bypassed while Keyboard Sequence is Off.
• Ch Seq Order: choose the channel for the 1st through 5th key slots. Every channel stays selectable. Choosing a channel already assigned elsewhere swaps those two slots, keeping all five assignments unique. Any of the eight channels can be assigned. Changes save with the song. During Channel Stack playback, changed or swapped slots release their old channels and start their new channels at step 1; unchanged keys keep playing. Reset order restores channels 1–5. Voice Stack does not use this order.
• On / Off: enable or leave Keyboard Sequence. Switching stops playback and disarms Record. For writing notes, choose Single note entry or Chord entry and arm Record.
• Quantize: align added or changed roots to Off, 1/4, 1/8, 1/16 or 1/32. The first key starts immediately. This is input quantization, separate from each lane’s Tempo grid and the sound-canvas Segment grid.
• Multi Division: off shares timing choices across roots; on lets each root draw its own divisions and feels from the same lane weights. It separates timing results, not the saved sounds or editable sublane values. Changing it stops playback.
• Latch: hold up to five notes together after release. New playing replaces the latched set. Turning Latch off with no keys held stops playback; Stop/Panic clears it. Latch also enables five-root holding in older single-root songs.
• Phrase Play: opens the performance canvas, enables sound and five-key Keyboard Sequence, and keeps the selected stack routing.

Play five channels with five fingers

1. Put notes in up to five channels and choose different sounds for them.
2. Use Ch Seq Order to assign those channels to the five key slots, then open Phrase Play → Canvas Settings and select Channel Stack. Return to the editor if you want to keep the patterns visible.
3. Hold one key: it plays the first assigned channel. Add another: the second assigned channel joins with its own sound and pattern. Continue up to five keys.
4. Enable Latch to keep the set playing after release. Use Quantize to line up root changes, or Multi Division for independent timing draws.

All pitch output still maps to the song’s selected key/scale. A root does not change the key setting. Input velocity scales performance velocity. Editor auditions play individual notes, not the complete phrase.

Copy/Paste a lane to prepare related patterns on different channels, then change its voice or use its dice or Sound / Mutate controls. Two fingers using the same channel still share that channel’s editable settings.

26. MIDI input, output and CC events

The built-in synth works independently of MIDI. To control hardware or a DAW instrument, connect your device or virtual MIDI port, press Connect MIDI, grant browser access, then choose MIDI out and/or MIDI in. Each lane has its own MIDI channel; there is currently one shared output port.

1. On the receiving instrument or DAW track, enable the same MIDI port and channel as the lane.
2. Start with a one-lane pattern and check that notes arrive.
3. Assign separate channels to separate parts before adding bends or controller automation.
4. If you only want external audio, switch the browser sound off. MIDI output can continue.

Lane channel menus use the MIDI channel numbers musicians expect. Avoid sharing a channel between parts when using pitch bend, bank/program or CC events: those messages affect the receiving channel as a whole.

CC events

In the CC editor, choose controller number 0–127, value 0–127 and Set event. The step sends that controller value during playback. Setting the same CC again replaces its value for that step; Remove deletes the event. Use the receiving instrument’s controller map. CC 74 often controls brightness, but mappings vary.

CC smoothing in Timing interpolates between successive event values. CC dropout can conditionally bypass them. CC events are outgoing MIDI messages; they do not automate arbitrary built-in synth knobs.

Browser and connection requirements

Use the HTTPS live page and a browser that supports Web MIDI. Desktop Chrome or Edge is a practical starting point; support depends on browser, operating system and device. Built-in browser audio working does not establish MIDI support. See Web MIDI requirements and compatibility.

MIDI clock/transport output, external-clock sync, separate output ports per lane, MIDI-file export and audio recording/export are not available in this edition. To capture a performance now, record the outgoing MIDI in your DAW or use an external audio recording setup.

27. Save, open and back up songs

Autosave keeps your working session in this browser. The Preset Menu library stores named snapshots you can return to independently. Neither is cloud sync; both belong to this browser profile and site origin on this device.

Opening, renaming, duplicating and deleting concern the chosen library record. The current/unsaved indicator tells you whether the working session matches the linked saved copy. Ordinary Undo/Redo is for project editing; it is not a library trash or recovery system for deleted snapshots.

The portable HTML build contains the synth and factory sounds without external assets. Open it in a browser that executes HTML/JavaScript, rather than a phone’s file-preview viewer. Landscape suits the editing workspace. Portable files and the live site can use separate browser storage; JSON transfers songs between them.

Back up and transfer

1. Use top-bar Export to download the current project as JSON.
2. Keep that file somewhere outside browser storage.
3. On another device or browser, open Phrase Step and choose Import.
4. Select the JSON file, then save it into that browser’s Songs library.

Presets versus local songs: Songs and the disk icon save privately in this browser. Saving never publishes a song.

The disk icon beside Songs saves or overwrites the current named song in one click. It creates a library copy if none exists; subsequent clicks update that song. ⌘S / Ctrl+S does the same. The status distinguishes saved snapshots from unsaved library changes. Cross-window revision conflicts still produce an error instead of overwriting a newer copy.

JSON stores the song’s phrases, notes, voices, timing, effects, lane/master filter envelopes, LFO/Mod Envelope settings, Keyboard Sequence options (including Transpose and Ch Seq Order), X/Y assignments, deterministic data and randomizer preferences. It is project data, not an audio mix or standard MIDI file. Importing replaces the working project after validation and can be undone.

If another window changes or deletes a saved record, revision checks reject a stale overwrite. Open the latest saved copy or use Save as new song rather than repeatedly clicking Save changes on an outdated link.

28. Shortcuts and quick reference

Computer shortcuts are ignored while typing in ordinary text/number fields. Click a non-editing area if a musical shortcut is not responding.

The computer-keyboard note row uses this chromatic order, starting from the selected lane’s New note pitch:

A W S E D F T G Y H U J K O L P ;

At the default starting pitch, these cover C4, C♯4, D4, D♯4, E4, F4, F♯4, G4, G♯4, A4, A♯4, B4, C5, C♯5, D5, D♯5 and E5 before key/scale mapping.

The left/right arrows beside Record move entry slots. The arrows in the Phrases row change phrases. The arrows in Channel tools rotate the pattern. These are separate jobs even though the symbols look similar.

29. Four practical workflows

30. Troubleshooting

My channel sound is not changing everywhere

Check Step Lock. When it is on, sound edits belong to the selected step. Switch it off to edit channel defaults, or use × to clear that step’s locks.

I see playback but hear nothing

Load a known demo. Check Sound ON, device/browser volume, master volume and lane levels. Check mute/solo, empty steps, step probability, Note chance, Velocity effects, Note dropout and very low filter cutoff. A long pad attack with a short gate can be almost inaudible. Explicitly switching Sound off keeps it off until you enable it again.

I cannot see the Notes or Step Actions editor

A lane’s pink strip is expanded. Close it to return the side editor. The sublanes use that space deliberately.

The arp repeats one note, or does not trigger

Check Arp On, a non-None style, trigger chance, and the corresponding source note. Custom pitches can map to the same scale note. Check the chord preset and selected pitches. A 0% arp chance plays the written step; a 0% Note chance prevents the step from sounding at all. Remember the arp sublane has its own length, independent of the selected main step number.

The arp is slower or the lane loop is longer than expected

Each arp slot takes a full interval and temporarily holds the parent step. A long style cycle at a slow division can last a long time; Stutter repeats each selected pitch and Up – Down adds the return path. Check the lane’s division and feel, Hold step and Repeat step before changing global BPM.

The tempo seems to change although BPM stays fixed

Weighted lane divisions and independent triplet/dotted feel change local playback speed. Check the lane’s Tempo grid, its independent sublane grids and any Phrase Play X/Y Tempo mapping. Arps inherit the triggering main step’s interval. Select Fixed for a steady reference lane.

A note is different from the stored MIDI number

Key transposes C-relative notes and scale maps the result. Check lane transpose, octave sublane and pitch Step Actions. Use the Audition output label. For raw note numbers, choose C Chromatic with zero transpose.

MIDI devices are missing, or the external synth is silent

Use HTTPS and a Web MIDI-compatible browser, grant access, connect the device before refreshing MIDI, and check the chosen port/channel at both ends. External audio must come from the receiving synth/DAW, not Phrase Step’s volume control. Browser audio support and Web MIDI support are separate.

An external synth bends out of tune or another lane changes

Match bend ranges, use separate MIDI channels, and check the external synth’s own pitch processing. Bank/program and CC messages are also channel-wide.

A saved song is missing

Check the browser profile, device and full website origin. The live site and a downloaded portable file keep separate storage. Import your JSON backup to transfer the song. The library is not an account-based cloud service.

Sound crackles or a background tab falls behind

Keep the synth in the foreground, close heavy tabs, reduce long overlapping chords/releases and try the lighter reverb network. Browser scheduling can be throttled in background tabs. Very dense eight-lane arps and extreme effects demand more processing.

I changed something and want the previous version

Use Undo for musical edits, a saved song for a named checkpoint, or Import for a JSON backup. Library Delete is permanent and is not reversed by project Undo.

31. Glossary

32. Current edition and limits

This manual describes the current browser edition: eight main lanes, 1–64 main steps per lane, 16 visible main steps, 1–16 slots per sublane, a full selected style cycle per triggered arp, and up to 99 phrases. The app is designed around an iPad-sized landscape workspace and also runs on desktop.

Included now: built-in subtractive/FM sound engines, the current FieldSynth factory library, per-lane sound editing, dedicated lane/master filter envelopes, synced and dotted master delay, LFO and Mod Envelope with Free/Note trigger, MIDI notes and CC, step actions, weighted timing, generative pattern and sound controls, deterministic playback, phrases, named local song saving, JSON transfer, Channel Stack and Voice Stack, Phrase Play’s two canvases, X/Y modulation and session-only sound-canvas gesture loops.

Not included yet: per-finger editable sound/pattern stacks, Alternation mode, saving/reopening generated sound canvases, automatic song arrangement, bar-queued phrase launches, MIDI clock/transport synchronization, per-lane output ports, standard MIDI-file export or built-in audio rendering/recording. Do not treat a phrase list as a song playlist or JSON Export as a WAV/MIDI export.

The browser and connected hardware determine audio latency, MIDI availability and background scheduling. Settings and generated notes are recoverable through your saved library and JSON backups; no account or cloud song sync is provided.