01 / PLAY THE SURFACE
A shape, under your hand.
T means triplets. 3–3–2 emphasizes existing hits on beats 1, 2&, and 4.
Choose a guide for an exact rate. Tempo aligns hits to the beat.
THE VOICE / ONE GESTURE, DIFFERENT RESPONSES
Knock, hat, or rattle.
Short knock starts dry and percussive. The longer voices remain available.
THE TOUCH
Choose the impact.
Short contact reveals more of the carving; long contact rounds the hit. Density changes timing. Each new phrase keeps the touch you chose.
03 / SAME GESTURE, ANOTHER BODY
Your phrase.
Exact score: change segment divisions in the builder to adjust density. This preserves rests and mixed rhythms.
Older sound version: this pair uses the current voices. Earlier lab sounds were not versioned precisely and may differ.
Replay keeps the recorded touch. Body, voice, pickup and pitch use the current settings. Length is held for bar captures. One-shot includes the tail. Loop wraps the resonance into exactly the captured bars. Neither includes the pulse. The phrase BPM below is retained even if you change the next-take BPM.
04 / WRITE A GESTURE
Build a phrase.
Segments play in order on the tempo grid, at the BPM and accents above and the touch you chose. Each one is a number of hits at one division, moving the strike from one place to another. Build replaces your current phrase (save it to the shelf first to keep it); after that it is a normal phrase: replay it on any voice, loop it, export it.
After you edit or build a recipe, it is saved with Pair JSON. Build replaces the current phrase; Replay keeps its exact hits until you build again.
05 / STEP MACHINE · LAB
Program a groove.
Four rows of 16 steps (one bar of sixteenths), each row its own voice, all playing together. Choose a paint, then tap steps; tap a lit step with the same paint to clear it. It loops, and edits land on the next bar. The WAV is the exact loop, tails wrapped round; Stems saves one file per row.
Objects worth returning to.
Body + phrase · saved on this deviceWhat is being modeled?
The shaker. A seeded cluster of small synthetic collisions excites a short resonator bank. Position changes fine/coarse texture; Loose changes collision spread. The variation repeats with the phrase.
The metallic hat. A separate designed voice: filtered noise excites 24 metallic resonances. Position selects overlapping frequency regions; openness controls decay. Metal–Air balances resonance and noisy attack. Held-open hits use a separate ringing layer; optional choking fades that layer on the next normal hit. It does not use the carved bar’s thickness, pickup or pitch. Each captured hit repeats its deterministic attack on replay and export.
A bending bar. One free-free Euler–Bernoulli beam with a variable thickness profile. Length 45 cm, width 4 cm, nominal thickness 12 mm. The material is a fixed elastic approximation, not a calibrated wood instrument. Its bending frequencies and spatial modes are computed from the profile; shear, twisting, and sound radiation are omitted. Very high modes are left out of audio playback.
A virtual contact pickup. Strike and listening positions select different mixtures of the computed modes. The output follows local vibration velocity. Each body gets a fixed level balance, preserving contrasts across its surface. A smooth peak guard catches loud buildup. These are musical controls, not a radiation model. The voice controls choose decay and partial balance. Ratio reference blends toward uniform-bar intervals after the solve, while retaining the carved mode shapes and masses: a hybrid voice, not inverse physical tuning. At zero blend the computed ratios remain. Pitch matching scales all frequencies together.
A repeatable performance. No timing jitter, compulsory landing, or density-driven brightness. Release stops new hits. Damp ends the phrase and its ring. Stop and Escape also stop the pulse. Free captures end at 8 seconds; bar captures span 1, 2, or 4 bars in 4/4 (up to 24 seconds). Shape and sound settings are held during playback.
Compare fairly. Try a single strike, a spacious roll, and a dense sweep on each shape. Use a matched pitch to hear beyond pitch alone. The same phrase on the uniform and edited bars tests what carving adds. Exported WAVs use the same body balance and peak control as playback; your Output slider is for listening. Older pairs open with their original 04.1 voicing. New pairs also store the voice controls. Switching voice keeps the same recorded hits. Short decay changes the sound, not the gesture range: Free runs from ⅓ to 24 hits/s; Tempo uses subdivisions within that range.