Back to the example index

Shared scores and interruption recovery

Preview and export the same fractional-note score. Exercise smooth controls, tuning, mix drive, voice stealing and safe shutdown. Start with a comfortable listening volume.

Engine configuration Interruption policy (apply after Dispose) Voice stealing (apply after Dispose) DSP quality (apply after Dispose) LFO waveform (next patch) Reference A4 (Hz) MIDI 60 tuning offset (cents; fractional notes interpolate) Pre-saturation mix gain (not host volume)

Sounding-note controls Expression Pan Modulation multiplier Ramp (seconds; 0 keeps immediate control) Feedback Independent LFO rate (Hz) AM depth PM depth (cents)

Musical transport and performance Tempo (BPM) Seek position (beats) Loop beats 0–8 Performance mode (new performance helper) Mono priority

Voice-bank lifecycle and export WAV format

Short preview retains a bounded WAV download. Long export requires a browser file-picker sink; it never retains the complete file in memory.

Loading example…

No context created.

Repeatable physical-device acceptance — manual evidence only

Use a physical iOS Safari or Android Chrome device over HTTPS. Test every scenario separately with cancel and preserve (Dispose before changing policy). Begin submits a held gate and a future onset; finish silences them. If the OS did not interrupt audio, say so: no policy proof. Desktop automation never earns physical acceptance. Checklist, support matrix and smoke workflow.

Evidence environment Device model OS version Browser version Named scenario

Subjective marker Marker note (240 characters) Scenario notes (route, battery, interruption, audible recovery; 1200 characters) I manually performed this scenario and observed/listened to the result on the declared device Manual judgment (not computed from lifecycle events)

No physical-device result recorded.

Local-only JSON: actual context sample rates, declared device/OS/browser, fixed policy, timestamps, bounded observations, subjective markers and separate manual/acceptance statuses. At most 24 captures, 256 observations and 64 markers per capture. No audio recording, persistent storage or upload. Active/absent captures remain unverified. Export before leaving; reloading clears captures.