Mist is an interactive fictional scenario: a spray contacts the selected bike after two simulation seconds, its engine stalls and it slows with the traffic model, then restarts ten seconds later. The droplets and timed engine response are illustrative, not a prediction of real water effects.

Web Audio synthesizes continuous engine and road sound at the active observer. The strongest 32 sources have individual engine voices; remaining modeled traffic contributes to the ambient mix. This perceptual audio mix is not the phase-resolved comparison solver. Audio starts off. Synthetic V-twin engines use uneven firing pulses, low exhaust resonance, and throttle-linked texture rather than recordings. Playback uses the calculated viewpoint contributions without a second distance penalty. A compressed listening range targets -28 dBFS at 80 modeled dBA, -23 dBFS at 90, and -13 dBFS at 110, before output processing. The playback meter measures digital output, not physical speaker SPL. Device volume, headphones, and speakers are uncalibrated. Background ambience remains quiet. The synthetic high-noise scenario uses a 136 dB(Z) full-load motorcycle reference at 1 m, with modeled idle at 86 dB(Z) and steady cruising up to 114 dB(Z) and a 20 dBA background; cars remain independent at 74 dB(Z). Estimated dBA still depends on engine spectrum, motion, source count, geometry, and propagation, not fixed distance bands. A compressor limits the combined mix. Device volume never changes the dBA calculation.

NYC street and building geometry; synthetic traffic demand, signals, source levels, and engine waveforms. No field calibration.

The live map includes every simulated source with travel delay, distance loss and approximate roof diffraction. First-order facade and treatment paths are resolved for nearby sources; farther sources retain direct and diffracted energy. Paths from each source combine with phase; different traffic sources use independent energies. Amber footprint outlines mark missing building heights and do not become invented acoustic walls. Marked cancellation nodes add an output-limited narrowband phase estimate with controller delay; this is not full broadband cancellation. Detailed comparisons use the separate waveform solver.

Colors show a 7-by-7 outdoor sample grid near street height over the viewed area. The active microphone and observation markers are measured separately at their actual heights. The timestamp identifies the latest completed sample; visual movement does not imply a fresh acoustic result.

-- physical sound speed. Optional trails are geometric propagation markers, not a full wavefield. The 8 kHz acoustic solver excludes frequencies above 4 kHz.

Passive surface interception and powered output have separate ledgers. A panel ledger is not a complete canyon energy audit. No guaranteed reduction, personal identification, or manufactured-reflector performance is claimed.

Street context: OpenStreetMap. Building geometry: NYC Open Data. Park exteriors: the same NYC Parks property-boundary snapshot used by Sunwalker; rims are not surveyed paths. The McCarren running track uses OpenStreetMap outer and inner boundaries, with eight interpolated lanes (ODbL, OpenStreetMap contributors). Trees combine mapped OSM positions with seeded supplementary park planting. Sports areas and paths use OpenStreetMap geometry. Additional visitors, strollers, dogs, court markings, and equipment are illustrative, not measured populations or acoustic sources. Canopy shapes and missing heights are illustrative, without invented acoustic shielding. Packaged snapshot: July 13, 2026.