Recorded sportbike listening comparison

All four WAVs are complete 25-second offline renders at 44.1 kHz, mono PCM16. Files01–03 use the same speed/throttle timeline below and engine/master settings (.75/.85), without separate loudness normalization. File04 is the current larger-displacement event demonstration with its own timeline. All previously shared bytes are preserved.

• 0–3 s: idle.
• 3–12 s: accelerate to 170 km/h through automatic shifts.
• 12–15 s: hold 170 km/h with throttle.
• 15–19 s: close throttle and brake to 60 km/h.
• 19–23 s: accelerate to 130 km/h.
• 23–25 s: hold 130 km/h.

01-previous-synthetic-engine.wav reconstructs the previous sine-based engine, speed-dependent bright noise and quiet CRF250 recorded layer. Its noise generator and some smoothing/startup behavior are approximations, so it is not an exact old-player capture. 02-recorded-sportbike-redesign.wav introduces real RPM layers, shifts, overrun and quiet air. 03-recorded-sportbike-refined.wav preserves more upper GSX detail and uses the dark Ninja recording as additional body.

Current larger-displacement demonstration

04-large-displacement-load-and-limiter.wav follows the user's request for a deeper, larger motorcycle with distinct rev matching, load changes and limiter. Its25-second trace came from the production engine-state class at creation. That source is now archived: the current game uses a newer physical manual/automatic transmission, which this historical preview does not validate.

• 0–2s: lower idle.
• 2–12s: accelerate to215km/h through the gears.
• 11.5–14s: loaded top-gear limiter.
• 14s: throttle lift; brake down to65km/h through19s.
• 17.57,18.23,18.93s: downshift rev matching.
• 19s: throttle reapplication, accelerating to145km/h by23s.
• 23–25s: cruise145km/h.

PCM peak is.236755, RMS.047319, with zero clipped samples. The different speed/load timeline makes04 an event demonstration rather than a matched loudness comparison with01–03. comparison-v4.json records13 events, the exact state/trace/output hashes and levels. engine-preview-v4.csv is the production-state trace.

These focused mixes are not Unity AudioSource captures. The engine state and gain/pitch behavior are reproduced offline; Unity resampling, loop-start phase, scheduling and the whole-game mix remain unmeasured. Tires, environment, rivals and game feedback are excluded. Numerical checks do not establish subjective sound quality, and no human listening acceptance is claimed.

Rebuild from the repository root using Python 3 with NumPy/SciPy and ffmpeg on PATH:

$env:OPENBLAS_NUM_THREADS='1'
$env:OMP_NUM_THREADS='1'
& 'Assets/Rampage/Resources/Audio/Sources~/run_audio_verification.ps1'
python 'Assets/Rampage/Resources/Audio/Sources~/build_sportbike_audio.py'
python 'Assets/Rampage/Resources/Audio/Sources~/verify_sportbike_audio.py'

The source recordings, page snapshots and scripts are retained in Assets/Rampage/Resources/Audio/Sources~, excluded from Unity runtime assets. Docs/RECORDED-AUDIO.json records hashes, licenses and edits. Runtime integration compiles offline; Unity playback waits for the user's driver-update pause to end. audio-validation.json retains47/47 digital delivery checks and31/31 explicitly historical preview04 state checks. The newer physical transmission has its own48/48 checks in Docs/QA/transmission-state-verification.json. These are separate from game-flow test totals. The renderer refuses to replace published04 with differing bytes; future revisions need a new filename.