✦ SPIELWELTENMoto Racer - Rampage ↗

Recorded world-tour atmosphere

BiomeAmbience.Configure(TrackDefinition, bool garage) now selects the actual chapter and stage, instead of collapsing the world tour into forest/desert/city. The engine, transmission, tire effects and event-only vibration are unchanged. Classic tracks retain a forest, exposed desert or quiet harbor setting through SetBiome.

Chapter Stage 1 Stage 2 Stage 3
Alps Sheltered leaf wind and distant temperate birds Less foliage; some exposed winter wind; no bird layer Quiet winter wind only; no summer bird/leaf loop
Tuscany Quiet foliage and sparse birds Slightly stronger bird layer among sheltered vineyards More exposed leaf wind, reduced birds
Mediterranean Restrained harbor waves Stronger cliff surf Cape surf and more dry wind
Egypt Dry air with low water layer Exposed dry desert wind Dry plateau wind; no river layer
Southeast Asia Thailand forest recording plus low foliage Distinct Borneo stream/cicada recording and water Quieter Thailand wildlife with more foliage wind
China Sheltered mountain woodland and quiet temperate birds Woodland with quiet birds More exposed foliage/wind; no bird layer
Grand Canyon Exposed dry wind Lower dry wind in the canyon Stronger open plateau wind

These are stage sound designs using licensed recordings, not claims that every bird species or recording location matches the fictional route. The Southeast Asian wildlife sources are from Thailand and Borneo. Amazon river audio is used as a generic water texture, not as an Amazon wildlife bed. Winter wind is recorded in Tasmania; desert wind in the Mojave. The snow excerpt ends well before the source's reported late arriving car. No additional synthetic engine whine, pitch-shifted birds, music or constant vibration is added.

Sources and preparation

All eight recordings are CC0-1.0, checked on the authors' original Freesound pages. The exact public HQ MP3 previews, source-page snapshots, sizes and SHA-256 hashes are retained under ArtSource/RegionalAmbience; the asset manifest records the actual files and edits. These are publicly available previews, not the larger original WAV files that require a Freesound login.

Tools/Art/prepare_regional_ambience.py downloads at most 25 MB per source from its published preview URL, verifies the CC0 page, decodes stereo, selects a bounded excerpt, applies restrained rumble/high-frequency filtering, removes DC, and creates a three-second internal equal-power overlap. Rotation leaves adjacent original samples at the external loop boundary. A fixed loudness target preserves variation within the recording; a peak ceiling prevents clipping. No generative replacement of the source is performed.

The eight authored WAVs total 67,032,352 bytes, stereo PCM16 at 44.1 kHz, 45–49 seconds per loop. Unity import is stereo Vorbis, quality 0.8, compressed in memory, no normalization, no background load. This compressed runtime size remains to be measured after import; the WAV size is not claimed as player memory use.

Runtime lifetime and transition behavior

The persistent application owns four AudioSources: two stereo layers in each of two banks. Steady playback uses at most two voices; a 2.4-second crossfade uses at most four. A rapid new selection replaces the quieter bank and lets the dominant bank fade out. Source count cannot grow with stage count. The eight imported Resources clips are warmed once at application initialization. No stage allocates synthetic PCM, calls AudioClip.Create, duplicates imported clips, or destroys clips used by another owner.

Garage/menu return fades the outdoor layers to zero and stops them. Pause explicitly pauses playing sources and freezes their crossfade progress; resume unpauses them. Disabling the component stops all four sources and resets selection/fades. Effect volume follows the existing settings; the AudioListener continues to apply master volume. Diagnostics emit RAMPAGE_AMBIENCE <profile-id> on outdoor changes, and AtmosphereId is available for a native capture.

Verification and remaining listening

The preparation manifest records hashes, source license evidence, PCM channels/rate, RMS/peak, zero clipped samples and loop-boundary steps below the normal 99th-percentile adjacent-sample step for every file. These are digital asset checks, not a listening claim.

Tools/Verification/verify-regional-ambience.ps1 compiles Runtime + Editor against the installed Unity assemblies and executes the actual runtime resolver for all 21 stages. It checks distinct stage selection, snow without summer birds, reduced highland wildlife, water only in the Nile stage, dry canyon profiles, classic fallback and out-of-range stage guards. The report is Docs/QA/regional-ambience-routing.json. This test never executes Unity native audio.

The offline check has completed: 85/85 resolver assertions, all 21 stages; Runtime 85 files and Editor 19 files compile with zero warnings/errors. The report records the exact BiomeAmbience.cs SHA-256. None of these checks executes native audio.

The next native player QA must confirm the imported assets load, pause/resume and menu silence, bounded AudioSources and AudioClips over stage cycles, and the complete engine/road/environment mix. Subjective listening has not been performed by the agent; the source MP3 quality and integrated balance are not certified as AAA. In particular, exposed wind must remain quieter than the engine and tropical cicadas should not become a constant sharp hiss.

Reproduce with system Python (NumPy/SciPy) and ffmpeg:

python 'Tools/Art/prepare_regional_ambience.py'
& 'Tools/Verification/verify-regional-ambience.ps1'