HARBOUR AFTER DARK — AUDIO CREDITS

Music: "Last Ferry", generated for Harbour After Dark with Eleven Music (music_v1).
Original 60-second instrumental. Runtime edit: 59.88 seconds, normalized peak,
120 ms wrap crossfade, gapless MP3 encoding.

Sound effects and harbour ambience: generated with eleven_text_to_sound_v2.
13 player/world cues, nine zombie vocal takes, one zombie footstep,
one harbour ambience loop. Zombie vocals revised with natural pitch,
alternate takes and a raw human performance direction.
Provider: ElevenLabs (https://elevenlabs.io)
The original 25 assets were generated on 2026-09-11 under verified Creator membership.
The retained source pack contains 88 assets, including the original boss pack, nine
first-revision spoken takes and twenty Australian horror vocal takes (including clarity and clean-language revisions), generated
under verified Creator membership. Runtime publishes and loads 60 files. Twenty-eight superseded or cancelled
voice/effect/music originals remain in the authoring archive, outside the game export. Voice-design previews are retained separately. Exact generation times are in generation.json.

Boss music: "Undertow" (The Tidekeeper), "Pressure Hull" (The Breakwater),
and "Overload" (The Livewire), generated with Eleven Music (music_v1).
Each is an original 60-second instrumental, prepared as a 59.88-second stereo
loop with a 120 ms wrap crossfade. Encounter music crossfades over 650 ms.
Each boss also has five distinct effects (entrance, swipe, slam, hurt and defeat),
generated with eleven_text_to_sound_v2. Originals and prompts are retained.
The earlier free-plan generations and CC0 music are archived locally and are
not included in the current game audio or export.

Usage follows the account's ElevenLabs agreement and applicable model terms.
Paid-plan sound effects can be incorporated in commercial projects; do not resell
or distribute them as a standalone sound library.
Creator music terms permit many commercial uses, but exclude Studio Games
(monetized games offered through more than one platform), film, TV and radio.
No blanket license beyond the provider's applicable terms is asserted.
https://help.elevenlabs.io/hc/en-us/articles/13313564601361-Can-I-publish-the-content-I-generate-on-the-platform
https://elevenlabs.io/eleven-music-model-specific-terms
https://elevenlabs.io/use-policy

Prompts and hashes: assets/audio/requests.json, generation.json and runtime.json.
Source files: output/audio/raw/. Plan and usage record: assets/audio/sources.json.
Generation uses a local key; the running game needs no API key or provider access.

Boss dialogue: generated with ElevenLabs eleven_v3. Premade voices: Callum
(Tidekeeper), Harry (Breakwater), Charlie (Livewire). Each has one entrance
and two hurt lines. Exact voice IDs, scripts and request hashes are recorded
in assets/audio/requests.json and generation.json. No voice cloning was used.

Current boss voices: three original Australian horror characters designed with
ElevenLabs eleven_ttv_v3 and performed with eleven_v3. Replaces the first spoken
revision on entrance, hurt and defeat. All prior recordings remain preserved.
Voice designs, selected preview IDs and hashes: assets/audio/aussie-boss-voices.json.
Design prompts and scripts: assets/audio/aussie-boss-voice-designs.json and requests.json.

Current scripts use clean threats without swearing or insults. Superseded takes
are excluded from public/audio and game exports by assets/audio/runtime-selection.json.

September 2026 level expansion: Training Gym room tone, generated with
eleven_text_to_sound_v2, is a 12-second stereo indoor ambience loop.
The original downloads, exact prompts and provider response metadata remain
in the same authoring archive and audio manifests listed above.

Australian wildlife: 13 short species cues generated with ElevenLabs
eleven_text_to_sound_v2 under the returned Creator plan on 2026-09-12 UTC.
Eleven moving species, plus kookaburra calls and shark-fin water movement.
Spiders and cockroaches use stylised tiny movement foley. Natural pitch,
44.1 kHz mono MP3, trimmed one-shots with -3 dBFS preparation peak target.
Exact requests, provider response metadata and preserved source/runtime hashes
are in the audio manifests. Subjective listening acceptance is pending;
browser playback and numeric checks do not establish species accuracy.