aetherbound-guild/docs/presentation/AUDIO_MUSIC_AND_VOICE_CATALOG.md

Aetherbound Guild: Audio, Music, And Voice Catalog

Authority: sonic identity, adaptive music, ambience, voice scope, event-to- sound map, mix priority, captions, accessibility, and production evidence.

Companions: SCREEN_AND_STATE_MAP.md, UI_AND_VISUAL_SYSTEM.md, and ART_ASSET_AND_ANIMATION_CATALOG.md.

Current status: specification only. No SFX, ambience, music, or voice is generated, licensed, recorded, integrated, human-listened, or release-accepted.

1. Sonic Thesis

A world of distinct local cadences being repaired without being forced into one voice.

The audio identity combines close, practical repair materials with separable regional instruments. Ceramic bells, wood, cloth, paper, glass, hand tools, water, pressure valves, struck metal, and controlled breath make action physical. Aether is a clean double-tone whose direction and stability remain audible. The score does not use a constant heroic orchestra or a dense fantasy loop. It leaves room for telegraphs, named adventurers, and the player's diagnostic attention.

The eight regions retain different meters, instruments, room tones, and route materials. The final chapter combines them without blending them into an indistinct wall. The player should hear difference cooperating.

1.1 Audio Priority

From highest to lowest:

1. accessibility/system block, save conflict, or destructive confirmation;
2. imminent enemy telegraph that requires a decision;
3. committed Directive, damage prevention, rescue, or failure cause;
4. spoken line selected by the voice-priority rules;
5. player navigation, equip, reward, Cache, and route commits;
6. ordinary combat action and profession resource feedback;
7. ambience and automatic/routine events;
8. music detail layers.

Lower layers duck, thin, or defer. They never solve priority by becoming louder. Mute, captions-only, mono output, and low-power modes preserve every gameplay decision.

2. Bus And Routing Contract

Master
|- Music
|  |- MusicBase
|  |- MusicPulse
|  |- MusicTension
|  `- MusicResolution
|- Ambience
|  |- Weather
|  |- Settlement
|  `- Machinery
|- Voice
|  |- Critical
|  |- Dialogue
|  `- Bark
|- SFX
|  |- CriticalTelegraph
|  |- PlayerAction
|  |- Combat
|  |- World
|  `- UI
`- Accessibility
   `- OptionalAudioCues

Master, Music, Ambience, Voice, and Effects are separately adjustable. The UI may expose a single Effects slider while preserving sub-bus mix internally. Master mute is immediate and reversible. Focus loss pauses/ducks according to platform policy; resume fades in over 150-350 ms and never restarts the track.

2.1 Voice And Event Caps

Family	Concurrent cap	Rate rule
Critical telegraph	2	never suppress the earliest; later one becomes visual/caption and queued short cue
Character semantic voice	1	next line waits or is replaced by caption; no six-character overlap
Character effort	3	group within 120 ms; lowest-priority effort suppressed first
Enemy signature	3	boss phase has priority over ordinary enemies
UI navigation	2	repeated focus movement grouped to max 8 cues/s
Damage impacts	6	same-material impacts inside 80 ms group into one layered event
Automatic/routine actions	2	same source max 1 cue/1.5 s unless a new player decision appears
Reward/commit stinger	1	no new stinger until prior transient completes, max 1/1.2 s

3. Loudness, Format, And Lifecycle Targets

These are starting engineering targets, not a substitute for listening:

• deliver masters as lossless WAV, 48 kHz, 24-bit; runtime compression selected by platform after ABX and memory tests;
• preserve at least 1 dB true-peak headroom in individual assets and mix;
• no DC offset, unintended leading silence, clipped transient, or cut tail;
• loops align at musical bars or edited ambience seams and survive ten repeats;
• critical SFX remain intelligible at low device volume and through mono fold;
• spoken language remains intelligible without requiring music below audibility;
• pause, suspend, headphones/output change, interruption, background/foreground, and device route change resume without duplicate or orphaned voices.

Every loop is machine-checked for format, duration, peak, silence, and seam proxy, then listened to for at least 10-15 minutes in its real context on speakers and headphones. Machine checks do not grant human listening.

4. Leitmotif And Material Vocabulary

Motif	Sound identity	Gameplay meaning
Guild repair	four-note open cadence, ceramic bell plus wood	practical shared work, objective, safe commitment
Aether route	two clean tones separated by a fifth, directional shimmer	path flow; interval narrows or breaks when ruptured
Quiet Choir	attractive unison that removes one voice on repetition	efficient uniformity becoming dangerous
Named difference	separate timbres answer without doubling pitch	different roles/actions creating stability
Unsecured Cache	open copper clasp, soft unresolved downward tick	owned only conditionally; risk remains
Secured result	clasp closes, low wood confirmation, no casino sparkle	atomic bank/commit
Command	mechanical four-point seal arm and precise blue-toned strike	player-authored intervention
Break	material fracture followed by exposed quiet center	enemy stability opened; high-priority window
Recovery	repair knot, breath, and incomplete cadence offered again	retained people and credible next plan

The motif is never the only information channel. Captions, geometry, values, and event logs remain authoritative.

5. Adaptive Music Scope: 30 Suites

Each suite is an authored musical identity, not one exported loop. A normal suite contains synchronized base, pulse, tension, and resolution stems/cues; low-density screens may omit pulse or tension. Total stem count is finalized after arrangement, but the complete scope is 30 suites.

ID family	Count	Emotional job	Starting tempo / loop guidance
MUS-TITLE	1	invitation: a broken line is repairable	66-74 BPM, 90-150 s base, sparse motif
MUS-GUILD	1	sustained ownership and working concentration	74-82 BPM, 120-180 s; eight optional regional timbre layers
MUS-FORMATION	1	quiet prediction and comparison	70-78 BPM, low percussion, phrases leave decision rests
MUS-RECORDS	1	recollection without museum solemnity	free/60-72 BPM, intimate material, no constant lead
MUS-RELATIONSHIP	1	attentive human space	58-72 BPM, small ensemble, dialogue-first ducking
MUS-RECOVERY	1	diagnosis, retained possibility, no shame dirge	62-70 BPM, unresolved-to-open cadence
MUS-CH01..08-ROUTE	8	regional travel and route anticipation	72-96 BPM, 120-180 s bases, branch pulse layer
MUS-CH01..08-BATTLE	8	readable automatic execution and rising pressure	88-116 BPM, 90-150 s, removable tension/percussion
MUS-CH01..08-BOSS	8	each chapter thesis under pressure	mechanic-timed but not rhythm-game locked; multi-phase stems
Total	30	complete campaign music identity	ending and Second Weave arrange existing motifs into new suite states

The ending uses the Guild suite plus eight independently audible regional layers and an authored resolution extension. Second Weave reorders meter, instrumentation, and stem-entry laws; it does not simply pitch or speed up the campaign music.

5.1 Chapter Music Briefs

Chapter	Route palette	Battle pressure	Boss distinction	Anti-goal
CH-01	open wind, wood pump, ceramic bell, plucked strings	dry frame drum and broken pulse gaps	silent missing beat becomes an Anchored counter rhythm	pastoral flute wallpaper
CH-02	water slap, reed breath, voiceglass harmonics	delayed ripples and numbered echoes	recital phrases replay with secondary clauses audibly missing	wet reverb obscuring timing
CH-03	work rhythm, anvil, pressure valve, low strings	resource-spend clicks accumulate into heat	alternating band rhythm vents the oppressive mint meter	constant loud forge percussion
CH-04	leaf brush, paper, ink scratch, woodwind pairs	bounded Echo phrases and intentional contradiction	repeated phrase is physically redacted, counter phrase remains	random mysterious whispering
CH-05	harbor rope, shell, two offset melodic lines	mirrored timing and left/right source separation	uniform lines collapse; asymmetric answer creates Break	stereo-only identity clue
CH-06	vane whistle, lift cable, cloud drum, rotating meter	directional gust accents with stable HUD pulse	four ward meters rotate around one calm, quiet center	exhausting thunder bed
CH-07	close lantern room, registry page, subdued market hands	comforting motif reveals a missing relational beat	false saint stays attractive while source markers expose it	generic dark-purple horror drone
CH-08	eight retained material timbres	linked telegraphs pass across distinct role motifs	unison mutes itself; three roles rebuild a plural cadence	maximal orchestral wall

5.2 State Transitions

• base begins on entry and never restarts for tab changes within a location;
• pulse enters after an expedition/formation commit at the next bar;
• tension follows committed pressure: boss phase, imminent linked telegraph, critical rescue window, or exposed high-value Cache, not arbitrary low HP;
• resolution is a 2-8 second authored cue, then yields to the next base;
• pause applies a low-pass/level reduction without losing musical position;
• dialogue ducks lead/pulse material by rule, not the entire world to silence;
• reduced-audio preference removes high-density pulse ornament but preserves state transition and critical motifs.

6. Ambience Scope

Ambience communicates place and process without becoming a second score.

Family	Count	Examples / states
Guild Yard	10	broken opening, eight accumulated regional service layers, completed federation; day/evening handled as layers
Chapter settlement	16	before and after restoration for each chapter
Chapter route	24	three route material/weather families per chapter
Battle stage	24	ordinary/camp, elite/sub-boss, and boss environment beds per chapter
Camp/Anchor	8	one region-specific quiet recovery bed per chapter
Narrative/indoors	12	workshop, clinic, records, training, ferry, archive, foundry, harbor, ward council, Nullmarket, First Junction, ending hall
Total authored ambience configurations	94	built from fewer reusable raw layers plus state-specific mixes

Every configuration specifies loop length, entry/exit fade, positional sources, foreground process events, low-power density, and silence budget. Machinery and weather layers respond to the restored service and chapter consequence. A completion state cannot use the unchanged pre-restoration ambience.

7. Voice Scope

All 30 adventurers have authored vocal identities. Voice acting is localized for English and Simplified Chinese; no locale uses an imitation of a protected performer. Casting, actor consent, contracts, union/territory requirements, voice cloning rights, and commercial terms must be verified before recording.

The Character Bible currently specifies exactly:

• 773 semantic combat/Hub lines across 30 characters;
• 144 character relationship/arc scenes;
• 30 distinct voice identities and localization notes.

7.1 Recording Budget Per Locale

Family	Count per locale	Purpose
Authored combat/Hub semantic lines	773	deployment, skill intent, warning, counter confirmation, rescue, reward, route/Guild response
Nonverbal effort palette	240	8 per character: light/heavy effort, hit, forced move, downed, rescue, relief, strain; no repeated words
Scene bookends	288	two voiced lines for each of 144 scenes: opening intention and closing changed position
Minimum recorded assets per locale	1,301	scene bodies remain fully authored text unless production later approves full scene voice
English + zh_CN minimum	2,602	retakes/alternates excluded from release count

This selective-voice contract keeps every gameplay-critical warning and character identity audible while avoiding an unbudgeted promise to voice every line in 144 scenes. Full scene voice is a later scope decision and cannot remove subtitle or transcript requirements.

7.2 Semantic Bark Matrix

Every character's authored line budget covers compatible rows; lines are not filled merely to hit a number.

Event	Priority	Cooldown / suppression	Caption behavior
First deployment / chapter return	medium	once per expedition/chapter context	speaker + line
Profession active intent	low-medium	signature line max once/45 s; efforts otherwise	speaker + action only when semantic
Correct counter recognized	high	once per mechanic/encounter	speaker + counter cause
Imminent danger warning	critical voice	one best-informed eligible speaker	speaker, target, event direction
Directive target/commit	high	selected actor or commander only	speaker + committed effect
Downed / Rescue Window	high	affected unit then rescuer; no group chorus	speaker + state and timer
Rescue success	high	one rescuer/returned unit exchange	two ordered captions
Strain gained / deployment blocked	medium	once at result/Guild	speaker + exact state
Reward/build interaction	medium	only character with a specific authored interaction	speaker + item relation
Anchor/route consequence	medium	one context owner, max one/decision	speaker + choice consequence
Victory	low	one character chosen by contribution/story priority	optional caption
Defeat recovery	high	one diagnosis line, never blame/shame	speaker + first preventable event

Voice selection is deterministic from relevance, story ownership, current condition, last-spoken time, and stable character ID. It does not use random rarity or favor the same lead character indefinitely.

7.3 Subtitles And Captions

• Subtitle includes localized speaker, semantic line, and offscreen direction.
• Information-bearing nonverbal sounds use captions: [Bell-Feeder winding up, center lane], [Cache clasp opens], [Anchor pulse reaches clinic].
• Character efforts are not captioned unless they communicate Downed, rescue, or another state not already explicit.
• Subtitle size, background opacity, speaker label, caption detail, and on-screen position are configurable. Captions never cover telegraphs or Directive controls.
• Event log stores critical spoken/caption information and links it to the authoritative event.
• A muted or missing voice file falls back to subtitle and event cue with no timing or gameplay change.

8. SFX Event Catalog

An event identity is not necessarily one file. Runtime sound composes reusable material transients, movement, tail, and one unique semantic sweetener. Unique mechanics retain a distinct cue while common body/weapon movement is shared.

8.1 System And Interface Events

Family	Event IDs	Examples
Navigation and focus	18	focus move, tab/dock, page, drawer, inspect, compare, filter, back, invalid
Confirmation and review	14	review open, legal confirm, cancel, destructive arm, stale rejection, receipt returned
Save and recovery	12	saving, saved, recovered, backup, migration, conflict, blocked; critical states include caption
Resource transaction	20	Crown/material gain/spend, Supply, Cache enter/loss/bank, Strain, Mastery, XP/level
Equipment and craft	20	equip by material, unequip, craft, upgrade, salvage review/commit, unique dormant, attune
Route and world	24	node forecast/select/commit, reroute, close, camp, Anchor, service inactive/start/stable
System/interface event identities	108	variants and material layers not included

8.2 Combat Events

Family	Event IDs	Contract
Core physical actions	32	six weapon/material families, projectile release/arrival, channel, device, movement, forced move
Damage/defense	24	light/heavy/true damage, Guard/Ward gain/absorb/break, block, barrier, immunity, heal, overheal
Status and control	24	each authoritative status family apply/tick/resist/cleanse/expire language
Telegraph and timing	18	arm, countdown thresholds, target change, interrupt, fizzle, Fracture Clock, phase
Directive	16	charge gain/full, legal/illegal, target, review, spend, effect families, duplicate rejection
Downed/rescue/retreat/result	18	Downed, window, rescue, lost, retreat channel/interruption, victory, defeat, diagnosis
Shared combat event identities	132	content-specific cues below layer on top

8.3 Content-Specific Events

Family	Required identities	Reuse boundary
Profession active skills	72	each owns a signature action/impact sweetener; may share material/body layers
Profession resources/passives	36	distinct low/full/spend cadence and restrained trigger
Character talents	30	identity-linked cue; voice may replace it only when cap permits
Enemy mechanics	100	one recognizable forecast/resolution signature per enemy
Boss phase mechanisms	24	minimum three additional phase/Break/defeat cues per boss
Behavior-changing items	160	trigger identity tied to item cause; repeated items rate-limited by unique group
Chapter hazards/services	48	three hazards and three service states per chapter minimum
Content-specific event identities	470	not 470 wholly unique raw recordings; every identity has an audible semantic difference

The complete catalog therefore specifies 108 + 132 + 470 = 710 SFX event identities. A material library, layering, pitch-safe variants, and source grouping reduce raw-file count. They do not reduce the number of gameplay events that require a reviewed audible outcome.

9. Material Library

Reusable source libraries are recorded with close, medium, and distant perspectives where needed:

• ceramic/enamel bell and plate;
• dry wood, wet wood, reed, paper, leaf, cloth, rope;
• clear, frosted, salt, and voice glass;
• brass, copper, iron, multi-metal tool/weapon impacts;
• water fill/drain/surge, wind direction/gust/calm, furnace/pressure/valve;
• pen, seal, abacus, clasp, latch, gear, vane, relay switch;
• six weapon movement/contact families;
• foot movement on eight regional stage materials;
• aether stable/flow/rupture/repair double tone;
• non-musical tonal markers for Command, Break, Hush, Anchor, Cache, Strain.

Pitch variation remains within identity-safe ranges and never changes urgency or suggests a different material. Critical telegraphs and transaction receipts are not randomly pitched.

10. Haptic Relationship

Haptics are optional and never the only cue:

• light: legal selection, equip, node latch;
• medium: Directive commit, Guard/Break, reward bank;
• strong but bounded: boss Break, rescue, chapter restoration;
• no haptic: disabled input, automatic routine, ambient animation, hover/focus;
• vibration-off produces identical event timing and results.

Repeated combat impacts are rate-limited independently from SFX. No sustained vibration follows music or environmental loops.

11. Production And Evidence Workflow

1. Lock the event, authoritative state change, material, scale, timing, bus, cooldown, caption, fallback, and exclusions.
2. Choose owned recording, licensed source, commissioned performance, or explicitly authorized generation. Record commercial license and rights.
3. Produce 2-3 candidates for signature SFX/music states; never accept the first response by default.
4. Machine-check format, duration, silence, peak/clipping, and loop seam proxy.
5. Human A/B listen in the intended game state on speakers and headphones.
6. Process only accepted source; retain rejected decision and reason.
7. Import, route, bind to committed events, and set concurrency/rate limits.
8. Capture real runtime: opening, first action, telegraph, Directive, failure, recovery, reward, restoration, mute, pause/resume, and output change.
9. Perform 10-15 minute fatigue/masking tests per representative music suite.
10. Record evidence states independently:

planned -> sourced/generated -> machine_checked -> human_selected -> processed -> imported -> integrated -> runtime_seen -> human_listened -> device_verified -> release_accepted

No current task authorizes paid SFX, ambience, or music generation. Voice/TTS standing authorization does not imply approval to cast synthetic final voices, clone a person, or replace performer rights review.

12. CH-01 Dynamic Audio Slice

The first presentation slice requires:

Moment	Audio contract
Opening	wind/pump/yard ambience, broken route interval, sparse CH-01 base
Bell-Feeder telegraph	centered mechanical inhale, lane direction marker, critical caption; music yields
Mara intercept	coat/step/shield-enamel layers, Guard absorb, optional relevant bark
Retune selection	Command mechanism arms; illegal/legal states distinct
Retune commit	charge spend, aether interval retunes, event position changes, no fake damage cue
Pressure/failure	knockback chain remains ordered; first divergence audible and logged, no shame sting
Reward to Cache	object material plus open copper clasp; unresolved risk tick
Anchor bank	clasp close and low wood receipt; atomically synchronized
Service restoration	route pulse crosses scene; clinic or tower adds a persistent new ambience layer

The slice music uses CH-01 route, battle, and boss candidates with real base-to- tension-to-resolution transitions. It is reviewed for telegraph masking, voice overlap, repetitive automatic actions, mono information, captions, mute, focus loss, pause/resume, speakers, and headphones.

13. Acceptance Gates

• Thirty character voice identities remain distinct and properly localized; 773 semantic lines and 144 scene bookends have production tracking.
• The 30 music suites have different emotional jobs, regional material, and state transitions; no constant dense loop fills the campaign.
• All 94 ambience configurations reflect place, restoration, and power state.
• All 710 SFX event identities agree with authoritative gameplay and follow concurrency/rate limits.
• Critical telegraph, Directive, rescue, failure cause, Cache, Anchor, save, and destructive confirmation are readable without music and without voice.
• Master mute, captions, mono compatibility, low volume, output change, interruption, background/foreground, and low power preserve play.
• Machine checks, human selection, runtime use, human listening, device, and release evidence remain separate.
• The CH-01 dynamic slice passes masking, fatigue, transition, caption, and state-truth review before bulk audio production.

Human listening, performer casting, commercial licenses, final provider/source, runtime integration, device verification, and release acceptance remain open.