Direction · 4 of 7
Each place has its layer, each person has their voice, and the player controls the volume.
The valley's sound is built from layers of the place: birds of the Recôncavo, sea, stream, campfire, insects by day, and crickets, frogs and owls by night, each layer with its own volume in SETTINGS and some of them positional — the sea at the pier, the stream at the river's two points, the field campfire when it's lit. Music alternates by time of day (morning, afternoon, dusk, night and small hours) with volume transitions, and the deep woods get tense music with whispers. Speech and footsteps are short events, tied to the villager or to the ground the character crosses.
Pedro, the traveler and 21 villagers have voices of their own, generated in ElevenLabs from a text with performance tags separate from the text that appears in the bubble (only mestre Quirino speaks in a bubble); Pedro narrates quests aloud when nearby, and the gravedigger's scoldings are queued so voices do not overlap. Production uses the tools in tools/elevenlabs/, with normalization and origin records, and consumes credits only on explicit request: the game uses the versioned files, and generating audio is not part of opening, running or testing the game.
Sound direction
Each place has its layer, each person has their voice, and the player controls the volume.
The valley's sound is built from layers of the place: birds of the Recôncavo, sea, stream, campfire, insects by day, and crickets, frogs and owls by night, each layer with its own volume in SETTINGS and some of them positional — the sea at the pier, the stream at the river's two points, the field campfire when it's lit. Music alternates by time of day (morning, afternoon, dusk, night and small hours) with volume transitions, and the deep woods get tense music with whispers. Speech and footsteps are short events, tied to the villager or to the ground the character crosses.
Pedro, the traveler and 21 villagers have voices of their own, generated in ElevenLabs from a text with performance tags separate from the text that appears in the bubble (only mestre Quirino speaks in a bubble); Pedro narrates quests aloud when nearby, and the gravedigger's scoldings are queued so voices do not overlap. Production uses the tools in tools/elevenlabs/, with normalization and origin records, and consumes credits only on explicit request: the game uses the versioned files, and generating audio is not part of opening, running or testing the game.
8 principles
What is not negotiable
Layers of the place, each with its own volume
ambiente_vale.gd keeps birds, sea, stream, campfire, insects and crickets as independent layers; each has a volume in the Sounds tab of SETTINGS, and the positional sources (AudioStreamPlayer3D) follow their source's layer: the sea at the pier, the stream at the river's two points, the field campfire when lit.
The forest sings in episodes, not in a loop
The forest loops (day: birds and insects; night: crickets, frogs and owls) crossfade with daylight, but they're shrill: they only come in short episodes of 6 to 12 s, 5 dB below the ambience, separated by pauses of 1 to 2.5 minutes. Birds, sea, stream and bonfire stay continuous.
Music by period and by place
Music in the morning, afternoon, dusk, night and small hours, with crossfades, a great kiskadee by day, and tense music with whispers in the deep woods. The menu has four tracks to choose from and two sets of button sounds (Original and Wood). The menu's music and ambience only start when the menu appears — the loading screen is silent.
One voice per person, and the text read isn't the text spoken
Pedro and the 21 voiced villagers have greetings and conversations that alternate at each encounter; the bubble shows the text and the voice reads the tts, the same text with performance tags for the eleven_v3 model (such as [sighs] and [whispers]). One voice per person: Weverton (Pedro), Manoel Lopes (Benedito), Edna (Zefa), Matheus Energetic (Cosme), Matheus Santos (Tonho), Lucinda (Filó), Raquel (Candinha) and Matheus Clear (Damião, also in the three cemetery scoldings, each angrier than the last); the fourteen villagers who came later have one too. Only mestre Quirino speaks in a bubble. The crossing narration remains in Portuguese in the four menu languages.
Footsteps and actions are short events tied to the ground
Footsteps follow the animation's rhythm and the ground underfoot: grass, earth, sand, wood (pier, bridge, canoe), shallow and deep water, mud, puddle and the stroke while swimming; running has its own sound per terrain; foam on the surface around whoever walks or swims. The 2D per-action effect table (axe, hoe, door, planting, watering, picking up) migrates to 3D as each system gains an action.
The player controls the volume, and the choice persists
The speaker icon in the top-right corner turns audio on and off; the options panel brings together music, effects and ambience, the valley layers, the menu tracks and footsteps in water (with a Listen button). Preferences are saved automatically to local preferences, and the valley volumes follow the Audio.volumes_alterados signal.
Generating takes an explicit request; the game uses what's versioned
The tools in tools/elevenlabs/ keep the production log and read local credentials through tools/comum/chaves.ps1. Generating speech, music or effects consumes credits and requires an explicit request; there is no need to generate audio to open, run or test the game. Provenance and credits live in assets/CREDITOS.md and in the ORIGEM records.
One audio catalog, two games
assets/audio/ is the catalog shared with the 2D project, including sources and metadata, and scripts/autoload/audio.gd is the same manager, independent of dimensionality. The 3D receives copies at the same relative paths (no symbolic links or dependency on another repository) and keeps its own preferences directory (MythsValleyPrototype3D), without overwriting the 2D one.
6 rules
In practice, day to day
- 24 s ambience loops in assets/audio/ambiente/ (mata_dia, mata_noite, riacho, fogueira), from ElevenLabs Sound Effects; aves_reconcavo and mare_mansa are the menu ambiences.
- Voices in assets/audio/vozes/<id>_fala_<n>.mp3, generated by tools/elevenlabs/gerar-falas-moradores.ps1; Damião's scoldings use stability 0.5 → 0.35 → 0.2.
- Manager API calls: Audio.efeito("plantar"), Audio.passo("terra"), Audio.tocar_musica(); all audio stays available, including what 3D gameplay doesn't use yet.
- The regression test tests/smoke_opening.gd checks that the 31 runtime audio files load; this doesn't replace listening to and judging the mix on the target computer.
- Technical reuse doesn't change the original licenses of the sources in assets/audio/fontes (a folder with .gdignore, outside the game).
- No key appears in any versioned log; credentials live in local files read by tools/comum/chaves.ps1.
Where it shows up
- The collection
Music by time of day and place, introduction and menu themes, ambience, effects, villagers' voices and narration. The soundtrack page brings together the published collection for listening; its cards record duration, origin and credits. - Damião's scoldings
Anyone who climbs onto a cemetery slab hears the gravedigger, angrier each time (damiao_bronca_1..3); on the third, he knocks the player down into the aisle between the rows. He only complains if he's in the cemetery, and he forgets after 90 s. - Pedro voices the quests
The 16 arrival steps and the farm steps have audio (pedro_guia_correr, pedro_fazenda_fazenda_ida…); at dusk he warns that it is going to get dark (pedro_anoitecer). Each of his lines has a voice, always Weverton's. - Footsteps in water and the shark
Footsteps in shallow and deep water, mud, puddle and swimming, plus the shark attack (tubarao_ataque.mp3), generated in ElevenLabs by tools/elevenlabs/gerar-efeitos-3d.ps1; the "Footsteps in water" option in AJUSTAR has a Listen button.
Where it comes from
The documents behind this area
- Sound strategy — music by time of day and ambient layers
- Valley audio — music, ambience, footsteps and voices
- The living valley — day and night, residents and the valley's sound
- Opening and audio — the crossing, the pier arrival and the opening sound
- 3D changelog — what changed in each build
- Development history — the steps before the standalone repository
- Asset credits — authorship and source of each piece