Sound and Voice

Published Wiki page · Versioned source · Suggest a source edit

Game sound

Arkline's game sound is generated procedurally from confirmed simulation state. It is machinery and event feedback, not random music or a recording of an unrelated world. No external sound library or music asset is introduced by this implementation.

The audio scene derives engine intensity from measured acceleration, mining loops from active confirmed modules/drone cycles, and travel sound from authoritative warp phases. Connection, docking, lock, extraction and other supported transitions can produce short cues. A confirmed nearby warp arrival within 200 km in the same system also produces a cue; stale arrivals do not replay. Repeated snapshots must not repeat rewards or their sounds; stale/disconnected state silences continuous machinery rather than implying that unobserved work continues. Sounds do not submit gameplay commands or award ore.

Event signatures

The reactor bed stays quiet under other events. Engine harmonics follow measured thrust; coasting does not produce a full-power engine loop. Mining has a spin-up, a restrained resonant loop, a short confirmed extraction cue and a spin-down. Continuous cycles do not replay start/stop cues each time a cycle rolls over. Warp uses related low-frequency layers for charge, entry, travel and arrival. Entry is emitted once when the confirmed trajectory crosses its warmup boundary; repeated frames, reconnects and stale samples do not replay it. Docking, drone operations, target lock, low capacitor and full cargo retain distinct signatures.

All layers are synthesized from original oscillator/envelope definitions. Filtered mechanical texture stays behind the tonal layers. A fixed voice pool, gain smoothing and limiter retain callback bounds and leave headroom for overlapping events.

Controls and output

Sound preferences provide Master, Ambience and Effects levels, a game-sound mute, Test sound and Retry audio. Defaults are 45%, 35% and 65%, with mute off. The settings are stored locally in sound.json in the client's data directory, using an atomic file replacement. Invalid, incomplete, oversized or out-of-range files fall back to defaults; invalid values are never saved. These preferences do not change server state.

The engine uses the system's default output device. Output initialization occurs on an audio worker; an unavailable device reports a status instead of making game connection depend on audio. Retry audio attempts to open the current default device again. Select speakers or headphones in the operating system. Explicit in-game output-device selection is not included.

Test sound is for checking the current output and gain path. Mute and zero levels must still be honored; it is not a bypass of saved settings. Normal game audio uses bounded cue storage, a fixed voice pool, smoothing and an output limiter. Those implementation limits do not establish that any particular speaker volume is comfortable; start with a low system output level.

Voice and remote ships

Voice input and spoken replies have separate controls. The mixer supports lowering game audio during spoken replies (22% of its normal gain). During microphone capture, the client mutes game audio to keep it out of the recognition input; ending capture retains the user's saved mute and volume choices. Speech recognition/output behavior and permissions remain separate from game-sound availability. No microphone is needed for game audio.

The scene planner accepts a single source pilot. The client uses the ship currently under confirmed control: normally the authenticated pilot's ship, or the authorized remote ship while its manual-control lease is active. Merely watching a remote camera does not switch the entire cockpit's sound source. The planner resets transition history when the source changes so another ship's old extraction or arrival does not replay as a new event. Returning to the owner's ship restores its sound source. Packaged validation should exercise these transitions; separate spatial feeds or remote audio streams are not claimed.

Validation boundary

Preference tests cover round-trip storage, invalid-file fallback and failed-save preservation. Scene/mixer tests can validate state transitions, bounded samples and mix rules without opening an audio device. They cannot prove audible output, latency, sound quality, voice ducking or device switching. Record those separately with a native playback test on the packaged build. No microphone capture is part of the settings tests.

Linux source builds need ALSA development headers (libasound2-dev on the CI Ubuntu image), alongside the existing window/OpenGL dependencies. Adding this CI prerequisite does not by itself establish that the hosted runner has usable audio hardware or that Linux playback was heard.