Synthesize every sound with in-repo scripts (tools/audio/): footsteps by surface, weapons, building, water, surges and UI. Compose an original adaptive score from sea-shanty-inspired harmony that builds with the tide.
Acceptance
Sounds are spatialized, and gunfire and footsteps are directional.
Every asset can be regenerated from source and has its provenance recorded.
No sampled or third-party audio, and no melodies from existing songs.
Levels are checked by a loudness script.
**Depends on:** A1, B8
Synthesize every sound with in-repo scripts (`tools/audio/`): footsteps by surface, weapons, building, water, surges and UI. Compose an original adaptive score from sea-shanty-inspired harmony that builds with the tide.
**Acceptance**
- Sounds are spatialized, and gunfire and footsteps are directional.
- Every asset can be regenerated from source and has its provenance recorded.
- No sampled or third-party audio, and no melodies from existing songs.
- Levels are checked by a loudness script.
claude
added this to the v0.6 Art, Audio and UI milestone 2026-11-09 05:28:42 +00:00
tools/audio/: a small synthesis kit (dsp.gd), 50 effect recipes (recipes.gd) and the score (score.gd): an original eight-bar tune and chord sequence in D Dorian, 6/8, written as note lists in the script, rendered as four looping layers (drone, frame drum, concertina, fiddle). make audio regenerates all 54 files into assets/audio/ in about 6 s, with manifest.json (hash, peak, loudness, generator). Provenance is in assets/PROVENANCE.md.
Loudness: make audio-check (tools/audio/loudness.gd) holds each sound to its category's target, keeps peaks under -1 dBFS, and checks the mix order: tide loudest; ambience, music and menus under the effects. The targets are in docs/STYLE.md.
SoundDirector, in the client session: shots (a report per weapon), impacts by what was hit, harvesting, footsteps by surface (grass, sand, rock, built pieces, shallows) for players near the camera, landings, splashes, swimming, building and breaking by material, the tide's bell, foghorn, surge and lighthouse, own hurts, eliminations, and surf and wind beds. Placed sounds play on AudioStreamPlayer3D voices at the event's position, so the camera's listener pans and attenuates them.
MusicDirector fades the layers on an AudioStreamSynchronized: full in menus and results, the drone alone while the first tide holds, then drum, concertina, and the whole tune for the last land. MenuSounds adds hover and press knocks to every menu button.
Tests:make check passes, 578 tests. New: test_audio_assets (8: regenerates every file and compares SHA-256, loudness, provenance, loops, nothing foreign in the folder), test_sound_director (13: positions and per-weapon reports, impacts, stride and alternating feet, surfaces on seed 2's island, distance culling, landing and splash, building, the tide's cues through a whole schedule, hurt and feed, quiet start on a new world, ambience), test_music_director (2), and a menu sounds test.
Not verified: nobody has listened to it. The host's audio devices are off limits, so I checked spectrograms and waveforms instead, not by ear. Panning is the engine's AudioStreamPlayer3D; the tests check voices are placed at the source, not what comes out of the speakers. R6 should listen.
Follow-ups
B8 (explosives) is still deferred. Its blast sound (keg_blast) is ready, and the impact of a powder_keg already plays it.
The Gitea server has no LFS, so .gitattributes now keeps assets/audio/*.wav in plain git (all under 1 MB; check_binaries.sh still guards the size). Anything bigger later will need LFS turned on, which is Eric's call.
Own shots are heard when the server's FX arrives, a round trip late. Predicting them locally would be a small follow-up if it feels laggy in R6.
Done in 2545337.
**What changed**
- `tools/audio/`: a small synthesis kit (`dsp.gd`), 50 effect recipes (`recipes.gd`) and the score (`score.gd`): an original eight-bar tune and chord sequence in D Dorian, 6/8, written as note lists in the script, rendered as four looping layers (drone, frame drum, concertina, fiddle). `make audio` regenerates all 54 files into `assets/audio/` in about 6 s, with `manifest.json` (hash, peak, loudness, generator). Provenance is in `assets/PROVENANCE.md`.
- Loudness: `make audio-check` (`tools/audio/loudness.gd`) holds each sound to its category's target, keeps peaks under -1 dBFS, and checks the mix order: tide loudest; ambience, music and menus under the effects. The targets are in `docs/STYLE.md`.
- `SoundDirector`, in the client session: shots (a report per weapon), impacts by what was hit, harvesting, footsteps by surface (grass, sand, rock, built pieces, shallows) for players near the camera, landings, splashes, swimming, building and breaking by material, the tide's bell, foghorn, surge and lighthouse, own hurts, eliminations, and surf and wind beds. Placed sounds play on `AudioStreamPlayer3D` voices at the event's position, so the camera's listener pans and attenuates them.
- `MusicDirector` fades the layers on an `AudioStreamSynchronized`: full in menus and results, the drone alone while the first tide holds, then drum, concertina, and the whole tune for the last land. `MenuSounds` adds hover and press knocks to every menu button.
**Tests:** `make check` passes, 578 tests. New: `test_audio_assets` (8: regenerates every file and compares SHA-256, loudness, provenance, loops, nothing foreign in the folder), `test_sound_director` (13: positions and per-weapon reports, impacts, stride and alternating feet, surfaces on seed 2's island, distance culling, landing and splash, building, the tide's cues through a whole schedule, hurt and feed, quiet start on a new world, ambience), `test_music_director` (2), and a menu sounds test.
**Not verified:** nobody has listened to it. The host's audio devices are off limits, so I checked spectrograms and waveforms instead, not by ear. Panning is the engine's `AudioStreamPlayer3D`; the tests check voices are placed at the source, not what comes out of the speakers. R6 should listen.
**Follow-ups**
- B8 (explosives) is still deferred. Its blast sound (`keg_blast`) is ready, and the impact of a `powder_keg` already plays it.
- The Gitea server has no LFS, so `.gitattributes` now keeps `assets/audio/*.wav` in plain git (all under 1 MB; `check_binaries.sh` still guards the size). Anything bigger later will need LFS turned on, which is Eric's call.
- Own shots are heard when the server's FX arrives, a round trip late. Predicting them locally would be a small follow-up if it feels laggy in R6.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Depends on: A1, B8
Synthesize every sound with in-repo scripts (
tools/audio/): footsteps by surface, weapons, building, water, surges and UI. Compose an original adaptive score from sea-shanty-inspired harmony that builds with the tide.Acceptance
Done in
2545337.What changed
tools/audio/: a small synthesis kit (dsp.gd), 50 effect recipes (recipes.gd) and the score (score.gd): an original eight-bar tune and chord sequence in D Dorian, 6/8, written as note lists in the script, rendered as four looping layers (drone, frame drum, concertina, fiddle).make audioregenerates all 54 files intoassets/audio/in about 6 s, withmanifest.json(hash, peak, loudness, generator). Provenance is inassets/PROVENANCE.md.make audio-check(tools/audio/loudness.gd) holds each sound to its category's target, keeps peaks under -1 dBFS, and checks the mix order: tide loudest; ambience, music and menus under the effects. The targets are indocs/STYLE.md.SoundDirector, in the client session: shots (a report per weapon), impacts by what was hit, harvesting, footsteps by surface (grass, sand, rock, built pieces, shallows) for players near the camera, landings, splashes, swimming, building and breaking by material, the tide's bell, foghorn, surge and lighthouse, own hurts, eliminations, and surf and wind beds. Placed sounds play onAudioStreamPlayer3Dvoices at the event's position, so the camera's listener pans and attenuates them.MusicDirectorfades the layers on anAudioStreamSynchronized: full in menus and results, the drone alone while the first tide holds, then drum, concertina, and the whole tune for the last land.MenuSoundsadds hover and press knocks to every menu button.Tests:
make checkpasses, 578 tests. New:test_audio_assets(8: regenerates every file and compares SHA-256, loudness, provenance, loops, nothing foreign in the folder),test_sound_director(13: positions and per-weapon reports, impacts, stride and alternating feet, surfaces on seed 2's island, distance culling, landing and splash, building, the tide's cues through a whole schedule, hurt and feed, quiet start on a new world, ambience),test_music_director(2), and a menu sounds test.Not verified: nobody has listened to it. The host's audio devices are off limits, so I checked spectrograms and waveforms instead, not by ear. Panning is the engine's
AudioStreamPlayer3D; the tests check voices are placed at the source, not what comes out of the speakers. R6 should listen.Follow-ups
keg_blast) is ready, and the impact of apowder_kegalready plays it..gitattributesnow keepsassets/audio/*.wavin plain git (all under 1 MB;check_binaries.shstill guards the size). Anything bigger later will need LFS turned on, which is Eric's call.