Skip to content

audio-assets.md

Latest commit

 

History

History
90 lines (61 loc) · 10.7 KB

audio-assets.md

File metadata and controls

90 lines (61 loc) · 10.7 KB
Raw

Bundled Audio Asset Provenance

Audit date: 2026-10-15. Applies to the v1.0.8 working tree and packaged Resources/data, not to older releases or Git history.

All 8 schemes / 53 WAV files now have recorded redistribution terms. The 21 recordings with existing Freesound notices are retained. The other 32 WAVs were replaced in v1.0.8 by original, reproducible mathematical synthesis. The project's MIT license does not relicense retained third-party recordings.

Stable scheme ID Display name WAVs Source and license v1.0.8 action
bubble Bubble 9 Glaneur de sons, Freesound; CC BY 3.0 Retained byte-for-byte
mechanical Mechanical 5 jim-ph, Freesound; CC0 1.0 Retained byte-for-byte
drum Drum 7 Veiler, Freesound; CC0 1.0 Retained byte-for-byte
typewriter Synth Typewriter 10 Tickeys Redux synthesis; project MIT Replaced every WAV
sword Synth Blades 9 Tickeys Redux synthesis; project MIT Replaced every WAV
Cherry_G80_3000 Synth Click 5 Tickeys Redux synthesis; project MIT Replaced every WAV
Cherry_G80_3494 Synth Thock 6 Tickeys Redux synthesis; project MIT Replaced every WAV
starwars Synth Space 2 Tickeys Redux synthesis; project MIT Replaced every WAV

Directory names, file names, scheme IDs, ordering, non_unique_count, and key_audio_map remain compatible with saved settings. Display names describe the replacement sounds. The synthetic packs contain no original recordings, third-party samples, film audio, or real Cherry keyboard recordings, and do not claim to reproduce a particular physical product or film sound.

Retained recording credits

These are adaptations of the recordings listed below: inherited Tickeys assets use renamed files and the Redux import includes volume processing and, for most files, mono mixing. Exact earlier edit recipes and the original-file-to-renamed-file correspondence were not recorded. Complete original pack credits are therefore preserved rather than guessing a one-to-one mapping. No further processing was applied to these 21 WAVs in v1.0.8.

Bubble — Glaneur de sons

Credit: Bubble sounds by Glaneur de sons, from Freesound, adapted for Tickeys / Tickeys Redux. Licensed under Creative Commons Attribution 3.0 Unported.

Mechanical — jim-ph

Credit: Vintage keyboard sounds by jim-ph, from Freesound, adapted for Tickeys / Tickeys Redux. CC0 1.0 Universal.

Drum — Veiler

Credit: Drum sounds by Veiler, from Freesound, adapted for Tickeys / Tickeys Redux. CC0 1.0 Universal.

Provenance evidence and replacement decision

The three download notices match the corresponding files in yingDev/Tickeys commit da238a94bec6545730571e59546aae535b2b69f1 byte-for-byte. The local import is commit 8e14d2a831933223a264ff2a2a05476c2335e962. Each retained WAV was compared with its same-named upstream WAV: all 21 have the same frame count and sample rate. After channel averaging, waveform correlation is at least 0.9996 (at least 0.9999998 except bubble/enter.wav), consistent with inherited gain/channel processing. This establishes the chain to the pack notices; it does not reconstruct undocumented historic edits.

The local import, upstream resource tree, and bounded search did not establish separate recording licenses for the five replaced packs. A repository-wide MIT file alone was not treated as evidence that third-party recordings could be relicensed. Those WAVs are absent from the current resource tree after replacement. Earlier commits/releases are unchanged. No audio files were found under legacy/; packaging uses assets/data/, not legacy/.

Synthetic pack generation

scripts/generate_audio_assets.py uses Python's standard library only, with no network calls or recording inputs. The script and newly generated WAVs use the project MIT license, reproduced in each synthetic pack's license.txt. That declaration does not apply retroactively to removed recordings.

  • Format: uncompressed PCM, signed 16-bit little-endian, mono, 44,100 Hz.
  • Determinism: base seed 0x5449434B, XORed with the first four SHA-256 bytes of scheme_id/file.wav, interpreted little-endian. Explicit xorshift32 supplies noise. Each file's actual seed is in the manifest.
  • A 44-frame attack fade and 441-frame end fade eliminate boundary discontinuities. DC removal uses the same envelope. Peak normalization targets 0.55 full scale, quantized to 18,022 / 32,767, about −5.2 dBFS. This is a per-file peak, not a limit on the application's mixed output.
  • Synth Click: high-passed noise, damped 1,850/820 Hz resonances, and a second 2,350 Hz tap; 110 ms.
  • Synth Thock: damped 235/470 Hz resonances, low-passed noise, and a softer return tap; 140–200 ms.
  • Synth Typewriter: noise attack and 690/2,110 Hz resonances, a delayed mechanism tap, extra scroll taps, and a 1,568/3,136 Hz return bell; 160–400 ms.
  • Synth Blades: filtered-noise swish and inharmonic 1,390/2,377/3,551 Hz resonances; 210–270 ms.
  • Synth Space: two downward frequency sweeps with 71 Hz phase modulation; 200 ms.
  • Frequency variation is deterministic (0.94–1.06 times nominal). Enter/space variants have longer clip durations where their filenames identify those keys.

The synthesis uses platform math functions; byte equality is checked on the build platform rather than promised across every possible math-library implementation.

Inventory and verification

assets/data/audio-manifest.json enumerates every distributed WAV, source/license, SHA-256, channels, sample rate, frames, duration, peak and synthetic seed. It is shipped in the App Bundle. Full CC texts in assets/data/licenses/ come from the official CC BY 3.0 text and CC0 1.0 text. Existing per-pack credits are preserved unchanged.

# Regenerate only the 32 synthetic WAVs, then inventory all 53 WAVs.
python3 scripts/generate_audio_assets.py
# Read-only check of generated bytes, manifest, mappings and WAV coverage.
python3 scripts/generate_audio_assets.py --check

The read-only check fails on missing/extra WAVs, invalid mappings, changed generated bytes, or manifest drift. Decoder testing and subjective listening are separate checks; valid PCM does not establish playback through a user's actual output device.

中文说明

v1.0.8 保留泡泡、机械键盘、鼓声三套共 21 个已有 Freesound 许可记录的音频，并将另外五套共 32 个来源未确认的 WAV 全部替换为程序合成音效。泡泡保留 CC BY 3.0 署名与原始下载记录；机械键盘、鼓声保留 CC0 记录。旧版的真实 Cherry 型号和星球大战显示名已改为“合成脆响／合成闷响／合成太空”等准确名称，旧设置使用的方案 ID 与按键映射继续有效。

合成文件不使用旧录音、电影素材或第三方采样；可从脚本重建，并按项目 MIT 许可证分发。现有第三方音频不因此改为 MIT。完整作者、来源、许可、历史修改说明和每个 WAV 的哈希见上文及随包清单。