compAIser recorded acoustic instrument pack The files in this pack are recordings of physical instruments. The current piano is CC BY 3.0; the organ, strings, winds and original compact piano are CC0 1.0. CURRENT PIANO: Accurate-Salamander Grand Piano 6.2 (Yamaha C5) Original recordings: Alexander Holm. Remaster: Chisato Yamauchi / Accurate-Salamander Project. Piano tuning: Hiroharu Narikawa. Full additional credits, source processing scripts and original license notices: assets/instruments/accurate-salamander-6.2/README.txt and README.orig.txt https://www.ir.isas.jaxa.jp/~cyamauch/AccurateSalamander/ https://github.com/cyamauch/NoctSalamanderGrandPiano License: https://creativecommons.org/licenses/by/3.0/ Pinned archive SHA-256: 03d9a02da210ea0f5968ead11c8e764b3ddde1ea0cba5a0432e1f094905cbb35 Archive URL, original hashes and adaptation details are in assets/instruments/piano-source.json. The bundled source license says CC BY 3.0; the mirror's inconsistent CC BY 4.0 metadata does not replace that notice. The piano supplies 480 main stereo recordings (30 sampled roots, 16 dynamics). Its 1,408 playing-key mappings cover all 88 keys and 127 positive velocities. The 24-bit WAV files are unmodified: no per-note normalization, downsampling, tail truncation or stereo folding. Nominal sample rate is 48 kHz; the source deliberately uses rate headers from 40,410 to 56,950 Hz as part of its tuning. BAIch preserves these headers and the additional per-key SFZ tuning. BAIch adaptations: import the neutral minimum/flat SFZ main-note regions; preserve dynamic boundaries, velocity tracking, key gains and note releases; apply the live mapping's additional 720-frame onset offset to remove nominal 15 ms latency; retain one stereo image for every piano voice; offer mid/side width control. No recorded WAV is altered. The standard non-pedal mapping is used; mechanical noises, half-pedal and sympathetic resonance are not emulated. The bundled auxiliary recordings/SFZs remain available for other compatible open-source players. The author's demonstration performances are not bundled. OTHER INSTRUMENTS AND LEGACY PIANO: CC0 1.0 https://creativecommons.org/publicdomain/zero/1.0/ Legacy Steinway piano and pipe organ: Versilian Community Sample Library (VCSL). https://versilian-studios.com/vcsl/ https://github.com/sgossner/VCSL Pinned commit: c1ea7bcc3c7309650ab0da9d15c9cd1fbc4a4c7e Strings and woodwinds: Versilian Studios Chamber Orchestra Community Edition. https://versilian-studios.com/vsco-community/ https://github.com/sgossner/VSCO-2-CE Pinned commit: 440300901dfe9275fd84e0b7763af1f8443ae62e The retained legacy Steinway B has thirteen pitch roots and two dynamic layers. It is preserved for comparison and provenance, not selected by normal playback. The organ is a recorded quiet pipe-organ registration. Violin is a solo bowed violin; viola and cello are recorded sections, not solo players. Flute, oboe, clarinet and bassoon are recorded woodwinds. No other instrument is relabeled as piano, organ, strings or winds. The three composition studios play these recordings; hAIrror additionally creates procedural sound as described below. Fracta's existing instrument ledger informed the violin/oboe source selection; BAIch downloaded fresh pinned upstream originals and did not modify Fracta. CURRENT CHAMBER BANK: baich-chamber-1, 741 unique stereo recordings, 742 regions. Sam Gossner and Simon Dalzell: recordings; Elan Hickler/Soundemote: sample cutting. Twenty original SFZ mappings are retained under instruments/vsco-chamber/sfz/. SFZ commit: 6dd651d55dde97fd4028699be9d4481f26917891 assets/instruments/vsco-chamber/source-plan.json records pinned Git blob IDs and mapping SHA-256 hashes. Active regions in manifest.json record original and prepared recording SHA-256 hashes, original URLs, gains, tuning and key zones. The author's key centers, not filenames' octave conventions, determine pitch. scripts/install_chamber.py preserves both original stereo channels, full note recordings and native 44.1 kHz sample rates. Preparation trims leading near- silence with a 3 ms pre-roll, adds a 2 ms attack fade and writes PCM24. Sources include both 16- and 24-bit material: PCM24 output does not invent precision. Sustained patches receive a matched loop boundary with an 80 ms crossfade; short and pizzicato recordings do not loop. There is no per-note normalization. Original SFZ dynamic/region gains and one balancing gain per instrument apply during playback. Recordings include sustain/vibrato, staccato or spiccato, two-take variations where supplied, and pizzicato for the three string patches. Not every pitch/articulation has the same dynamic depth or alternate takes. The legacy ledger assets/instruments/source-plan.json and its preparation script remain for the compact bank and organ: 22 Hz DC/highpass cleanup, 32 kHz PCM16, mono ensemble preparation and shorter crossfaded sustain loops. That script preserves installed upgraded piano/chamber banks unless the corresponding --legacy-piano/--legacy-chamber option is explicitly selected. Octave labels are calibrated to sounding pitch: organ/section/wind upstream filenames generally call middle C 'C3'; the Steinway and solo violin use 'C4'. The manifest's integer pitches use middle C = 60 for every instrument. Intermediate pitches use resampled roots; sampled legato bow transitions and breaths are not separately modeled. The piano preserves its complete main-note recording set. The app and offline exporter share the same source recordings, region selection, gain and stereo plan. Offline masters default to 48 kHz/24-bit stereo, with optional 16-bit output. CC0 allows redistribution and modification without attribution. CC BY 3.0 requires the piano attribution/license and indication of adaptations; preserve this ledger and the included original notices when redistributing the bank. hAIrror: PROCEDURAL AND TRANSFORMED SOUND hAIrror combines independently generated oscillator/modal/noise voices with granular and randomized-phase spectral transformations of this pack. Its source catalog (assets/instruments/aimbient-catalog.json) records the selected originals, their hashes, timing, tuning, licenses and credits. Listening crops a bounded region, folds it to mono, resamples it, applies normalization and edge fades, and creates new granular/spectral material. The original audio files remain unchanged. Piano derivatives retain the Alexander Holm / Chisato Yamauchi / Accurate-Salamander attribution and CC BY 3.0 license above; other sources retain the Versilian Studios attribution and CC0 provenance. AImbient: HARMONIC SPECTRAL RESYNTHESIS AImbient shares the licensed catalog above. It extracts the relative strengths of a recording's harmonic partials using its documented root pitch, adds quieter integer harmonics, and resynthesizes that envelope on its own shared harmony. Windowed randomized-phase overlap-add stretches this material into sustained textures. The original recordings are unchanged. The piano attribution and CC BY 3.0 terms above apply to these adaptations; the other recordings retain their Versilian Studios / CC0 provenance. Generic binaural filtering uses an adaptation of the Spatial Audio Framework's Genelec Aural ID KEMAR data, provided upstream by Aki Makivirta and Jaan Johansson. Copyright 2020 Leo McCormack, ISC license. The pinned source, transformations, checksum and complete permission notice are in web/aimbient/spatial/PROVENANCE.txt (also served at /aimbient/spatial/PROVENANCE.txt).