Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
15 commits
Select commit Hold shift + click to select a range
448f2d4
dsp(sonku P8): keep original stereo LF pressure while highband remain…
masarray Nov 14, 2026
11098cc
fix(sonku P8): route ROOM OFF through LF transparency and keep 0%-wet…
masarray Nov 14, 2026
a4ab0a3
build(sonku P8): compile-gate LF stereo transparency and identical na…
masarray Nov 14, 2026
a158e5e
fix(sonku P8): compile-safe no-op low-band symbols when experiment di…
masarray Nov 14, 2026
1db0368
test(sonku P8): objective 1/3-octave-like frequency sweep, LF stereo …
masarray Nov 14, 2026
95b7161
bridge(sonku P8): expose physical experimental LF transparency only v…
masarray Nov 14, 2026
34bf225
ux(sonku P8): clearly identify the real opt-in transparent-stereo-bas…
masarray Nov 14, 2026
a820d95
docs(sonku P8): isolate bass/room physical trial from unproven dearVR…
masarray Nov 14, 2026
4b48556
ci(sonku P8): objective PCM acoustic transfer evidence gate plus nati…
masarray Nov 14, 2026
7c71abf
test(sonku P8): respect clickless MIX smoothing before asserting bit-…
masarray Nov 14, 2026
1d0562f
dsp(sonku P8.1): derive bounded common-ear timbre EQ from measured KE…
masarray Nov 14, 2026
8ad550f
dsp(sonku P8.1): apply measured common-ear 2k resonance and high-air …
masarray Nov 14, 2026
1b46871
test(sonku P8.1): reject measured 2k honk and assert offline HRIR-der…
masarray Nov 14, 2026
0f70ba6
docs(sonku P8.1): record real KEMAR measured spectrum and data-derive…
masarray Nov 14, 2026
d2d8fb0
fix(sonku P8 CI): hash and package the SAME exact field FAT guide
masarray Nov 14, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
214 changes: 214 additions & 0 deletions .github/workflows/sonku-p80-acoustic-windows-field.yml
Original file line number Diff line number Diff line change
@@ -0,0 +1,214 @@
name: SonKu P8.0 measurable stereo-bass and room native field candidate
# Isolated opt-in full JUCE app build; no public installer, tag or release.
on:
pull_request:
types: [opened, reopened, synchronize]
paths:
- '.github/workflows/sonku-p80-acoustic-windows-field.yml'
- 'src/Main.cpp'
- 'src/SonKuNativeBridgeContract.h'
- 'src/LiveAudioEngine.h'
- 'engine/builtin_fx/SonKuSpatialProcessor.h'
- 'engine/builtin_fx/sonku/**'
- 'site/preview-v2/assets/mixer-native-sonku.js'
- 'CMakeLists.txt'
- 'tests/sonku_p72_*'
- 'tests/sonku_p75_*'
- 'tests/sonku_p80_*'
workflow_dispatch:
permissions:
contents: read
concurrency:
group: sonku-p80-physical-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
jobs:
preflight:
name: Isolated fast PCM and GUI preflight
runs-on: ubuntu-latest
timeout-minutes: 8
steps:
- uses: actions/checkout@v4
with:
ref: ${{ github.event.pull_request.head.sha || github.sha }}
persist-credentials: false
- name: Native JS and source GUI capability checks
run: |
set -euo pipefail
node --check site/preview-v2/assets/mixer-native-sonku.js
node tests/sonku_p73_gui_contract_test.js
grep -q "'FIELD_PREVIEW_FULL_CTEST_NOT_RUN'" scripts/m3e_fat_evidence_gate.mjs
python -m unittest discover -s tests -p native_peq_assembly_test.py
- name: 44.1 and 48 kHz prepared PCM source regression
run: |
set -euo pipefail
g++ -std=c++20 -O2 -Wall -Wextra -Werror -pedantic \
-DSONKU_P7_MEASURED_ELEVATION_RESEARCH=1 \
-DSONKU_P71_DIRECT_XYZ_RESEARCH=1 \
-DSONKU_P72_NATIVE_XYZ_RT=1 \
tests/sonku_p72_realtime_xyz_test.cpp -o /tmp/sonku-p75-pcm
/tmp/sonku-p75-pcm
g++ -std=c++20 -O2 -Wall -Wextra -Werror -pedantic \
-DSONKU_P7_MEASURED_ELEVATION_RESEARCH=1 \
-DSONKU_P71_DIRECT_XYZ_RESEARCH=1 \
-DSONKU_P72_NATIVE_XYZ_RT=1 \
-DSONKU_P75_MUSIC_SAFE_RESEARCH=1 \
tests/sonku_p75_stereo_integrity_test.cpp -o /tmp/sonku-p75-music
/tmp/sonku-p75-music
g++ -std=c++20 -O2 -Wall -Wextra -Werror -pedantic \
-DSONKU_P7_MEASURED_ELEVATION_RESEARCH=1 \
-DSONKU_P71_DIRECT_XYZ_RESEARCH=1 \
-DSONKU_P72_NATIVE_XYZ_RT=1 \
-DSONKU_P75_MUSIC_SAFE_RESEARCH=1 \
-DSONKU_P80_TRANSPARENT_BASS_RESEARCH=1 \
tests/sonku_p80_tonal_room_evidence_test.cpp -o /tmp/sonku-p80-evidence
/tmp/sonku-p80-evidence | tee /tmp/sonku-p80-acoustic-evidence.csv
grep -q '^PASS P8:' /tmp/sonku-p80-acoustic-evidence.csv
windows-field:
name: Actual full Windows JUCE executable and verified portable field ZIP
needs: preflight
runs-on: windows-2022
timeout-minutes: 55
steps:
- uses: actions/checkout@v4
with:
ref: ${{ github.event.pull_request.head.sha || github.sha }}
persist-credentials: false
- name: Verify immutable PR head and physical candidate identity
shell: pwsh
env:
EXPECTED_SHA: ${{ github.event.pull_request.head.sha || github.sha }}
run: |
$actual = (& git rev-parse HEAD).Trim()
if ($LASTEXITCODE -ne 0 -or $actual -ne $env:EXPECTED_SHA -or $actual -notmatch '^[0-9a-f]{40}$') {
throw "SONKU_P80_SOURCE_MISMATCH:$actual"
}
Write-Host "SONKU_P75_EXACT_SOURCE=$actual"
- uses: actions/checkout@v4
with:
repository: juce-framework/JUCE
ref: 8.0.15
path: third_party/JUCE
persist-credentials: false
- name: Obtain pinned WebView2 compilation SDK
shell: pwsh
run: |
nuget install Microsoft.Web.WebView2 -Version 1.0.3485.44 -OutputDirectory .deps -NonInteractive
if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }
- name: Configure full native app with research XYZ explicit ON
shell: pwsh
run: |
cmake -S . -B build -G "Visual Studio 17 2022" -A x64 `
-DJUCE_ROOT="$env:GITHUB_WORKSPACE/third_party/JUCE" `
-DJUCE_WEBVIEW2_PACKAGE_LOCATION="$env:GITHUB_WORKSPACE/.deps" `
-DARMIX_SONKU_EXPERIMENTAL_NATIVE_XYZ_RT=ON `
-DARMIX_SONKU_EXPERIMENTAL_MUSIC_SAFE=ON `
-DARMIX_SONKU_EXPERIMENTAL_P8_TRANSPARENT_BASS=ON `
-DBUILD_TESTING=ON
if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }
$cache = Get-Content -LiteralPath build/CMakeCache.txt -Raw
if ($cache -notmatch 'ARMIX_SONKU_EXPERIMENTAL_NATIVE_XYZ_RT:BOOL=ON' -or
$cache -notmatch 'ARMIX_SONKU_EXPERIMENTAL_MUSIC_SAFE:BOOL=ON' -or
$cache -notmatch 'ARMIX_SONKU_EXPERIMENTAL_P8_TRANSPARENT_BASS:BOOL=ON') {
throw "SONKU_P80_RESEARCH_NOT_ENABLED"
}
$tests = (& ctest --test-dir build -C Release --show-only=json-v1 | Out-String | ConvertFrom-Json)
if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }
$required = @('NativeSonKuP72AtomicXyzLiveEditSource',
'NativeSonKuP72JuceV3XyzLiveAckV1V2Restore',
'NativeSonKuSessionV1V2AndAudioAppliedAck',
'NativeSonKuP75MusicStereoPhaseNoNullAndChunkParity',
'NativeSonKuP80TransparentBassStereoPhaseAndRoomDiagnostic')
$available = @($tests.tests | ForEach-Object { $_.name })
foreach ($test in $required) {
if ($test -notin $available) { throw "SONKU_P80_MISSING_CTEST:$test" }
}
- name: Compile actual ARMixConsole Windows GUI and focused real JUCE tests
shell: pwsh
run: |
cmake --build build --config Release --parallel 2 --target `
ARMixConsole ARMixSonKuP72SourceTests ARMixSonKuP72NativeTests ARMixSonKuNativeStateTests ARMixSonKuP75MusicIntegrityTests ARMixSonKuP80TonalEvidenceTests
if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }
$exe = 'build/ARMixConsole_artefacts/Release/AR MIX Studio.exe'
if (-not (Test-Path -LiteralPath $exe -PathType Leaf)) { throw "SONKU_P75_NO_GUI_EXE" }
if ((Get-Item -LiteralPath $exe).Length -lt 4MB) { throw "SONKU_P75_WRONG_GUI_TARGET" }
- name: Scoped native XYZ state, callback ACK and legacy restoration
shell: pwsh
run: |
ctest --test-dir build -C Release --output-on-failure `
-R '^NativeSonKu(P72AtomicXyzLiveEditSource|P72JuceV3XyzLiveAckV1V2Restore|SessionV1V2AndAudioAppliedAck|P75MusicStereoPhaseNoNullAndChunkParity|P80TransparentBassStereoPhaseAndRoomDiagnostic)$'
if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }
- name: Package real native portable mixer, never an installer
shell: pwsh
run: |
& ./scripts/package_native_windows.ps1 -BuildDir build -OutputDir dist
if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }
python scripts/factoryfx_release_bundle.py stage --lock config/factoryfx-approved-releases.json --output dist --best-effort
if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }
- name: Record exact experiment and physical qualification pending
shell: pwsh
run: |
$path = 'dist/BUILD_PROVENANCE.json'
$p = Get-Content -LiteralPath $path -Raw -Encoding utf8 | ConvertFrom-Json
$sha = (& git rev-parse HEAD).Trim()
if ($p.sourceCommit -ne $sha) { throw "SONKU_P80_ARTIFACT_SHA_MISMATCH" }
$p.fatStatus = 'FIELD_PREVIEW_FULL_CTEST_NOT_RUN'
$p | Add-Member -NotePropertyName sonkuExperiment -NotePropertyValue @{
enabled = $true
nativeXYZ = $true
musicSafeStereo = $true
p8TransparentBass = $true
p8LowBandCrossoverHz = 220
musicSafeAlgorithm = 'MEASURED_CENTER_MID_PLUS_UNCONDITIONAL_STEREO_SIDE_RESEARCH'
stateSchema = 3
fullWindowsExecutable = $true
publicRelease = $false
physicalAsioQualified = $false
dearVrPerceptualParityProven = $false
qualification = 'SONKU_P80_FULL_JUCE_APP_5_SCOPED_CTEST_PASS_REAL_ASIO_LISTENING_PENDING'
licenseRedistributionApproval = $false
mitKemarNoticeSha256 = (Get-FileHash -LiteralPath docs/THIRD_PARTY_8D_SONKU_HRTF.md -Algorithm SHA256).Hash.ToLowerInvariant()
fieldFatGuideSha256 = (Get-FileHash -LiteralPath docs/SONKU_P80_STEREO_BASS_ACOUSTIC_EVIDENCE_20261114.md -Algorithm SHA256).Hash.ToLowerInvariant()
} -Force
$p | ConvertTo-Json -Depth 20 | Set-Content -LiteralPath $path -Encoding utf8
- name: Independent GUI PE, approved resources, provenance and ZIP verification
shell: pwsh
run: |
& ./scripts/verify_native_windows_artifact.ps1 -OutputDir dist
if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }
$p = Get-Content -LiteralPath dist/BUILD_PROVENANCE.json -Raw | ConvertFrom-Json
if (-not $p.sonkuExperiment.enabled -or
-not $p.sonkuExperiment.musicSafeStereo -or
-not $p.sonkuExperiment.p8TransparentBass -or
$p.sonkuExperiment.stateSchema -ne 3 -or $p.sonkuExperiment.publicRelease) {
throw 'SONKU_P75_WRONG_FIELD_ARTIFACT'
}
- name: Include mandatory MIT KEMAR notice and physical FAT guide
shell: pwsh
run: |
# Standard distribution allowlist was verified above. Field-only notice.
Copy-Item -LiteralPath docs/THIRD_PARTY_8D_SONKU_HRTF.md -Destination dist/THIRD_PARTY_8D_SONKU_HRTF.md
Copy-Item -LiteralPath docs/SONKU_P80_STEREO_BASS_ACOUSTIC_EVIDENCE_20261114.md -Destination dist/SONKU_P80_FIELD_FAT.md
$p = Get-Content -LiteralPath dist/BUILD_PROVENANCE.json -Raw | ConvertFrom-Json
$checks = @{
'THIRD_PARTY_8D_SONKU_HRTF.md' = [string]$p.sonkuExperiment.mitKemarNoticeSha256
'SONKU_P80_FIELD_FAT.md' = [string]$p.sonkuExperiment.fieldFatGuideSha256
}
foreach ($item in $checks.GetEnumerator()) {
$file = Join-Path dist $item.Key
if ($item.Value -notmatch '^[0-9a-f]{64}$' -or
-not (Test-Path -LiteralPath $file -PathType Leaf)) {
throw "SONKU_P75_FIELD_NOTICE_MISSING:$($item.Key)"
}
$actual = (Get-FileHash -LiteralPath $file -Algorithm SHA256).Hash.ToLowerInvariant()
if ($actual -ne $item.Value) { throw "SONKU_P75_FIELD_NOTICE_HASH_MISMATCH:$($item.Key)" }
}
if (Test-Path -LiteralPath dist/Studio) { throw 'SONKU_FIELD_USER_DATA_MUST_NOT_SHIP' }
Write-Host 'SONKU_P75_FIELD_READY: GUI EXE + verified MIT notice + FAT guide'
- uses: actions/upload-artifact@v4
name: Upload isolated SonKu Windows x64 field candidate
with:
name: ArSonMix-SonKu-P80-Bass-Room-Research-Windows-x64-${{ github.run_id }}
path: dist/
if-no-files-found: error
retention-days: 21
compression-level: 0
29 changes: 28 additions & 1 deletion CMakeLists.txt
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,15 @@ if(ARMIX_SONKU_EXPERIMENTAL_MUSIC_SAFE AND
NOT ARMIX_SONKU_EXPERIMENTAL_NATIVE_XYZ_RT)
message(FATAL_ERROR "P7.5 music-safe requires explicit native XYZ research ON")
endif()
# Research-only P8 acoustic diagnostic. Does NOT change public LIGHT,
# shipping presets, or P7.5 without explicit operator build flags.
option(ARMIX_SONKU_EXPERIMENTAL_P8_TRANSPARENT_BASS
"Enable P8 original stereo low-frequency pressure with measured high-band HRTF"
OFF)
if(ARMIX_SONKU_EXPERIMENTAL_P8_TRANSPARENT_BASS AND
NOT ARMIX_SONKU_EXPERIMENTAL_MUSIC_SAFE)
message(FATAL_ERROR "P8 requires the isolated P7.5 music-safe research profile")
endif()
set(CMAKE_CXX_STANDARD 20)
set(CMAKE_CXX_STANDARD_REQUIRED ON)

Expand Down Expand Up @@ -217,7 +226,8 @@ target_compile_definitions(ARMixConsole PRIVATE
$<$<BOOL:${ARMIX_SONKU_EXPERIMENTAL_NATIVE_XYZ_RT}>:SONKU_P7_MEASURED_ELEVATION_RESEARCH=1>
$<$<BOOL:${ARMIX_SONKU_EXPERIMENTAL_NATIVE_XYZ_RT}>:SONKU_P71_DIRECT_XYZ_RESEARCH=1>
$<$<BOOL:${ARMIX_SONKU_EXPERIMENTAL_NATIVE_XYZ_RT}>:SONKU_P72_NATIVE_XYZ_RT=1>
$<$<BOOL:${ARMIX_SONKU_EXPERIMENTAL_MUSIC_SAFE}>:SONKU_P75_MUSIC_SAFE_RESEARCH=1>)
$<$<BOOL:${ARMIX_SONKU_EXPERIMENTAL_MUSIC_SAFE}>:SONKU_P75_MUSIC_SAFE_RESEARCH=1>
$<$<BOOL:${ARMIX_SONKU_EXPERIMENTAL_P8_TRANSPARENT_BASS}>:SONKU_P80_TRANSPARENT_BASS_RESEARCH=1>)
if(WIN32)
# Windows retains its proven ASIO-first / WASAPI / static WebView2 distribution.
# Never link Windows loader or sound APIs into macOS/Linux builds.
Expand Down Expand Up @@ -348,6 +358,23 @@ if(BUILD_TESTING)
SONKU_P75_MUSIC_SAFE_RESEARCH=1)
add_test(NAME NativeSonKuP75MusicStereoPhaseNoNullAndChunkParity
COMMAND ARMixSonKuP75MusicIntegrityTests)
if(ARMIX_SONKU_EXPERIMENTAL_P8_TRANSPARENT_BASS)
# The native JUCE adapter and the PCM fixture must compile identical
# P8 DSP flags. The default P7.5/ship branch retains its original DSP.
target_compile_definitions(ARMixSonKuP72NativeTests PRIVATE
SONKU_P80_TRANSPARENT_BASS_RESEARCH=1)
add_executable(ARMixSonKuP80TonalEvidenceTests
tests/sonku_p80_tonal_room_evidence_test.cpp)
target_compile_features(ARMixSonKuP80TonalEvidenceTests PRIVATE cxx_std_20)
target_compile_definitions(ARMixSonKuP80TonalEvidenceTests PRIVATE
SONKU_P7_MEASURED_ELEVATION_RESEARCH=1
SONKU_P71_DIRECT_XYZ_RESEARCH=1
SONKU_P72_NATIVE_XYZ_RT=1
SONKU_P75_MUSIC_SAFE_RESEARCH=1
SONKU_P80_TRANSPARENT_BASS_RESEARCH=1)
add_test(NAME NativeSonKuP80TransparentBassStereoPhaseAndRoomDiagnostic
COMMAND ARMixSonKuP80TonalEvidenceTests)
endif()
endif()
endif()
# Pure control-plane first-run routing selection: no JUCE, drivers,
Expand Down
38 changes: 38 additions & 0 deletions docs/SONKU_P80_STEREO_BASS_ACOUSTIC_EVIDENCE_20261114.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,38 @@
# SonKu P8.0 — Acoustic objective evidence and stereo-bass field test

**Experimental only. NOT dearVR-equivalent or approved for public release.**
User headphone feedback: P7.4 severe stereo cancellation; P7.5 shrill/prominent mid, thin bass, flat room, weak natural externalization.

## Code changes and limits
- Only opt-in `ARMIX_SONKU_EXPERIMENTAL_P8_TRANSPARENT_BASS=ON` together with P7.5 and native XYZ research. The default LIGHT/public plugin is untouched.
- Preserve original channel-specific low-frequency pressure through a prepared 220 Hz complementary one-pole crossover, while spatializing its remaining upper band through the measured center KEMAR HRTF. This is **not** a guessed +dB bass boost, but an original-signal transparency path.
- Return the highpass-only SIDE in the P7.5 protection bed, never doubling low frequency after restoring original L/R. At WIDTH 60° and ROOM OFF, a fully inverted stereo program should pass through at exact user trim without cancellation.
- Use actual x0/x1 stereo source inputs for the six physical early reflection images and the eight-line late diffuse field (instead of two duplicated MID inputs). Their room size/delay and source angular spread now follow the two inputs; sound-quality realism and rearward localization still need independent headphone evaluation.
- Remove prior shelf on the new experimental path so ROOM ON/OFF cannot activate a spectral boost. Dry-only MIX=0 should be mathematically exact independent of GAIN.
- No FFT, allocations, locks, background worker, extra bridge graph, or extra per-insert data movement in the realtime callback. Original KEMAR 140-tap FIR and native JUCE/ACK persist.
- **Still not a fully spatialized stereo object renderer:** above-crossover stereo SIDE remains a direct width bed. This is a targeted bass/room diagnostic milestone, not completion of P8.1 or P8.2.

## P8.1 measured HRTF tonal calibration (same P8 opt-in)

Observed native P8.0 transfer at 48k, FRONT, ROOM OFF, duplicated mono:
63Hz 0.670, 125Hz 0.516, 1kHz 0.614, **2kHz 1.521**, 4kHz 1.303 and **8kHz 0.145** relative to original. These data justify the owner's "mid honking, thin/squeaky music" report independently from ASIO buffer correctness.

The independently licensed MIT KEMAR *horizontal* bank was examined: at 48k its geometric mean ear magnitude across all 37 azimuths and 2 ears is 0.677 at 1k, 1.375 at 2k, 0.958 at 4k, 0.261 at 8k. A measured broad 2kHz surplus is real; a simple +bass shelf cannot fix it.

A new `SonKuHrtfTonalCalibration.h` computes a horizontal-bank *common-ear* corrective EQ ONCE at prepare; no FFT/sin/cos/log/alloc on the realtime callback:
- Presence cut calculated as the measured 2k mean relative to geometric mean of 1k and 4k: capped between 0 and -5.5dB; applied as original RBJ peaking minimum-phase biquad centered at 2.2k, Q 0.85.
- High-air shelf calculated from only **40%** of the measured 1k-vs-8k deficit, maximum +4dB, at 6.2k. This deliberately does not erase the localized KEMAR high-frequency pinna notch.
- Same scalar coefficients for both ears on DIRECT HRTF ONLY, preserving frequency-wise interaural amplitude ratio and nearly identical group-delay shift at the two ears. Never equalize dry stereo SIDE, original bass, directional ER or late reverb with this common ear filter.
- All EQ state is reset on a prepared rebind. Dynamics and plugin state v1/v2/v3 remain unchanged. P7.5/default when P8 flag OFF remains byte-identical.

This is **not** a patented dearVR Clarity algorithm, not a universal headphone correction, not diffuse-field calibration and not a perfect spectral match. Tests now reject a FRONT 2k gain > 1.30 (P8.0 baseline ~1.52), document 8k return, and still require actual music listening.

## Bench and acceptance
- `tests/sonku_p80_tonal_room_evidence_test.cpp` records per-signal RMS transfer vs 63/125/250/500/1k/2k/4k/8k at 44.1 and 48k; hard-left and dual-mono; room OFF and ROOM ON impulse energy. It enforces finite, nonclipped samples, dry bypass equivalence, anti-phase energy, width 60° exact anti-phase waveform, room late tail and chunk partition parity.
- Log the source SHA and compile flags; ensure CTest source fixture and native JUCE adapter share P8 flag. Output metrics from synthetic waves are diagnostic ONLY, not target equalization, no real-music spectral score, no headphone claim.
- For safe physical evaluation: use **new isolated Windows tester**, start with **headphone volume low**, one P8 SonKu insert, ROUTE MAIN confirmed, Master Mono OFF, ROOM OFF, ORBIT OFF, SPATIAL MIX 100%, SIDE WIDTH 60°, GAIN -6 dB. Compare true plugin bypass vs P8 ROOM OFF, then ROOM Square 7.2m with Reflections 55 and Ambience 30, and afterwards dearVR with Bass Boost OFF and Clarity reported. Same licensed local track, output level matched and identical headphones/ASIO.
- A/B on (1) bass-heavy stereo drums/bass, (2) vocals, (3) acoustic reverb tails, (4) hard-left signal, (5) anti-phase high frequency and (6) wide chorus; *record source/output PCM dry and wet* if still thin, nasal or flat. If bypass itself sounds wrong, inspect channel mode and WIRE/MAIN/ASIO mapping first.
- Do NOT publish a Windows release or merge the original draft P7.4/P7.5 simply because synthetic CI passes. Further P8.1 full stereo positioning and P8.2 reference-calibrated broadband room remain mandatory for any dearVR-comparable claim.
- MIT KEMAR attribution in `docs/THIRD_PARTY_8D_SONKU_HRTF.md` is required in physical artifacts.

Architecture gate: `docs/SONKU_P8_DEARVR_BENCHMARK_ACOUSTIC_ARCHITECTURE_20261114.md`. The goal is original, independent implementations benchmarked to dearVR, not copying proprietary IR/Clarity algorithms.
Loading
Loading