- AquaSynthDaemonService owns compile/render/stream/live-session execution for patch.compile, instrument.sample, bounded instrument.stream automation, instrument.open, instrument.control, instrument.block, and instrument.close. It delegates DSP to AquaSynth.Faust, writes compile/session/render/stream/live/live-session-state/operator .cc witnesses under the configured .aquasynth store, writes .f32/.wav sample artifacts, keeps native streaming patch instances for live sessions, rewrites retained aquasynth.live_instrument_session.v1 state for open/closed status, current controls, next sequence, and last command/receipt ids, releases live sessions without deleting reusable compiled patch witnesses, and emits CultMesh-compatible audio/control stream descriptors plus per-block packet receipts. AquaSynthCultNetDaemon owns the local CultNet command-observation boundary:typed command documents enter the CultNet database, the daemon serializes execution through AquaSynthDaemonService, and typed receipt/provider documents come back out. JSON-lines modes are only operator/xeno edges; they must not become second audio/session authorities.
- AquaSynthDingsHost is the single per-user playback owner:one named shared-mode WASAPI session, one mixer, one Aqua daemon child, and the live Windows session master volume/mute authority. AquaSynthDingsMcp instances are typed named-pipe bridges only and contain no audio output or fallback mixer. Event IDs own the Minimal Dings motif; curated instrument IDs own agent timbre; AquaSynthDaemon remains the synthesis authority. The shared AquaSynth.Dings contract owns catalog and MessagePack command/receipt types. Research provenance and timing policy live in docs/research/auditory-notifications/README.md.
- Song-snippet parity is scene-audio pressure, not redistributed fixture data. The worker may derive local tempo/register/scale hints from private clips, but agents must express live rhythm and pitch motion through owned AquaSynth controls until native sequencing sugar has an explicit DSL/lowering owner.
- Notes own frequency, gate duration, and host/MIDI source; envelopes own ADSR level shape. Do not put note duration or SFXR punch back into the general envelope model.
- Faust-managed MIDI polyphony is preferred:emit one voice graph with standard `freq`, `gain`, and `gate` controls plus `[midi:on][nvoices:n]` instead of inflating AquaSynth graphs for local allocation.
- AquaSynth Faust output may be rendered through Faust-generated C# for candidate audio comparison; reference parity still needs a real external render or captured licensed fixture.
- Reference-patch correctness and synth parity tests live in AquaSynth, not AquaSynth.
- Parameter binding uses field-site references like `lpf=@/macro/brightness`; do not add target-routing semantics unless a later map proves a new authority is needed.
- DX7 feedback lowers through `Dx7SysEx.OperatorFeedbackAmount`; feedback value 5 remains 0.19 for PRC, while feedback value 7 is much hotter at 2.2 to restore community-patch buzz. Do not restore PRC-local `feedback * constant` folklore.
- DX7 voice transpose lowers through `Dx7SysEx.NoteFrequencyHz`; transpose 24 is neutral, 12 is one octave down. Do not hardcode `freq=261.6256` for imported voices.
- DX7 fixed-frequency operators lower through `Dx7SysEx.FixedOperatorFrequencyHz` as absolute Hz divided by the graph base frequency; do not treat fixed mode as a note ratio.
- DX7 ROM `COM` output compensation is not a carrier-balance authority in AquaSynth lowering; `OperatorOutputCompensation` returns unity and graph gain owns loudness recovery.
- DX7 operator key-scaling and rate-scaling lower against the effective played MIDI note after voice transpose; do not hardcode MIDI 60 inside imported operator level or envelope timing.
- DX7 route lowering is topology-aware:isolated direct branches use `OperatorModulationRouteIndex` (`6.275`); summed edges and direct edges feeding summed modulation use `SummedOperatorModulationRouteIndex` (`6.0`).
- `SummedOperatorModulationRouteIndex` is backed by the hard PRC target, a project-authored algorithm-8 stack probe, and split cascade/summed-pair probes; do not restore the older under-bright `1.6` scale.
- DX7 calibration must proceed through isolated rungs in `docs/dx7-calibration-plan.md`; do not keep tuning hard PRC-specific constants as a substitute for a level/index model.
- Staged operator envelopes own multi-target operator level contours through `env=rl rates=... levels=...`; do not force DX7-style contours back into the general ADSR voice envelope model.
- Staged operator envelopes may use per-segment curves through `curves=lin,exp,...`; this belongs to the rate/level contour authority, not to the standard ADSR envelope.
- DX7 envelope trace evidence now lives in `Dx7SysEx.TraceEnvelope`, `Dx7SysEx.TraceInterpolatedEnvelope`, and `artifacts/parity/dx7-envelope-trace/egstep.csv`; Dexed-style audio uses block-interpolated applied gain, not the raw EG state alone.
- Do not add an `env=dx7` runtime surface until it matches Dexed plugin audio with the block-interpolated gain model, not only the raw trace helper.
- DX7 applied envelope lowering now lives in `Dx7SysEx.ApproximateAppliedRateLevelEnvelope`; it normalizes traced applied gain so operator output level remains a separate authority.
- Current DX7 analog1 algorithm-2 lowering caps non-carrier applied-envelope Level1 at 0.92 to avoid modulator peak blowout; treat this as pressure on exact DX7 EG/output modeling, not as permission to add broad peak clamps.
- `MELLOWSOLO` is a pressure artifact, not passing parity; its render window is 1.25s so the candidate release tail is not cut at the artifact boundary.
- Public-domain reference fixtures, GPL Zyn instrument-bank fixtures, dexed-py helpers, and pinned ZynAddSubFX source are test/development material only; published NuGet packages must ship only AquaSynth.Core/AquaSynth.Faust surfaces.
- ZynAddSubFX import starts as feature extraction and then explicit, scoped rebuilds. Upstream GPL `.xiz` instruments may be used as test-only pressure/parity feedstock with provenance reports, while project-authored `.xiz` fixtures remain the clean baseline. The Zyn engine source and instrument bank are pinned under `external/zynaddsubfx` for parity oracles only, not AquaSynth runtime code.
- Named `layer` declarations own Zyn-style kit/layer identity:engine tag, optional key range metadata, default gain, and effect-send labels. They lower to ordinary voices for now; do not smuggle PAD/free-envelope/effect behavior into them.
- Additive `harmonics` declarations own authored ratio/gain partial banks under named layers. They lower to ordinary voices for now; do not pretend this covers Zyn ADDsynth phase, bandwidth, oscillator shaping, or free-envelope semantics.
- Normal voices may use staged `env=rl` rate/level envelopes. Notes still own gate timing; envelopes own level shape. This is not arbitrary Zyn free-envelope point editing.
- PAD `spectrum` declarations own authored partial clouds under named layers and split table root from playback pitch:`root` is source/base frequency, while `freq` is the note played by the spectral voice. They generate PADsynth-style FFT wavetables with frequency-domain spreading and deterministic random phase, then emit Faust table readers before ordinary voice treatment. Richer PAD table shaping uses neutral `pad_mode`, `pad_bandwidth`, `pad_bwscale`, `pad_profile`, and `pad_position` syntax. The parser may accept older `zyn_*` parity-artifact aliases, but generated/authored `.aqua` should not expose reference-synth names for general DSP intent. Do not pretend this covers every Zyn OscilGen modulation/waveshaping/filter/adaptive-harmonic path yet.
- Zyn OscilGen oscillator filters use one-based harmonic indices:first harmonic is `1`, not `0`. Preserve that index when translating filter types or spectral peaks move to the wrong partial.
- Zyn PAD analog high-pass filters are filter authority too:HP1/HP2 type, stages, and FILTER_ENVELOPE motion lower through `hpf_order` and `hpf_env=rl`, not ad hoc layer gain or brightness compensation.
- Zyn PAD state-variable filters (`category=2`) are filter authority:LP/HP (`type=0/1`) lower through `lpf`/`hpf`, and band/notch (`type=2/3`) lower through `bpf`/`notch`.
- Band-pass and notch are native AquaSynth filter authorities (`bpf`, `bpf_q`, `bpf_order`, `notch`, `notch_q`, `notch_order`), not Zyn-only compatibility. Zyn PAD analog/SV band-pass and notch filters lower into those fields; formants remain vowel/body color and PAD spectrum remains source-table construction.
- Zyn PAD selected table roots are not a global default in AquaSynth lowering. Keep sample-0 roots unless a source rule earns the selected root; currently OscilGen shift-base sources (`base_function=7`) may use the selected note table root because Organ Choir proved sample-0 roots leave the octave voice misaligned.
- Zyn OscilGen base-function modulation is source-table authority:apply it as phase modulation before base-function sampling, not as later oscillator modulation or filter/EQ compensation.
- Zyn adaptive harmonics remap the prepared oscillator spectrum at the PAD table root frequency before PAD bandwidth/profile spreading. Do not skip it for sources whose wind/body character depends on sparse remapped harmonics.
- Zyn coverage is now tracked as handled/counted/unknown buckets by `ZynInstrumentSurvey.CoverageDirectory`; do not treat broad parse counts as expressivity. Unknown PAD filter/waveshaper/OscilGen buckets and counted ADD/SUB/LFO/free-envelope/formant/effect buckets need parity workouts before they graduate into AquaSynth model changes.
- Voice `vowels=` owns authored formant-frame motion:each frame is a static formant bank and `vowel_hz` scans/blends the frames in Faust. Zyn `FORMANT_FILTER` vowel sequences may seed this surface, but Zyn's exact cutoff-position driver, formant slowness, and vowel clearness are still pressure, not hidden compatibility logic.
- Filter damping is first-class:`lpf_q`/`lpq` owns explicit low-pass Q when a reference synth exposes analog filter damping. Low-pass order remains slope authority even when Q is present; Faust lowering cascades resonant low-pass stages instead of replacing order with one generic resonant filter. Do not fake Q by secretly moving cutoff, slope, or layer gain.
- Structured `.aqua` stock lives under `patches/` as development source material. Every library patch must parse/export, and the patch library must stay out of the NuGet package unless intentionally promoted.
- Project-authored Dexed patches may be generated inside tests for calibration; do not ship them in the NuGet package or treat them as runtime dependencies.
- Do not consume external synth targets that cannot be backed by parity tests. Parsing is inventory; reproducing behavior through terse, readable AquaSynth DSL is the proof.
- The speech lane is the active center:IPA parity, utterance embeddings, synth-driver automation, rendered loss, and compiled Faust control sweeps take priority over new DX7/Zyn digestion.
- Every learned speech arrow must stay inspectable:utterance metadata -> embedding, embedding plus phonetic features -> automation, automation -> rendered audio, rendered audio -> loss/gradient.
- Pink Trombone parity is structural primitive-vocal pressure, not permission to add more formant proxy knobs or revive the old waveguide lane. AquaSynth has a `tract` DSL command with PT-shaped runtime controls, but default vocal acoustics now lower through `VocalNetwork`; `Voice.Tract` is authoring sugar, not a second audio authority.
- Physical voice modeling should generalize around continuous acoustic morphologies, acoustic port networks, source ports, branch/radiation ports, and wave-clock policies. A larynx, syrinx, or alien voice should differ by topology and source-port configuration, not by special-case modules. Faust `physmodels.lib`/`delays.lib` provide useful primitives (`pm.chain`, `pm.l2s`, `pm.waveguideFd*`, `de.fdelay*`, Thiran/allpass and Lagrange delays), but AquaSynth should still generate tract/branch scattering where live area controls and inspectable training parameters matter.
- The physical voice research packet now includes a practical implementation/researcher map (`research/physical-voice-modeling/PEOPLE.md`) and talk/transcript ledger (`research/physical-voice-modeling/talks/TALKS.md`). Birkholz/VocalTractLab pressures multi-level articulatory control and gestures; Story/TubeTalker pressures continuous parametric area functions and length/coupling; Smith pressures bidirectional delay-line scattering and fractional-delay waveguides; Fels/ArtiSynth pressures separation between biomechanics, acoustics, timeline controls, and diagnostics. These are reusable DSL semantics, not modules to clone.
- Physical voice implementation has moved PT pressure into neutral acoustic owners:`AcousticPath`, `AcousticTerminal`, `AcousticConnection`, `AcousticSourcePort`, `AcousticBranch`, `AcousticRadiationPort`, `WaveClockPolicy`, and `AcousticPortNetwork`. Parser commands `path`, `terminal`, `connect`, `source_port`, `branch`, `radiation_port`, `wave_clock`, and `acoustic_network` exist. Existing `tract_shape`, `glottis`, `tract_injection`, `nasal_branch`, and `tract` commands populate acoustic records as authoring aliases; the legacy `Voice.Tract` resonator/waveguide renderers are removed.
- `acoustic` / `acoustic_voice` render an `AcousticPortNetwork` only through the compiled bidirectional Faust segment graph. Graph lowering splits paths at terminal positions, aggregates co-located terminals into graph nodes, orders connection ports deterministically, scatters same-path two-port area discontinuities, scatters arbitrary non-bypass connection groups through one passive N-port area law, injects source terminals even at connected nodes, and sums radiation terminals. Segment delay follows physical segment length and the network `wave_clock` strategy through unit-grid, half-sample, linear fractional, Lagrange, or Thiran/allpass Faust delay primitives. Source, branch, radiation, terminal, and connection controls can bind through ordinary Aqua parameters into emitted Faust. PT-shaped `tract` authoring routes through `propagation=graph`; legacy propagation values now fail fast instead of producing proxy audio.
- Pink Trombone log-mel parity now writes graph candidates only. Generated tract graphs mirror live `/pink/...` tract bindings into acoustic source, terminal, radiation, and path-area fields. Graph generation stopped baking first-frame tongue/constriction/lip morphology into the path; `AcousticPath.AreaControl` now turns tongue/constriction/lip controls into live area expressions at graph sample points. Velum owns nasal branch entrance diameter, which lowers through a squared parameter-binding transform into side-port admittance; generated nasal connections keep full connection coupling, and the nasal terminal carries the side-port admittance limit. The first nasal tube sample is no longer allowed to decide the velum opening. Lip opening owns path area and radiation aperture, not terminal area scaling. Graph radiation now uses aperture as a boundary impedance owner: lip/beak opening blends boundary reflection from near-closed termination to declared open-end reflection, names per-port boundary flow (`incoming - outgoing`), and weights output by bounded local terminal admittance rather than a fixed 4x output scalar. Latest static graph evidence after radiation admittance normalization is under `artifacts/parity/pink-trombone-logmel/20260527T170548495`: open/front/nasal/ma/sibilant/closure cosine 0.6675/0.7216/0.7251/0.7098/0.4862/0.6513; latest utterance evidence under `artifacts/parity/pink-trombone-utterance-logmel/20260527T170618466`:mama/papa/thrombosis cosine 0.7428/0.7830/0.5442 with rmsRatio 2.7955/2.4594/2.7397 and articulation 0.1288/0.1226/0.1151. The output is less absurdly loud, but articulation and speech-band motion are still the live failure.
- `docs/acoustic-graph-flow-audit.md` compares Aqua's graph to Birkholz/VTL, Story/TubeTalker, Smith waveguides, ArtiSynth, and PT/SndKit. The graph records can name the research-backed architecture, but current Faust lowering still lacks fully shaped graph-native closure pressure storage. The kept physical fixes move nasal coupling from global connection damping into transformed side-port admittance and use one Kelly-Lochbaum/PT-style traveling-wave convention for ordinary two-port area discontinuities and three-port branch junctions. The previous pressure-continuity rewrite was wrong for the graph's wave variables and produced the phone-vibration failure. Radiation now has named flow/admittance probes and local terminal-admittance weighting, reducing over-radiation without restoring the old damper. The graph clock for generated tract graphs is continuous fractional delay so adding terminals does not silently lengthen the morphology. Node source/incident-pressure debug probes expose source-layer failures; a raw release-reservoir boost was tested and cut after the KL fix made it overdrive papa. Latest utterance artifact `artifacts/parity/pink-trombone-utterance-logmel/20260527T223931573` has papa at cosine 0.8828/RMS 1.0186 while thrombosis still fails from silence/articulation. Next cut is source impedance/constriction filtering for release energy, not global gain or PT-shaped modules.
- PT utterance metrics now include cross-utterance separability because log-mel parity alone hid the user's actual listening failure. Artifact `artifacts/parity/pink-trombone-utterance-logmel/20260527T225833392` shows `mama/papa` candidate collapse:reference log-mel similarity 0.3791 versus candidate 0.6069, with candidate envelope similarity 0.8885. Interior source terminals no longer decide topology or node area/reflection; they attach to existing graph nodes and inject energy, while non-source terminals own geometry. Keep this source/topology cut even if other broken layers get louder.
- Generated lip radiation terminals no longer bake first-frame lip opening into terminal `area_scale`; lip opening owns live path area plus radiation aperture. This removes a stale initial-state authority where words starting with closed lips kept closed-lip admittance. Static PT graph artifact `artifacts/parity/pink-trombone-logmel/20260527T231611586` passes, while utterance artifact `artifacts/parity/pink-trombone-utterance-logmel/20260527T231056242` still fails `mama/papa` collapse and silent `thrombosis`. Next work remains source/reservoir/radiation coupling, not restoring terminal-scale double gates.
- Graph voice repair pass on 2026-05-28 fixed four model-level ownership leaks:glottal excitation is source-normalized with local `tanh`; segment propagation loss is area-dependent so closures absorb instead of ringing forever; generated nasal side-port area maps velum diameter squared directly instead of through the old hidden 0.15 damper; generated tract lips use live path-end geometry as the aperture owner and no longer double-gate radiation opening/boundary flow. Utterance artifact `artifacts/parity/pink-trombone-utterance-logmel/20260528T070759312` passes smoke and no longer collapses `mama/papa` (`referenceLogMel=0.3791`, `candidateLogMel=0.3651`); `thrombosis` improves to cosine 0.4366 but remains weak/quiet. Static artifact `artifacts/parity/pink-trombone-logmel/20260528T071240442` passes with `closure-release` demoted to sealed-static evidence (`peak=0.000005`). Next pressure is true graph-native closure reservoir/release through the scattering path.
- `docs/acoustic-graph-dsp-audit.md` is the current Jenga audit for graph voice DSP. It says the path graph, typed terminals, KL/PT two-port scatter, source/topology separation, velum side-port area, generated lip geometry ownership, area-dependent segment loss, and utterance separability should stay. The first audit cut replaced the generic connection fallback pressure/admittance blend and PT-shaped three-port branch recognizer with one passive N-port area-scattering law for all non-bypass connection groups; emitted Faust no longer contains `graph_connection_pressure_*` for graph connections. The second audit cut moved graph radiation color/admittance into one named local primitive with emitted `graph_radiation_reference_area_*`, `graph_radiation_differentiation_*`, `graph_radiation_highpass_*`, `graph_radiation_admittance_*`, and `graph_radiation_flow_*` signals instead of inline `30/70` and `area + 1.0` expressions. The third audit cut hoisted graph turbulence/release into named per-source `_closure`, `_release`, `_reservoir`, `_reservoir_drive`, `_pressure_drive`, `_release_pressure`, `_noise_gate`, `_release_gate`, `_sustained`, `_burst_noise`, and `_release_pulse` locals. Reservoir charging now uses directed upstream-minus-downstream pressure for ordinary two-port path nodes and positive local incident pressure as fallback; latest artifacts `artifacts/parity/pink-trombone-utterance-logmel/20260528T115452268` and `artifacts/parity/pink-trombone-logmel/20260528T115719882` still show no metric movement. This points away from source-local reservoir charging and toward missing contact-aware closure state that participates in scattering/flow. Remaining suspect owners are unowned segment loss constant, still-empirical radiation/source constants, nearest-node source attachment, and missing contact-aware closure state.
- Graph source lowering now names per-node incident pressure and node-local source sums inside the recursive graph loop. Generated tract injection source banks now use cell-centered positions and `1 / sections` index scaling, so turbulence/release energy is born inside tube cells rather than directly on graph/radiation nodes. Latest source-position artifact under `artifacts/parity/pink-trombone-utterance-logmel/20260527T172949293`:mama/papa/thrombosis cosine 0.6654/0.7746/0.5522, rmsRatio 3.0010/2.5712/2.4943, articulation 0.1444/0.1177/0.1193. Static closure-release RMS improved to 1.0078 under `artifacts/parity/pink-trombone-logmel/20260527T172843945`, but papa articulation worsened; keep the cell-centered source placement because it is the correct graph-level source geometry, not because metrics liked it.
- Source ports now have an explicit `impedance`/`load` control and graph lowering uses node incident pressure to make glottal, labial/syrinx, reed, click, synthetic, and turbulence sources load-aware. Generated tract area terminals are `Junction` geometry owners again; generated `voices_0_contact_*` terminals are a sparse contact overlay capped at four interior sections, with named graph-owned `graph_contact_*_closure`, `_reservoir`, and `_release` signals so closure pressure storage is no longer either every area sample or only a turbulence-source trick. Eight generated contacts plus inline occlusion overflowed the native debug Faust compile; four sparse contacts pass the opt-in thrombosis graph probe. `patches/advanced/bird-syrinx.aqua` is the first checked-in differentiable syrinx patch:paired labial source ports, bronchial paths, a tracheal path, an area-scattering merge, and beak radiation, all through `AcousticPortNetwork`.
- Historical graph compaction pressure:before the primitive cut, generated PT-style `tract` lowering capped oral area terminals at 10, moving injection source ports at 8, and contacts at 4 to keep Faust alive. That path is no longer the default vocal owner; keep the lesson, not the generator.
- Moving tract injection placement now uses a raised-cosine kernel over the compact source bank instead of a hard triangular source-position weight. Keep this direction:smoother continuous heuristics over fewer generated ports, not more discrete source terminals. Artifact `artifacts/parity/pink-trombone-utterance-logmel/20260529T140038717` still fails articulation but improves thrombosis active ratio/silence/speech-band slightly without increasing graph size.
- Generated lip radiation opening must mirror live tract lip opening. The path-end geometry and radiation aperture are one mouth owner, not separate truths. Artifact `artifacts/parity/pink-trombone-utterance-logmel/20260529T141822464` improves papa's plosive failure sharply (RMS 0.8369, motor-band ratio 0.0555), but mama still sounds vowel-poor and thrombosis is still mostly absent; next pressure is vowel body/radiation color and fricative continuity.
- Proprioception correction says baby words are downstream witnesses, not the rebuild owner. The live cut is to collapse generated graph-vocal lowering into a small set of instrumented physical primitives, compare primitive/sample-flow behavior against VTL, ArtiSynth, Story/TubeTalker, Smith/Faust, and PT/SndKit, then return to utterances once the tract body is observable. `docs/acoustic-graph-proprioception.md` owns this cut line.
- First primitive vocal teardown cut is live. `tract`, `glottis`, `tract_injection`, and `nasal_branch` now adapt PT fixtures into `AreaFunction`, `WaveguidePath`, `SourcePort`, `ConstrictionContact`, `BranchPort`, `RadiationLoad`, `ProbeTimeline`, and `VocalNetwork` records instead of emitting generated `voices_0_area_*`, `voices_0_contact_*`, or `voices_0_inj_*` acoustic banks. The obsolete `EnsureTractAcousticNetwork` generator is deleted. Faust lowering for `Voice.VocalNetwork` emits compact primitive path/source/contact/branch/radiation locals using `de.fdelay*` delay families and debug probes under `/debug/...`; `ProbeTimelineReport` writes deterministic primitive CSV samples. Next pressure is passivity/energy timeline samples plus PT/SndKit timeline comparison, not intelligibility tuning.
- Primitive path telemetry now reports `incoming_wave`, `outgoing_wave`, `energy_in`, `energy_out`, and `passivity_ratio` in both Faust debug probes and `ProbeTimelineReport`; contact telemetry includes resistance. `PrimitiveReferenceReport.ComparePinkTrombone` compares static primitive area sections, length, reflection count, delay estimate, branch admittance, contact state, and radiation fields against local PT/SndKit fixture controls without adding a native reference dependency. Next pressure is per-block comparison against PT/SndKit internals, not static control matching alone.
- `PinkTromboneReferenceRenderer.RenderTimeline` now exposes PT internal oral/nasal path, branch, contact, and radiation samples from the renderer's own diameter/wave/reflection/transient arrays. `PrimitiveReferenceReport.ComparePinkTromboneTimeline` role-normalizes Aqua primitive samples against PT timeline samples for path/branch/contact/radiation signals. This is the first local reference-internal comparison surface for the primitive rebuild; next pressure is fixture-suite artifact reporting and then tightening laws where the timeline drift points.
- `PinkTrombonePrimitiveTimelineComparisonWritesArtifacts` writes per-fixture Aqua primitive timelines, PT internal timelines, comparison CSVs, and a summary under `artifacts/parity/primitive-vocal-timeline/<timestamp>`. This makes primitive drift inspectable across the PT fixture suite without native Faust or external VTL/ArtiSynth dependencies. Use these artifacts to choose the next physical-law correction.
- First comparison-driven primitive law correction:`SourcePort` owns `FlowScale` (default `.02`) so pressure/tension/opening no longer emit hidden unit-scale waves; `RadiationLoad.Reflection` owns reflection directly and aperture no longer rewrites that coefficient; `ProbeTimelineReport` radiation flow now uses path outgoing wave like Faust instead of reporting `aperture / impedance`. Artifact `artifacts/parity/primitive-vocal-timeline/20260529T155613895` drops fixture-suite mean absolute timeline error from roughly `1.2-1.3` to `0.15-0.27`. Remaining large drift is branch/contact dynamics, which needs an owner decision about gesture slew versus primitive-local smoothing.
- Gesture motion ownership is now separate from acoustic primitives. `ControlSurface` exposes normalized runtime controls for primitive fields, and `ControlSpline` owns future-time linear/hold/Bezier gesture motion over those surfaces. Faust lowering emits surface and spline point controls, maps effective normalized values back to physical primitive fields, and keeps old `@/param` bindings as compatibility base controls. `tract_motion` adapts velopharynx opening into a branch spline. A generated contact-opening spline was tested and cut because PT timeline evidence showed it invented an open-then-close transient; contact closure must come from articulator/area ownership. Focused verification:`PatchScriptTests` pass 90/90 and primitive timeline artifact `artifacts/parity/primitive-vocal-timeline/20260529T185955965` keeps `closure-release` at mean absolute error `0.155925`.
- Public gesture authoring API is now explicit. `ControlSurfaceCatalog.FromPatch(patch)` exposes normalized surface metadata for hosts; `ControlSplineTimeline` lets clients edit future points and emit `ControlValuesAt(t)` as ordinary surface-control dictionaries. Native Faust compiled patches carry `ControlSurfaces` and expose `ControlValuesAt(timeline,t)` to filter semantic timeline values to controls exported by the optimized DSP. This is the runtime path for user-authored gestures without recompilation. Verification:`PatchScriptTests` pass 91/91 and `NativeFaustRuntimeTests` pass 5/5.
- First IPA/anatomy sugar is live. `phoneme_gesture` accepts known IPA symbols or explicit descriptors such as `descriptor=voiceless_labial-velar_fricative`, maps place/manner/voicing onto primary anatomical surfaces, and emits ordinary `ControlSpline`s. It is deterministic seed material for local reference-data golf, not exact articulatory truth. Patches may still use normal FM/AM/modulators/envelopes/filters around the anatomical control stream. Verification:`PatchScriptTests` pass 92/92.
- IPA training scoring contract is documented. Reports must keep `gesture_score`, `clean_vocal_score`, and `full_parity_score` separate. Gesture score measures descriptor-to-surface coverage, motion direction, spline contour/timing, primitive timeline consequences, and optional external articulation landmarks. Full parity patches may use FM/AM/modulators/envelopes/filters, auxiliary animated voices, and recording-condition helpers, but those helpers cannot count as articulation evidence unless gesture and clean-vocal scores move coherently.
- `IpaGestureExperiment.WriteRound` is the first frozen batch harness for IPA hypothesis work. It writes `manifest.yaml`, candidate `.aqua` scripts, primitive `ProbeTimelineReport` CSVs, gesture-layer `metrics.csv`, and per-candidate `evidence.jsonl` under a caller-provided artifact root. `IpaGestureExperiment.AnalyzeRound` writes `analysis/metric-summary.csv`, `analysis/candidate-clusters.csv`, and `analysis/science-brief.md` for the science-worker handoff. This harness owns frozen descriptor/spline/primitive evidence only; clean vocal identity, full spectrogram parity, optimizer checkpoints, and worker orchestration stay downstream. Focused verification:`PatchScriptTests` pass 93/93.
- `IpaTrialOrchestrator.RunAsync` is the first local render/scoring worker for IPA trials. It generates candidate patches through `IpaGestureExperiment`, renders them through Faust, compares against local PT fixture references with `AudioAnalyzer`, writes WAV/report artifacts, and stores typed `IpaTrialResult` records in `ipa-trial-results.cc`. The five-seed trial run at `tests/AquaSynth.Dsl.Tests/bin/Debug/net10.0/artifacts/parity/ipa-trials/20260529T214127955/five-seed-trials` showed voiced vowels/nasals became audible only after primitive `SourcePort` lowering gained a true oscillatory voiced carrier; sibilants are pressure, plosives remain weak. Verification:opt-in IPA trial test passed, and focused `PatchScriptTests|IpaTrialResultsRoundTripThroughCultCache` pass 94/94.
- External IPA exploration now has a worker/tool boundary. `tools/IpaTrialWorker` owns `seed`, `score`, `search`, `show`, and audit-only `dump` commands over the typed `.cc` trial-result store; `score` renders external Codex-authored `.aqua` candidates through `IpaTrialOrchestrator.RunCandidateScriptsAsync`, while `search` is the semantic retrieval organ for hypothesis agents. `IpaTrialResult` records now embed extracted primitive timeline facts from `ProbeTimelineReport` for contact closure/reservoir/release, source load/flow, branch admittance/exchanged flow, radiation boundary/output, path area/delay/energy, and passivity; `search`, `show`, `dump`, and summary CSV output surface release/radiation/passivity witnesses directly from the `.cc` record. `search` follows VoidBot's Qdrant shape: `.cc` records plus timeline facts and referenced local artifacts become evidence chunks, local Ollama `qwen3-embedding:0.6b` produces 1024-dim vectors, Qdrant collection `aquasynth_ipa_trial_results` stores stable SHA-256 UUID points with trial/candidate payload metadata, and the report surfaces vector scores, excerpts, tags, metric facts, timeline facts, contrast candidates, `class_focus`, per-hit `class_match`, and `show` commands. Qdrant is a derived index; CultCache remains truth, and lexical/metric search is fallback when the vector stack is unavailable. Retrieval tuning added phonetic class-affinity scoring and candidate-specific mixed-lane ownership so `mix-p` can support stop seams without every mixed candidate matching stop evidence. `tools/run-ipa-trial-loop.ps1` orchestrates external `codex exec` hypothesis/evaluator workers around that store. Source-edit-enabled rounds now allow agents to golf DSL syntax, parser/model records, Faust lowering, worker evidence surfaces, and tests when evidence points at a repeated owner failure; agents must write `source-change-plan.md`, and the loop runs a focused build before scoring. `hypotheses.md` is the hypothesis worker report of record; `hypothesis-agent.md` is only the external agent's final chat summary. The first loose source-edit round produced no source edits and no aggregate metric win, then exposed a deeper learning-loop fault: family ideas were rotated across unrelated targets instead of aligned to phonetic owners. Production prompts now pass the previous evaluation path into the hypothesizer, require `Previous Round Learning`, require `Target-Class Lane Plan`, forbid round-robin family assignment, and let the evaluator fail class alignment when family/target assignments do not match phonetic owner classes or explicit transfer-test mutations. Candidate reports must include a `Scene Voice Model` naming primary-articulatory, tract-noise, release/transient, background-room, and recording-condition voices, with general AquaSynth synth primitives allowed for helper scene voices but not accepted as articulation evidence unless lower-layer witnesses move. At least one scene/noise or recording-condition family in a source-golf round must implement real auxiliary `.aqua` `voice` declarations instead of satisfying scene modeling with prose. The active trial shape is five target sets of five phoneme lanes each:vowels, nasals/approximants, fricatives, stops, and mixed generalization. Candidate names are exactly `<targetId>__<family>__<hypothesis-name>.aqua`, with lowercase kebab-case family ids. Three dry external Codex smoke passes refined hypothesizer/evaluator prompts into exact 25-lane output, retrieval receipts, claim audits, evidence-quality ledgers/verdicts, and fail-closed evaluator compliance. A later five-pass prompt optimization showed the best science comes from fixed receipt packets plus constrained reasoning; production prompts now require contrast pairs, `show` metric values, primitive timeline facts/excerpts or explicit `missing`, class-consistent reference matrices, scene voice models, `X owns Y so Z remains true` owner sentences, at least five micro-sweep perturbations before a seam is actionable, and a novelty gate rejecting loudness/dressing-only families as loss-
- Song-snippet parity now treats target analysis as CultMesh database evidence. `song-prepare` writes each 30-second clip's log-mel spectrogram CSV, per-band first/second derivative stats, RMS envelope, envelope autocorrelation, and analysis report. `song-score` upserts those as typed `SongChallengeEvidenceDocument` records in the same `.cc` store as `IpaTrialResult`, adds target metrics and `cultmesh://aquasynth/song-challenge-evidence/...` artifacts to every result, and `search/index` derive Qdrant chunks from the database records. `pattern` and `scale` DSL sugar are live authoring conveniences that lower into normal parameters and hold-loop curves; prompt agents may propose richer Strudel-like syntax, but candidate patches must use implemented syntax.
- Song-scene scoring now records instrument-role ownership so the music-generator curriculum does not learn that simple oscillator blips are good instruments by default. `song-score` emits `instrument_voice_syrinx`, `instrument_drum_subtractive`, `instrument_pad_additive`, `musical_instrument_score`, and `chip_distress_risk`; verdict/distillation ranking rewards owned roles and demotes chip-distress-risk candidates unless preserved as contrast. Swarm prompts now require voice-like leads to prefer syrinx/acoustic sources with pressure/opening/radiation motion, percussion to prefer subtractive pitched bodies plus filtered noise skins, and pads/beds to prefer additive/PAD `layer`/`harmonics`/`spectrum` banks.
- Song-scene harness supports full-song curriculum and in-agent revision. `song-prepare --duration-seconds 0` freezes the whole decoded song from sample zero, and `run-song-snippet-trial-swarm.ps1 -FullSongs` samples recursively from the Reinier parent folder including `.m4a/.mp4`. Target and candidate analysis now include whitened spectral autocorrelation over band-whitened log-mel flux. Candidate-side Faust-render artifacts include log-mel spectrogram/stats, RMS envelope/autocorr, whitened spectral autocorr, and an analysis markdown report. Swarm prompts require five local score/revise attempts per target before publishing the final `.aqua`.
- Patterned AquaSynth percussion now needs explicit voice-local retrigger ownership. Sequenced gain (`gain=@/seq/kick`) only changes level; without `trigger=<step_seconds>`, a one-shot drum envelope, pitch fall, FM decay, filter ramp, or phaser motion keeps aging from song start and later hits become dead-gate mush. Music agents should start 808 percussion from `patches/808/kick.aqua`, `snare.aqua`, `hat.aqua`, and `clap.aqua`, split body/click/skin/clap/hat layers explicitly, add `trigger`, and use drive/fold/sample-hold texture instead of broadband noise pretending to be a kit.
- TurbulenceJet now includes a release-timed burst-noise component in addition to the deterministic pressure pulse, but a stronger trial barely moved papa; the source is not the main remaining articulation bottleneck. Graph radiation now mixes 30% raw boundary flow with 70% high-passed flow after local admittance weighting, down from the over-bright 15/85 blend. Latest radiation/source artifact under `artifacts/parity/pink-trombone-utterance-logmel/20260527T174236049`:mama/papa/thrombosis cosine 0.6578/0.7692/0.5585, rmsRatio 3.0006/2.6153/2.3488, articulation 0.1437/0.1173/0.1191. Static `closure-release` under `artifacts/parity/pink-trombone-logmel/20260527T174139661` has rmsRatio 0.9900 and speechBandRatio 0.0842. This is still not speaking; next pressure is true tract resonance/passive scattering and pressure storage, not more burst gain.
- Same-path two-port area discontinuities now emit `graph_area_energy_in_*` and `graph_area_energy_out_*` probes alongside the three-port branch `graph_connection_energy_*` probes. Use these to inspect passivity and formant-body failures at the scatter sites rather than judging only final output loudness.
- `FaustExportOptions.DebugProbeUi` wraps graph probe signals in Faust `vbargraph` UI controls under `/debug/...` without changing audio output arity. Native streaming Faust patches now separate writable controls from read-only probe zones and can read/snapshot probe values after `ProcessBlock`; use this for acoustic graph passivity/radiation inspection, leave it off for normal parity renders.
- `PinkTromboneGraphDebugProbesWritePassivityReportWhenNativeFaustIsInstalled` streams the `nasal-vowel` graph with debug probes and writes `artifacts/parity/pink-trombone-graph-probes/<timestamp>/passivity-report.txt`. Latest run `20260527T200828013` shows same-path area and nasal-branch energy out/in ratios at 1.0, so current gross failure is not basic scatter energy explosion; next cuts should inspect frequency-dependent loss/radiation color and closure pressure storage.
- PT expressive parity should decompose into reusable Aqua primitives before audio golf: `tract_shape` owns continuous diameter/area geometry over normalized position, physical length (`length_cm`), interpolation/resampling, derived reflection coefficients, and per-cell acoustic delay; `AcousticAreaControl` owns live sampled morphology controls on a path; `WaveClockPolicy` owns the legal delay floor instead of forcing every segment to one sample; generated graph tracts now use `HalfSampleGrid` to match the PT/Kelly-Lochbaum half-sample regime; ordinary same-path graph interiors now own signed two-port area-discontinuity scattering; `glottis` owns excitation controls with mild harmonic waveshaping and a small higher-harmonic brightness term for graph sources; `tract_injection` owns positioned frication/burst controls; `AcousticSourcePort.Transient` owns graph-native release/click energy as a pressure pulse rather than filtered noise; sustained `TurbulenceJet` lowering owns a narrow-open aperture gate so closed constrictions no longer hiss; `AcousticSourcePositionControl` owns differentiable runtime weights over fixed source terminals so moving articulators can move source energy without recompiling the graph; `AcousticRadiationPort` owns aperture-dependent boundary reflection, flow radiation, and current post-flow normalization; `nasal_branch` owns a second diameter/area tube and branch aperture while `AcousticConnection.Coupling` owns structural max admittance; `tract_motion` owns control slew and obstruction threshold; `AcousticPortNetwork` owns graph lowering. `PinkTromboneReferenceRenderer` ports the pinned MIT `chdh/pink-trombone-mod` DSP source (`external/pink-trombone-mod` at 359c2d3b42b10280404c1650dc601902112b4c90), and reference WAVs before `artifacts/parity/pink-trombone-utterances/20260526T211902375` are invalid. The accepted PT utterance curriculum is mama, papa, and thrombosis; lulek was cut because its control sketch taught a bad artifact. Latest graph-only utterance smoke under `artifacts/parity/pink-trombone-utterance-logmel/20260527T150836912`:mama/papa/thrombosis remain `not-accepted-articulation` with cosine 0.8337/0.8893/0.5317 and articulation scores 0.2118/0.1890/0.2594. `AudioComparison.Articulation` reports envelope cosine, active-frame ratio, silence mismatch, envelope/spectral flux ratios, motor-band ratio, speech-band ratio, and an articulation score so log-mel cannot forge a listening receipt. A direct LF coefficient source was tested and cut because it overdrove the graph and broke static open vowel; a broader harmonic-tail source was tested and cut because it worsened voiced-band behavior. A naive midpoint admittance scatter was tested and cut because it collapsed `mama`; the kept area-discontinuity scatter needed the opposite reflection sign for this graph state convention, and re-testing the literal SndKit sign after the clock fix made `thrombosis` loud/unstable. A long noisy release transient was tested and cut because it turned papa into motor-band energy; deterministic pressure release with PT-scale decay is structurally kept but still not sufficient. Next pressure is true graph pressure storage/release at severe closures; output coloring alone is not allowed to impersonate articulation.
- If dotnet builds hang in this workspace, stop orphan dotnet workers and retry with `--disable-build-servers -p:UseSharedCompilation=false -p:BuildInParallel=false`; plain parallel/shared builds can stall without useful diagnostics.
- eSpeak NG is an early parity reference for timing, IPA coverage, and log-mel pressure, not an anatomy oracle or final target.
- Do not expand into prosody, emotional context, personality, or broad Weksa expressivity until the tiny IPA utterance harness can show useful loss movement.
- Speech training receipts must keep time as a witness:every request, render result, and checkpoint should record stage decision time, latency spent, latency budget, confidence, and artifact/model identifiers.
- CultMesh is the speech render/scoring transport and worker-orchestration layer, not distributed checkpoint authority. AquaSynth now has payload manifests, peer/lease admission receipts, worker assignment, result collection, and central checkpoint application. Workers may render, score, return gradients, artifacts, and timing receipts under a lease; the authority trainer or shard-primary applies gradients through `SpeechBackpropagationPipeline` and commits checkpoints through normal CultCache/CultMesh shard authority.
- Do not use semantic text embeddings as a proxy for IPA or alien phonetic strings. Phonetic/IPA realization starts as PanPhon sequence evidence and is compressed by AquaSynth's phonetic sequence encoder.
status:"Boundary contracts, packed neural networks, utterance encoder, phonetic sequence encoder, synth-driver mapper, chained backprop, optional eSpeak rendered fixtures, Faust finite-difference loss surface, compiled Faust speech-loss controls, distributed render request/result/checkpoint contracts, timing/confidence receipts, sound-claim cards, and a Weksa-emitted v0.1 tiny PanPhon batch exist. CultMesh speech training now has payload manifests, peer/lease admission receipts, worker assignment, result collection, and central checkpoint application; real network artifact transfer/process supervision remains CultMesh infrastructure. The active need is a loader/training pass that consumes Weksa's batch artifact and tests many controller weights as rendered speech output against simple ground truth."
- "Keep the utterance schema small and testable: segments, spans, duration/stress/boundary hints, speaker/morphology identity, pitch/loudness/rate/breath controls, and named context vectors."
- "Use the CultMesh speech training step for compiled Faust render/scoring jobs: admit compatible workers under leases, collect `SpeechRenderResult` gradients/receipts, and apply centrally or by one shard-primary trainer. Do not let worker payload transport become optimizer authority."
status:"Initial ZynInstrumentReader reads plain or gzip-compressed .xiz XML from project-authored fixtures and classifies active ADD/SUB/PAD engines, enabled kit items, layering, envelopes, free envelopes, LFOs, filters, formants, and effects as ReferencePatch features. `ZynInstrumentSurvey.CoverageDirectory` now classifies observed Zyn features into handled/counted/unknown buckets and writes `artifacts/parity/zyn-upstream-bank-survey/coverage.md` from the upstream GPL bank. First AquaSynth rebuilds exist for additive lead, PAD texture, and vocal/formant layer and now use named `layer` declarations. Additive rebuilds can use `harmonics` banks attached to layers, PAD rebuilds can use `spectrum` PADsynth-style FFT wavetable banks attached to layers, and PAD/vocal rebuilds use staged voice `env=rl` envelopes through layer defaults. Voices can now use `vowels=` formant frames plus `vowel_hz`; `ZynInstrumentReader.RebuildFirstFormantMotionAsAquaSynthScript` seeds that surface from Zyn `FORMANT_FILTER` vowel sequences for Wide Bass-style workouts while keeping exact Zyn position-driver/smoothing as named missing features. `ZynInstrumentReader.RebuildFirstPadAsAquaSynthScript` now emits per-preset PAD `.aqua` rebuilds for enabled PAD kit items with OSCIL harmonic expansion, oracle-reported table base frequency, PAD detune, Zyn PAD profile/bandwidth/position semantics, static/filter-envelope low-pass mapping, low-pass note tracking, explicit low-pass Q, logarithmic Zyn envelope times, and calibrated PAD output gain. ZynAddSubFX source is pinned under `external/zynaddsubfx` at 3ab608c432996ba4d582176572c0b0f82328c825 for test-only parity work, and `tools/zyn-reference/build-zyn-reference.ps1` now builds an ignored patched source copy with MSYS2/MinGW, proves `PadNoteTest`, and builds `ZynPadReference`, including a test-only `harmonics` mode for dumping normalized OscilGen amplitudes before PAD generation, plus `ZynVoiceReference` for ADD note reference renders. The optional PAD reports emit note and source-table WAV/listening artifacts. The optional formant batch renders Wide Bass, Ghost Ensemble, and Ghost Choir under `artifacts/parity/zyn-upstream-formant-fixtures`; initial scores are 0.280990, 0.312813, and 0.363644, which makes the next workout real rather than flattering. Latest accepted upstream PAD pressure has DoublePadBass close enough by ear with log-mel 0.247748 / score 0.650029 after Q/order cascade, while Ghost Ensemble improves to log-mel 0.369402 / score 0.57229. The remaining Ghost gap is no longer first-32 OscilGen harmonic extraction; oracle harmonic dumps match those ratios. The upstream GPL instrument bank is initialized at `external/zynaddsubfx/instruments` commit c5c912131b31df5fdf372d2f06a25aaf2375837f; survey tests parse 1318/1318 .xiz files, write a pressure report and coverage map, and render a five-patch upstream PAD listening batch under ignored artifacts."
- "Use Wide Bass formant-frame output as the foothold before attempting exact Zyn formant cutoff-position driving."
- "Additive harmonic banks, PAD spectral clouds, staged voice envelopes, and vowel-frame motion now exist; prove any next syntax across more than one target with parity, not parser-only fixtures."
- "Cut any new graph-vocal feature that lands before primitive-collapse and flow-probe instrumentation make the tract body's sample/flow behavior observable."
