diff --git a/docs/pink-trombone-dsl-parity-plan.md b/docs/pink-trombone-dsl-parity-plan.md index 5e7ccb6..d341fa8 100644 --- a/docs/pink-trombone-dsl-parity-plan.md +++ b/docs/pink-trombone-dsl-parity-plan.md @@ -238,6 +238,13 @@ gain did not restore oral vowel RMS. The remaining issue is oral energy: the side-branch law is less wrong, but the graph still needs better power/flow normalization before RMS parity becomes honest. +Frication also had a low-level ownership bug: generated graph `TurbulenceJet` +ports were using burst as source pressure, so sustained sibilants were choked +by a transient control. They now use `max(turbulence, burst)` as pressure while +keeping turbulence as the noise amount. This only nudges the current sibilant +fixture (0.2302 cosine), but the primitive semantics are cleaner: turbulence +owns continuous noisy pressure, burst owns transient pressure. + ## Cut Line The current `Voice.Tract` scalar proxy may survive only as an adapter over real diff --git a/src/AquaSynth.Core/PatchScript.cs b/src/AquaSynth.Core/PatchScript.cs index d5d2e79..ef86d68 100644 --- a/src/AquaSynth.Core/PatchScript.cs +++ b/src/AquaSynth.Core/PatchScript.cs @@ -1254,10 +1254,10 @@ public static class PatchScript primaryPathName, NormalizeTractIndex(injection.Position, tract.Sections), AcousticSourceKind.TurbulenceJet, - injection.Burst, + Math.Max(injection.Turbulence, injection.Burst), 0, injection.Diameter, - injection.Turbulence); + Math.Max(0.001f, injection.Turbulence)); AddAcousticSourcePortRecord(injectionSource); AddAcousticTerminalRecord(new AcousticTerminal(injectionSource.Name, injectionSource.Path, injectionSource.Position, AcousticTerminalKind.Source, injectionSource.Name)); sourceNames.Add(injectionName); diff --git a/state/evidence.jsonl b/state/evidence.jsonl index 7705aa1..f51b732 100644 --- a/state/evidence.jsonl +++ b/state/evidence.jsonl @@ -155,3 +155,4 @@ {"timestamp":"2026-05-26T20:25:00+01:00","topic":"pink-trombone-graph-source-radiation-evidence","claim":"Graph source and radiation semantics are now the live parity levers. LF-style glottal flow with a small harmonic pressure blend, graph radiation high-pass leakage, and aperture-weighted high-passed `TurbulenceJet` sources raise graph open/front vowel cosine above the legacy waveguide lane and make closure-release positive. Nasal remains over-loud and sibilant cosine remains weak, so the next pressure is nasal branch/radiation gain or junction coupling plus frication/transient detail.","evidence":"Latest optional PT log-mel run under `artifacts/parity/pink-trombone-logmel`: graph open 0.6393 vs legacy 0.6143, graph front 0.6596 vs legacy 0.6349, graph nasal 0.4607 with RMS ratio 1.6914 vs legacy nasal 0.5425, graph sibilant 0.2320 vs legacy 0.3744, graph closure 0.1170 vs legacy -0.0820."} {"timestamp":"2026-05-26T20:55:00+01:00","topic":"physical-voice-side-branch-scattering-blocker","claim":"Nasal graph parity is blocked by the side-branch port law, not by a missing scalar gain. Terminal-specific admittance in connection pressure is structurally cleaner but does not move PT log-mel metrics. Per-port outgoing coupling and squared generated velum opening both suppress nasal leakage only by making oral fixtures nearly silent, showing that the current pressure-wave normalization lets a weak side branch become an accidental loudness path. The next architecture needs power/flow-consistent branch-port scattering that separates junction admittance from outgoing wave normalization.","evidence":"Focused optional PT log-mel runs after per-port outgoing coupling produced open/front graph RMS ratios near 0.0017 with nasal cosine 0.5338; squared generated velum opening produced open/front RMS ratios near 0.0038/0.0039 and nasal RMS 1.5729. Both were cut. The retained terminal-admittance pressure solve preserves the previous source/radiation metrics."} {"timestamp":"2026-05-26T21:20:00+01:00","topic":"physical-voice-side-branch-admittance-progress","claim":"Side-branch graph scattering now uses terminal-specific admittance plus a floored area-ratio pressure normalization for outgoing branch waves. This is not final power parity, but it improves nasal cosine/RMS without erasing the oral vowel cosine win. Raw per-port coupling, squared velum mapping, and lip/beak radiation gain were tested and cut.","evidence":"Latest optional PT log-mel run: graph open 0.6202 RMS 0.1703, graph front 0.6417 RMS 0.1752, graph nasal 0.5159 RMS 1.4815, graph sibilant 0.2268 RMS 1.0643, graph closure 0.1117 RMS 1.8085. Previous source/radiation baseline was open 0.6393 RMS 0.3559, front 0.6596 RMS 0.3661, nasal 0.4607 RMS 1.6914, sibilant 0.2320 RMS 0.9158, closure 0.1170 RMS 1.6650."} +{"timestamp":"2026-05-26T21:35:00+01:00","topic":"physical-voice-frication-source-pressure","claim":"Generated graph `TurbulenceJet` ports now treat sustained turbulence and burst as separate pressure authorities: continuous frication uses max(turbulence, burst) for source pressure instead of being throttled by burst alone, while turbulence still gates noise amount. Fully ungated noise was tested and cut because it hurt closure-release.","evidence":"Optional PT log-mel after pressure fix: sibilant graph cosine 0.2302 / RMS 1.1387 and closure graph cosine 0.1145 / RMS 1.6627. Setting noise to 1 barely changed sibilant (0.2303) and worsened closure (0.1113 / RMS 1.8777), so that part was reverted."} diff --git a/state/spine.yaml b/state/spine.yaml index 211db3e..f0330a1 100644 --- a/state/spine.yaml +++ b/state/spine.yaml @@ -118,7 +118,7 @@ live_invariants: - 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 started moving 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` now exist. Existing `tract_shape`, `glottis`, `tract_injection`, `nasal_branch`, and `tract` commands populate acoustic records as compatibility aliases while the current `Voice.Tract` Faust renderer still owns audio. - `acoustic` / `acoustic_voice` can now render an `AcousticPortNetwork` through a compact Faust response proxy or, when typed terminals are present, a compiled bidirectional Faust segment graph. Graph lowering splits paths at terminal positions, aggregates co-located terminals into graph nodes, orders connection ports deterministically, scatters area-weighted/pressure/bypass connection groups, 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 can opt into graph audio with `propagation=graph`; legacy `propagation=waveguide` remains the comparison baseline. Remaining pressure: deepen connection-law semantics and run log-mel parity against the graph renderer. Do not expand the proxy into a second fake synthesizer. - - Pink Trombone log-mel parity now writes both legacy waveguide and graph candidates. Latest graph evidence after source/radiation/side-branch cuts: graph open/front cosine is 0.6202/0.6417, still slightly above the legacy waveguide lane at 0.6143/0.6349; nasal graph cosine improved to 0.5159 and RMS ratio dropped from 1.6914 to 1.4815; sibilant graph cosine remains weak at 0.2268; closure-release graph cosine stays positive at 0.1117 and beats the old negative waveguide baseline. The useful changes were LF-style glottal shaping with a modest harmonic pressure blend, graph radiation high-pass leakage by radiation kind, aperture-weighted high-passed `TurbulenceJet` source semantics, terminal-specific admittance in connection pressure, and floored area-ratio pressure normalization for branch outputs. Raw per-port outgoing coupling, squared generated velum opening, and lip/beak radiation gain were cut as compensators. Remaining pressure is oral RMS/energy normalization and better frication/transient behavior, not more static shape paperwork. + - Pink Trombone log-mel parity now writes both legacy waveguide and graph candidates. Latest graph evidence after source/radiation/side-branch/frication cuts: graph open/front cosine is 0.6202/0.6417, still slightly above the legacy waveguide lane at 0.6143/0.6349; nasal graph cosine improved to 0.5159 and RMS ratio dropped from 1.6914 to 1.4815; sibilant graph cosine remains weak at 0.2302; closure-release graph cosine stays positive at 0.1145 and beats the old negative waveguide baseline. The useful changes were LF-style glottal shaping with a modest harmonic pressure blend, graph radiation high-pass leakage by radiation kind, aperture-weighted high-passed `TurbulenceJet` source semantics, terminal-specific admittance in connection pressure, floored area-ratio pressure normalization for branch outputs, and generated frication ports using max(turbulence, burst) as pressure instead of burst alone. Raw per-port outgoing coupling, squared generated velum opening, lip/beak radiation gain, and un-gated turbulence noise were cut as compensators. Remaining pressure is oral RMS/energy normalization and better frication/transient behavior, not more static shape paperwork. - 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; `tract` may resample that geometry to a chosen compiled `sections` grid without making authoring sample count anatomy; `glottis` owns excitation controls; `tract_injection` owns positioned frication/burst controls and now distributes frication pressure into waveguide section updates; `nasal_branch` owns a second diameter/area tube; `tract_motion` owns control slew and obstruction threshold; `propagation=waveguide` lowers tract voices into a Faust-friendly `wg_loop ~ si.bus(...)` feedback component with compiled right/left traveling-wave state, live per-section diameter/area/reflection fields, obstruction-history burst pressure, and legacy substep clock intent. Waveguide tracts now default to an acoustic unit-delay grid from physical length at the current 44.1 kHz target, so a 17 cm oral tract emits about 22 sections instead of treating PT's 44 half-sample cells as anatomy; `VocalTract.IndexScale` maps runtime PT-index controls onto that emitted grid. PT's 44 oral and 28 nasal cells are reference discretization, not Aqua morphology authority. `PinkTromboneParityFixtures` defines fixed Aqua DSL workouts, `PinkTromboneReferenceRenderer` renders deterministic PT-style fixture references, and `AudioComparison` includes log-mel cosine similarity. Current waveguide baseline is weak but measured: cosine about 0.61 open vowel, 0.63 front vowel, 0.54 nasal, 0.37 sibilant, and -0.08 closure-release. Remaining low-level gap is fractional-delay waveguide lowering whose clock derives from physical length, propagation speed, and sample rate so runtime morphology motion does not require recompilation; closure/release timing is still badly wrong. - 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.