diff --git a/docs/ipa-vocal-tract-roadmap.md b/docs/ipa-vocal-tract-roadmap.md new file mode 100644 index 0000000..91cec26 --- /dev/null +++ b/docs/ipa-vocal-tract-roadmap.md @@ -0,0 +1,426 @@ +# IPA Vocal Tract Roadmap + +AquaSynth can get close to a Pink Trombone-class tract model, then move past it +by treating IPA as an authoring input rather than the synth itself. The machine +we want is not a giant table from symbol to sound. It is a pipeline: + +```text +IPA text + -> phonetic tokens + -> language phonology profile + -> timed articulatory gesture plan + -> morphology-specific tract and excitation curves + -> Faust DSP + -> rendered speech, song, chant, and nonhuman vocalization +``` + +The target user feeds Weksa text as IPA or IPA-like phonetic strings. AquaSynth +keeps the real authority below that surface: phonetic features, gestures, +anatomy, excitation, and acoustic simulation. + +## Objective + +Build a research-grounded vocal-tract instrument for human and alien speech: + +- parse IPA into structured phonetic intent; +- lower phonetic intent through a language profile into articulatory gestures; +- render those gestures through human and nonhuman tract morphologies; +- keep coarticulation continuous rather than stepping between phoneme presets; +- make every layer testable with mocks, fixtures, and deterministic outputs; +- grow supported IPA exponentially as the tract model earns new physics. + +The Weksa goal is conlang speech rendering for Zyphos worldbuilding. IPA gives +authors a familiar entry point. The tract model gives us the power to make Weksa +sound like Weksa instead of like English wearing ceremonial markup. + +## Research Spine + +The first implementation should stand on old, durable work: + +- Kelly-Lochbaum / digital waveguide vocal tract models simulate 1D wave + propagation through connected tube sections using reflection coefficients. + This is the practical spine behind Pink Trombone-like realtime tract synthesis. +- Pink Trombone demonstrates that a simplified tract, glottis, turbulence, oral + cavity, and nasal coupling can produce interactive intelligible speech-like + sound in realtime. +- Liljencrants-Fant-style glottal source models give a better voiced excitation + than a naive sawtooth or pulse train, and can parameterize voice quality. +- Articulatory phonology and task dynamics provide the right control model: + phonetic units become overlapping gestures, not isolated frames. +- Vocal-tract area-function work is the middle layer between anatomy and DSP: + vowels and constrictions can be expressed as cross-sectional area over tract + distance before being lowered to reflection coefficients. +- IPA feature databases and phoneme-table implementations are reference + material for metadata, not authorities over the sound. They help keep the + parser and feature mapper honest while the gesture and morphology layers own + synthesis. +- Source-filter speech theory still matters, but this lane should not collapse + back into static formant filters. Formants are useful diagnostics; tract shape + is the synthesis authority. + +Useful starting sources: + +- Pink Trombone / Neil Thapen interactive tract model: + +- Pink Trombone overview at IMAGINARY: + +- Recent Pink Trombone optimization paper describing PT as a realtime + Kelly-Lochbaum tract model: + +- Story 2005 vocal-tract area-function model: + +- Kelly-Lochbaum / time-domain articulatory synthesis summary: + +- Waveguide vocal tract acoustics: + +- Voice source model comparison including LF: + +- LF model fitting for speech synthesis: + +- Haskins TADA gestural planning model: + +- PanPhon IPA feature vectors: + +- PHOIBLE phoneme inventories and distinctive features: + +- eSpeak NG phoneme-table implementation reference: + + +## Architecture + +### 1. IPA Parser + +Owns Unicode IPA tokenization and notation, not sound. + +Inputs: + +- IPA string; +- optional language tag; +- optional inline stress, length, tone, syllable, and boundary marks. + +Outputs: + +- ordered `PhoneticToken` records; +- diacritic attachments; +- suprasegmental marks; +- source spans for diagnostics. + +The parser should be table-driven and deterministic. It should reject unknown +clusters with useful diagnostics instead of guessing heroically and making a new +little bureaucracy out of Unicode. + +### 2. Phonetic Feature Mapper + +Owns IPA-to-feature interpretation. + +Examples: + +- `/p/` -> voiceless bilabial stop; +- `/ɬ/` -> voiceless alveolar lateral fricative; +- `/qʼ/` -> uvular ejective stop; +- `/aː/` -> long open vowel, front/central depending on profile defaults. + +Outputs: + +- `PhoneticSegment` with manner, place, voicing, airstream, length, stress, + vowel height/backness/rounding, and diacritics. + +This layer should support language profiles. IPA has broad and narrow uses, and +Weksa may assign phonemic categories to articulatory targets that are only +human-adjacent. PanPhon, PHOIBLE, and eSpeak NG are useful references for feature +schemas and inventory sanity checks, but AquaSynth should own its feature record +instead of importing a third-party schema as the permanent internal model. + +### 3. Gesture Planner + +Owns time and coarticulation. + +Inputs: + +- phonetic segments; +- language timing profile; +- prosody vector; +- speaking style; +- morphology capability map. + +Outputs: + +- overlapping gestures for tract constriction, tongue body, tongue tip, lips, + velum, glottis, pressure, turbulence, burst release, and auxiliary organs. + +This is where consonants and vowels smear into each other. Stops close before +their release. Vowels pull neighboring consonants. Fricatives hold a narrow +constriction and noise source. Nasals open a branch. Ejectives and clicks become +pressure/airstream events rather than cute punctuation. + +The gesture planner should resemble a small task-dynamics compiler: phone +features become overlapping constriction tasks and source events. It should not +linearly interpolate phone presets unless a test proves that shortcut works for +the current stage. + +### 4. Morphology Model + +Owns anatomy. + +Human baseline: + +- tract length; +- section count; +- palate/tongue mapping; +- lip opening/protrusion; +- velum and nasal branch; +- one glottal source; +- oral tract radiation. + +Alien/Weksa extensions: + +- longer or shorter tract families; +- asymmetric or split oral branches; +- additional resonant sacs; +- secondary glottal or membrane exciters; +- nonhuman nasal/side-channel coupling; +- constrained articulator ranges by species or caste; +- morphology-specific mappings from human IPA-like features to physical organs. + +IPA does not get to own alien anatomy. IPA names intent; morphology realizes it. + +### 5. Tract DSP + +Owns sound. + +First spine: + +- 1D Kelly-Lochbaum-style tube sections; +- reflection coefficients from section area; +- glottal excitation at the source; +- lip radiation; +- nasal branch switch/coupling; +- turbulence injection at constriction points; +- closure and burst primitives for stops. + +Use area functions as the stable middle representation: a morphology and gesture +plan produce tract-area curves over time; the DSP compiler turns those curves +into reflection coefficients, source injection points, and coupling events. + +Later spine: + +- losses and dispersion; +- lip/larynx impedance refinements; +- side branches and sacs; +- multiple sources; +- differentiable or search-assisted parameter fitting for vowel targets; +- morphologies that are not merely scaled human throats. + +Generated Faust should remain boring enough to inspect. If tract DSP becomes +large, generate a specialized Faust module from a typed tract plan instead of +stuffing a hundred magic controls into ordinary `voice`. + +## Supported IPA Roadmap + +Do not support all IPA first. That is the trap where the parser looks mighty and +the synth says five ugly vowels. Expand by doubling the useful surface each time +the physical model earns another category. + +### Stage 0: Token Spine + +Goal: lossless-enough IPA tokenization and diagnostics. + +Support: + +- whitespace and word boundaries; +- `/.../` and `[...]` wrappers; +- primary and secondary stress; +- length marks `ː` and `ˑ`; +- common tie mark for affricates; +- combining diacritic storage without full interpretation. + +Tests: + +- round-trip token spans; +- unknown-symbol diagnostics; +- diacritic attachment fixtures. + +### Stage 1: Vowels And Simple Voice + +Goal: intelligible sustained vowel targets through a human tract. + +Support: + +- monophthong vowels: `i y ɨ ʉ ɯ u ɪ ʏ ʊ e ø ɘ ɵ ɤ o ə ɛ œ ɜ ɞ ʌ ɔ æ ɐ a ɶ ɑ ɒ`; +- length and stress as duration/prosody modifiers; +- modal voiced glottal source; +- simple pitch, intensity, and breathiness controls. + +Machine growth: + +- vowel feature space; +- tract area targets informed by area-function references; +- glottal source model; +- vowel fixture renderer. + +Validation: + +- check approximate formant regions for `/i e a o u ə/`; +- compare tract-area curves against known vowel-shape expectations; +- keep listening fixtures because a formant table can still lie with confidence. + +### Stage 2: Core Pulmonic Consonants + +Goal: CV, VC, and CVC syllables with recognizably different places and manners. + +Support: + +- stops: `p b t d k g ʔ`; +- nasals: `m n ŋ`; +- fricatives: `f v s z ʃ ʒ h`; +- approximants: `w j l ɹ`; +- basic aspiration `ʰ`; +- syllable and word timing. + +Machine growth: + +- closure/release events; +- plosive burst noise; +- fricative turbulence source; +- nasal branch; +- simple coarticulation curves. + +Validation: + +- verify voice onset timing for stops; +- verify burst timing and broad noise bands; +- verify nasal/oral contrast; +- reject fricative support that is only a filtered noise preset with no + constriction/source location. + +### Stage 3: Expanded Human IPA + +Goal: useful conlang coverage without pretending to be a universal phonetician. + +Support: + +- places: dental, alveolar, postalveolar, retroflex, palatal, velar, uvular, + pharyngeal, glottal; +- manners: trills, taps/flaps, lateral fricatives, lateral approximants, + affricates; +- symbols: `θ ð ɕ ʑ ç ʝ x ɣ χ ʁ ħ ʕ ɸ β ʂ ʐ ɲ ɳ ɴ ɾ r ʀ ɭ ʎ ʟ ɰ`; +- affricate tie sequences such as `t͡s`, `d͡ʒ`, `t͡ɬ`; +- voiceless/voiced and dentalized/palatalized/labialized/velarized diacritics + where they map to real gestures. + +Machine growth: + +- tongue tip/body split; +- lateral constriction; +- trill/tap event generators; +- profile-specific place maps. + +This is the point where a language profile starts earning its keep. The same IPA +symbol may need different timing, place bias, or allophonic behavior by language; +that belongs in the profile, not in the global token parser. + +### Stage 4: Airstream And Phonation + +Goal: Weksa can use sounds that feel physically intentional instead of being +ornamental apostrophes. + +Support: + +- ejectives: `ʼ`; +- implosives: `ɓ ɗ ʄ ɠ ʛ`; +- clicks: `ʘ ǀ ǃ ǂ ǁ`; +- breathy, creaky, slack/stiff voice diacritics; +- voiceless vowels and sonorants; +- tone marks if Weksa needs lexical or prosodic tone. + +Machine growth: + +- pressure reservoirs; +- glottal closure timing; +- ingressive/egressive event modeling; +- richer glottal source parameterization; +- non-pulmonic test fixtures. + +### Stage 5: Weksa And Alien Morphologies + +Goal: IPA can drive nonhuman vocal anatomy without reducing aliens to exotic +human accents. + +Support: + +- morphology profiles with declared organ inventory; +- feature remapping per morphology; +- unsupported-feature diagnostics that explain the missing organ or gesture; +- dual-source phonation; +- resonant sacs and side branches; +- nonhuman lateral or split-tract continuants; +- authored Weksa phoneme inventories and allophones. + +Machine growth: + +- morphology capability negotiation; +- tract graph beyond a single tube plus nasal branch; +- multi-source DSP; +- Weksa fixture corpus; +- Zyphos scene hooks for speaker, body, place, and emotional pressure. + +## Testing And Mocking + +The implementation should be built as a set of injectable services: + +- `IIpaTokenizer` +- `IPhoneticFeatureMapper` +- `IPhonologyProfile` +- `IGesturePlanner` +- `ITractMorphology` +- `ITractPlanCompiler` +- `ITractFaustEmitter` +- `IAudioRenderer` +- `IReferenceFixtureStore` + +Mock seams are not ceremony here. They protect the machine: + +- tokenizer tests do not need DSP; +- feature tests do not need coarticulation; +- gesture tests can snapshot control curves; +- morphology tests can assert capabilities and remaps; +- Faust emitter tests can verify generated structure without rendering; +- renderer tests can compare short deterministic buffers; +- Weksa profile tests can prove unsupported IPA fails clearly. + +Golden fixtures: + +- IPA tokenization fixtures; +- feature bundle fixtures; +- gesture-plan JSON fixtures; +- tract-area curve fixtures; +- rendered vowel and CVC WAV artifacts; +- Weksa phrase fixtures once the language profile exists. + +Metrics: + +- formant target distance for vowels; +- envelope and burst timing for stops; +- noise-band distance for fricatives; +- nasal/oral spectral contrast; +- intelligibility/listening notes for complete syllables; +- script/IPA terseness only after audio is credible. + +## First Work Packets + +1. Add `docs/ipa-vocal-tract-roadmap.md` and preserve the boundary doctrine. +2. Add an IPA tokenizer and feature model with Stage 0/1 fixture tests. +3. Add a human baseline vowel planner that emits tract-area targets, not audio. +4. Add a tiny tract renderer prototype in AquaSynth.Faust or a test-only lab + lane, then decide the package boundary once the prototype proves itself. +5. Render five vowels and three CVC syllables through fixtures. +6. Add the first Weksa phonology profile with explicit unsupported-feature + diagnostics. +7. Only then add Stage 2 consonant synthesis to the shipping path. + +## Non-Goals + +- Full CFD, finite-element airflow, or anatomical medical simulation. +- A black-box neural TTS voice that hides anatomy and cannot explain itself. +- A complete IPA parser before the tract can render basic syllables. +- Alien sounds as random effects. Nonhuman output needs morphology, not glitter. diff --git a/docs/reference-synth-roadmap.md b/docs/reference-synth-roadmap.md index 43c118d..d61d1cc 100644 --- a/docs/reference-synth-roadmap.md +++ b/docs/reference-synth-roadmap.md @@ -51,6 +51,29 @@ missing graph feature instead of smuggling it through unrelated parameters. ## Target Families +### New Lane: IPA Vocal Tract / Weksa + +This is not merely another reference synth. It is a new authoring and rendering +lane: IPA text lowers into phonetic features, timed articulatory gestures, +morphology-specific tract/excitation curves, and Faust DSP. The roadmap lives in +[`docs/ipa-vocal-tract-roadmap.md`](ipa-vocal-tract-roadmap.md). + +Pressure on AquaSynth DSL: + +- IPA as an input surface, not an internal patch model. +- Phonology profiles separated from tract morphology. +- Gesture planning and coarticulation before audio rendering. +- Human and alien vocal anatomies driven by the same structured phonetic intent. +- Mockable tokenizer, feature mapper, gesture planner, morphology, emitter, and + renderer seams. + +First deliverable: + +- Stage 0/1 IPA tokenization and vowel feature fixtures. +- Human baseline vowel tract-area planner. +- A small render proof that can say a few vowels before the supported IPA table + expands. + ### Tier 1: DX7/Dexed Use this first. DX7 SysEx is compact, structured, and musically meaningful: diff --git a/state/evidence.jsonl b/state/evidence.jsonl index a7ef5a0..c6e43b8 100644 --- a/state/evidence.jsonl +++ b/state/evidence.jsonl @@ -104,3 +104,4 @@ {"date":"2026-05-17","topic":"zyn-formant-parity-fixtures","lesson":"The first formant-motion parity workout needs multiple multi-frame ADD/formant fixtures and a real Zyn ADD note oracle. Avoid letting drum-kit envelope/source failures masquerade as formant evidence. Use Wide Bass, Ghost Ensemble, and Ghost Choir as the initial chorus; all currently fail enough to be useful.","evidence":"Added test-only `tools/zyn-reference/ZynVoiceReference.cpp`, wired it into the MSYS reference build, and added `ZynVoiceReferenceRendererSurveysUpstreamFormantFixturesWhenBuilt`. The batch writes ignored artifacts under `artifacts/parity/zyn-upstream-formant-fixtures/`; baseline scores were Wide Bass 0.280990, Ghost Ensemble 0.312813, Ghost Choir 0.363644. The native Zyn reference build succeeded and the focused formant fixture test passed."} +{"date":"2026-05-19","evidence":"Added docs/ipa-vocal-tract-roadmap.md with staged IPA support from tokenization/vowels through pulmonic consonants, expanded IPA, airstream/phonation, and Weksa alien morphologies; linked it from docs/reference-synth-roadmap.md and state.","lesson":"AquaSynth should treat IPA as an authoring input for phonetic intent, not as the internal synth model. The planned machine is tokenizer -\u003e feature mapper -\u003e phonology profile -\u003e gesture planner -\u003e morphology -\u003e tract plan -\u003e Faust renderer, growing supported IPA only as tract/excitation physics earn each category.","topic":"ipa-vocal-tract-roadmap"} diff --git a/state/map.yaml b/state/map.yaml index 2f123d6..f30b962 100644 --- a/state/map.yaml +++ b/state/map.yaml @@ -16,6 +16,7 @@ surfaces: patch_library: patches tests: tests/AquaSynth.Dsl.Tests reference_roadmap: docs/reference-synth-roadmap.md + ipa_vocal_tract_roadmap: docs/ipa-vocal-tract-roadmap.md faust_toolchain_boundary: docs/faust-toolchain-boundary.md dx7_public_domain_fixtures: tests/AquaSynth.Dsl.Tests/Fixtures/Dx7/PublicDomain zyn_reference_source: external/zynaddsubfx @@ -36,6 +37,7 @@ modules: dx7_reference_render_tests: tests/AquaSynth.Dsl.Tests/Dx7ReferenceParityTests.cs zyn_reference_reader: src/AquaSynth.Core/ZynInstrument.cs zyn_coverage_survey: src/AquaSynth.Core/ZynInstrument.cs + ipa_vocal_tract_roadmap: docs/ipa-vocal-tract-roadmap.md package_boundary_tests: tests/AquaSynth.Dsl.Tests/PackageBoundaryTests.cs sfxr: src/AquaSynth.Core/SfxrParams.cs analysis: src/AquaSynth.Core/AudioAnalysis.cs @@ -57,6 +59,7 @@ boundaries: - Faust source generation in AquaSynth.Core - target-code generation, compile-product contracts, dynamic patch compilation policy, cache identity, generated artifact manifests, and Faust toolchain selection in AquaSynth.Faust - live patch compilation, native Faust factory lifetime, and offline sample-buffer render proof in AquaSynth.Faust + - IPA-to-phonetic-feature planning, articulatory gesture planning, and human/alien vocal-tract roadmap design dev_only: - public-domain DX7 SysEx fixtures - pinned GPL ZynAddSubFX source for parity tests diff --git a/state/memory.json b/state/memory.json index b48a6ab..4901175 100644 --- a/state/memory.json +++ b/state/memory.json @@ -79,10 +79,16 @@ "ZynAddSubFX instruments for additive, subtractive, PAD, formant, and layered patches", "Faust examples/libraries for general DSP graph topology", "Surge XT for modern hybrid synth modulation", - "Cardinal/VCV-style patches for modular cable/rate semantics" + "Cardinal/VCV-style patches for modular cable/rate semantics", + "IPA/Weksa vocal tract synthesis for phonetic gesture planning, human and alien tract morphologies, and conlang speech rendering" ], "rule": "Add DSL abstractions only when reference rebuilds prove the missing invariant. Do not consume references that cannot be backed by parity tests; parsing alone is inventory, not evidence of expressivity." }, + "ipa_vocal_tract_path": { + "status": "Roadmap drafted for an IPA-to-vocal-tract lane. IPA is an authoring input only: tokenization and feature mapping lower into language phonology profiles, timed articulatory gestures, morphology-specific tract/excitation curves, and Faust DSP. The staged support plan starts with tokenization and vowels, then core pulmonic consonants, expanded human IPA, airstream/phonation, and Weksa/alien morphologies.", + "lesson": "Do not build a complete IPA parser before the tract can say basic syllables. The durable machine is tokenizer -> feature mapper -> phonology profile -> gesture planner -> morphology -> tract plan -> Faust renderer, with dependency-injected seams and mocks at each boundary. Weksa should use IPA-like phonetic input without letting human IPA semantics own alien anatomy.", + "next": "Implement Stage 0/1 fixtures: IPA token spans, vowel feature bundles, human baseline vowel tract-area targets, and a tiny render proof before expanding consonants." + }, "dx7_reference_path": { "status": "DX7 SysEx parser extracts voice edit buffers, packed voices, packed 32-voice banks, six operators, envelopes, algorithm, feedback, oscillator sync, LFO, transpose, voice name, checksum validation, ReferencePatch features, algorithm topology from the reverse-engineered OPS algorithm ROM table, envelope-to-ADSR approximations, measured carrier output amplitude, isolated two-op modulation route scale, summed/cascaded route scale, calibrated feedback amount table, ratio detune lowering, voice transpose lowering, fixed-frequency operator lowering, ROM COM output compensation cut, operator key/rate scaling at the effective played note, topology-aware route lowering, staged operator envelopes with per-segment curves, raw and block-interpolated DX7 envelope tracing, applied-envelope curved lowering with optional sustain floor, two AquaSynth DSL rebuild attempts, first-class OperatorGraph support with graph-level vibrato, public-domain SysEx fixture coverage, dexed-py reference rendering, project-authored Dexed patch rendering, AquaSynth candidate rendering through Faust-generated C#, one thresholded rendered parity test for analog1.syx voice 13, a hard PRC SYNTH1 parity probe with listening WAV artifacts and log-mel <= 0.145, project-authored algorithm-8 summed-stack/cascade/summed-pair parity, rendered community parity for Mooger, Piano Bass, RES SYNTH1, and ANLGSYN 1 with a sustained high-band gate, a MELLOWSOLO pressure artifact test, and renderable delayed operator feedback.", "lesson": "Algorithm 32's additive six-carrier shape maps cleanly to six AquaSynth voices. Algorithm 8 justified OperatorGraph: operators, modulation edges, carriers, feedback, and graph-level pitch motion belong to a graph authority, not Voice fields. The first rendered parity rung is behavioral and sine-like, not exact DX7 operator execution. A harder algorithm-8 target exposed that feedback must be delayed recursion in Faust, not a direct cyclic self-reference. Static DX7 output-level approximation helped import/rebuild math but was mathematically wrong for carrier amplitude. A single-carrier Dexed sweep shows carrier amplitude follows 2^((outputLevel - 99) / 8). Isolated two-op DX7 FM maps full-level modulation to AquaSynth route index 6.275; summed/cascaded algorithm-8 branches now use 6.0 after PRC listening, a project-authored combined stack probe, and split cascade/summed-pair probes showed lower values were under-bright. Feedback lowers through a fitted table instead of PRC-local constants; ANLGSYN 1 showed max feedback value 7 needs a hotter AquaSynth amount 2.2 to recover buzzy high bands, while value 5 remains 0.19 for PRC. ANLGSYN 1 then exposed a sustained-buzz gap: output soft clip helped but did not stop mid-note drift; max-feedback source operators need an applied-envelope sustain floor so the feedback loop stays driven after the attack. ANLGSYN 1's remaining beat is pitch LFO: AMD is nonzero but all operator AMS values are 0, while PMD/PMS lower cleanly to operator-graph vibrato with faded delay. Ratio-mode detune belongs in operator frequency lowering. DX7 voice transpose belongs in graph base frequency: transpose 24 is neutral, transpose 12 lowers MIDI note 60 to 130.8128 Hz. DX7 fixed-frequency operators are absolute Hz, not note ratios; ANLGSYN 1 proved fixed op1/op3 must lower to ratios around 0.007615/0.009227 against a 130.8128 Hz graph base to preserve low buzzing modulation. ROM COM carrier compensation was cut: user listening found no-COM renders harmonically closer and mainly too quiet, so `OperatorOutputCompensation` returns unity and graph gain owns loudness recovery. Operator key-scaling and rate-scaling must lower against the effective played MIDI note after transpose; hardcoded MIDI 60 made RES SYNTH1 too soft on attack and Piano Bass less harmonically accurate. Project-authored Dexed patch rendering now gives lawful calibration targets. Reusing the operator output curve for EG levels was tested and cut because it failed hard-target log-mel; later output-level-dependent and naive traced-release applied-envelope experiments were also cut because they worsened broader parity. Direct envelope tracing shows current linear `env=rl` cannot express DX7 EG spike/drop behavior. Dexed-style audio samples EG once per 64-sample block and interpolates applied gain, so raw EG parity alone is not audio parity. Staged operator envelopes now support linear/exponential segments and supranormal levels. `ApproximateAppliedRateLevelEnvelope` uses that surface and now picks the first near-peak before gate so slower rising modulators hit early instead of swelling late; it can also preserve a floor for max-feedback source sustain. Community gains with no COM and played-note scaling: Mooger 0.75, Piano Bass 0.72, RES SYNTH1 0.75. Algorithm-2 non-carrier applied envelopes currently cap the first peak at 0.92 to prevent ANLGSYN 1/RES SYNTH1 peak blowout while preserving Piano Bass. MELLOWSOLO is a pressure artifact test, not passing parity; its render window is 1.25s so the candidate release is audible instead of cut at the artifact boundary. Mooger needs a more aggressive attack contour, and RES SYNTH1's phase-like attack issue is likely relative operator attack/peak shape because oscillator sync is enabled. PRC remains passing on log-mel/envelope/RMS, with score floor lowered to 0.74 after rate scaling changed the aggregate score.",