Screaming Chicken Synth

A UI-first Maru instrument based on the product brief and reference mock-up. The portable face retains the reference's 4:3 proportions and dense choir, excitation, synthesis and effects panels, using shared Maru controls and one module-owned parameter schema.

Current stage: five coupled physical hypotheses, a harmonic baseline and seven presets. MIDI notes excite a bounded air reservoir and a five-layer squeaker choir. The modulation wheel (CC1), or the on-screen MOD fader, controls air.airflow: up increases volume, upper resonances, drive and air depletion; down is softer, smoother and slower. Use the host's Play button, on-screen keyboard or MIDI input to sound notes.

Squeeze Study is the default preset for new instances.

Try Coupled Scream for a pressure-driven closing/free-reed pair, and Squeeze Study for the same physical family with a learned 3.4-second human squeeze contour. Both are experimental comparisons against recorded toys. Settings selects the reed model, enables the eight-point squeeze contour and sets its duration. The existing contour editor now changes actual squeeze force.

In the coupled models, breath noise now follows actual aperture airflow and its bandwidth follows pressure-derived air speed through breath.texture. It stops without flow and changes through squeeze and recovery. The harmonic baseline retains its existing breath behavior. The articulation study documents the remaining compressed dynamics and weak high scream, with explicit temporal measurements.

Settings also exposes an optional Attack reed: a separately tuned, pressure-driven transient layered over the existing body. Its short chamber vent, level, tuning, damping and opening are editable through the usual host parameters. Disabled preserves the earlier body exactly. See the attack model and measured attack candidates.

Reed Chicken retains the stable, harmonica-inspired baseline. Try Staccato Clucks for quick taps, Slow Deflate for a held key, Breathy Sustain for continuous air and Flock Reeds for a wider ensemble. All seven are in the existing preset selector.

Finite Air is the starting behavior: pressure loss reduces vibration and brightness, with only a gentle pitch sag (default maximum 0.6 semitone). Human adds smooth irregular pressure/pitch motion; MOD → Pressure rate / Pressure depth adjust it. P → Pitch Amount / Curve / Range shape the descending pitch. Squeeze Release now rounds the dying reed and airflow tail. Auto Breath and Infinite Sustain remain selectable in settings. See reed voicing and presets.

Effects, sequencing, advanced voice modes/priority/legato and measured physical animation remain pending. The illustrated envelope/response plots are parameter models; the main scope and meters read actual host output. This is a musical model measured against public lossy toy previews, with substantial remaining spectral error and unmeasured physical geometry; no native/plugin release is qualified. Measured comparisons retain both improvements and the independent-toy regression against the prior retuned voice.

Ownership

• src/contract.ts: schema, ranges, units, choices and defaults.
• src/ui/index.tsx: the single AudioModuleFaceProps face and scoped CSS.
• src/ui/artwork.tsx: paint-only dial skins and product-specific SVG glyphs.
• src/ui/assets/chicken-rig.png: generated decorative render; no baked controls.
• src/factory-bank.ts: canonical factory voices completed from the schema.
• src/dsp/reed.ts: product harmonic reed, composing shared sine/phase primitives.
• src/dsp/processor.ts: the sole pressure/reservoir, squeaker and note-allocation owner.
• src/dsp/*-reed.ts: module-owned pressure/aperture/load hypotheses selected by the same Processor; no second host or audio path in the face.
• src/dsp/squeeze-gesture.ts: smooth force interpolation of the canonical eight excitation points, with enablement/duration owned by the schema.
• Hypothesis benchmark and executable study: versioned Audio Fit rounds, complete reference provenance, health checks and listening exports.
• Airflow voice: physical choices, CC ownership and evidence.
• Implementation journal: reference geometry, decisions, verification and next DSP work.
• DSP implementation plan: audited shared leaves and the physical-voice gap.
• Artwork provenance: generation prompt and asset identity.

The host owns parameter storage, MIDI, presets, audio, device access and meters. No audio engine or device listener lives in the face.

Catalog promotion and publication evidence: catalog publication. Qualified update timing and the narrower DSP dependency closure: publication measurement.

Reference-audio research and executable Audio Fit workflow: reference fitting.

Compare and rate the retained candidates in Audio Fit's human review UI: reference/candidate switching, scored spectra and saved judgments for each recording.

Pressure/reed diagnostics: actual state traces and physical hypotheses, reference mode analysis, and swept-flow ablation results. These are experimental comparisons, with E/F retained as listening anchors. The human reviewer shows every clip's scored spectrogram and a shared low-body frequency view.

For the remaining changing rasp at onset, see the sourced attack roughness research: mechanical modes, slot geometry, airflow history and the temporal-analysis gap. It defines the next isolated experiments without claiming a solved attack.

The reconciliation round compares exact main-instrument factory states with E/F and isolated plate-offset geometry. All23 candidates remain available in the local human reviewer; the main physical synth and the fit studies share the same Processor. Plate offset defaults to zero.