SonicReassurance

Development biography · earlier archive

Development profile compiled from project source and retained records in October 2026. Dates and test results describe those records; they do not imply a new build, live acceptance test or public release.

_Source-backed, read-only. Scope: the project archive Delivered inline ._

Biography

Sonic Reassurance (product name "MAUMAU") is a small, sharply-focused SwiftUI application that listens to the microphone and renders nearby sound as a restrained halftone field of dots. It is deliberately not a music visualizer in the usual sense — its design intent is the opposite of spectacle. Where a visualizer adds richness and surprise, this adds legibility: it gives a live sound a second, time-locked, predictable sensory channel, so the event can be coded as one benign thing rather than an ambient surprise to scan for. The README bounds the claim honestly — "an accessibility-oriented interface prototype, not a diagnostic or medical device" — which is the right posture for something aimed at people who find unpredictable soundscapes disorienting (for example some autistic or sensory-sensitive users).

The mapping from sound to image is intentionally simple and consistent, because predictability is the therapeutic idea. Loudness drives local dot size and a broad "surface pressure"; a sound onset fires a single expanding pressure ring; low frequencies cause broad swelling, mid frequencies clustered movement, high frequencies a fine shimmer; stereo balance shifts the field's centre to approximate horizontal direction; a stability measure settles the motion when a soundscape is steady; and everything decays gently back to equilibrium when sound stops . The same sound feature always maps to the same visual feature — that fixed grammar is what makes the surface reassuring rather than agitating.

Under the hood it is a clean, dependency-free DSP + SwiftUI pipeline across 13 Swift files (~900 LOC). MicrophoneAudioAnalyzer installs an AVAudioEngine input tap (256-frame buffer, 5 ms preferred I/O buffer on iOS) on a userInteractive queue and streams buffers to AudioFeatureExtractor. The extractor downmixes to mono, computes an RMS→dB→normalised level, runs fast/slow envelope followers (0.52 / 0.055 coefficients) whose difference yields onset detection, tracks transient strength and age with exponential decay, and applies an asymmetric attack envelope (0.40 rising / 0.085 falling). Every half-window it runs a 1024-point Hann-windowed FFT, split into three dB-normalised bands (low 40–250 Hz, mid 250 Hz–4 kHz, high 4 kHz–min(12 kHz, Nyquist·0.48)). Stability is 1 − 4.5·stddev over the last ~70 levels; direction is (R−L)/(R+L) from per-channel RMS. The FFT itself, Radix2FFT, is a hand-written iterative radix-2 transform with bit-reversal and precomputed sin/cos tables — notably with no Accelerate dependency, keeping the app self-contained and portable. Each cycle emits a SonicVisualFrame, a small Sendable/Equatable value type carrying level, transient, transientPhase, low/mid/highEnergy, direction and stability.

Rendering is a single SwiftUI Canvas inside a 60 fps TimelineView. SonicDotFieldView lays out a 23×31 dot grid and, per frame, superposes a broad Gaussian "pressure" (centred and nudged by direction plus idle drift), a clustered mid-energy sinusoid, a high-energy shimmer, and a Gaussian transient ring at an expanding radius — then maps the combined pressure to each dot's radius (0.9 up to the profile maximum) and opacity (capped at the profile maximum). Behaviour is shaped by four SensoryProfile presets — Still / Soft (default) / Responsive / Expressive — each a tuple of idleDrift, maximumOpacity, maximumRadius, transientGain, spectralGain, motionRate and responseGain. Still removes almost all idle motion; Expressive raises contrast and motion for creative use. VisualSettings exposes live sensitivity (0.35–1.8), surface movement (0–1.5), and toggles for direction, status text and auto-start.

Interaction is minimal by design: tap the surface to pause/resume listening, long-press (0.45 s) for settings, with a small LIVE/PAUSED pill top-right. The app auto-starts on launch and when returning to the foreground, and stops when backgrounded, returning the field to rest via a .silent frame. Privacy is a first-class property: PrivacyInfo.xcprivacy declares no tracking, no collected data types and no accessed-API reasons, and there is no networking code anywhere in the project — microphone audio is analysed entirely on-device and never leaves it. Originally an iPhone/iPad app, it was ported to a native (non-Catalyst) macOS build via one multiplatform target, so the same calm surface now runs on the Mac as a resizable window.

User journey: open the app → grant mic access → the field wakes and breathes with the room → tap to pause when wanted → long-press to pick a calmer (Still) or livelier (Expressive) profile and tune sensitivity → background the app and it quietly stops.

Modules / features (what a visitor can see, hear, try)

Authored functionality:

• Live dot-field "sound surface" (23×31 grid, 60 fps) — the main thing you see; reacts to level, onset, bands, direction, stability, decay. (SonicDotFieldView)
• Four sensory profiles Still/Soft/Responsive/Expressive — visibly change motion, contrast, dot size. (SensoryProfile)
• Settings sheet — profile segmented picker; sensitivity slider (0.35–1.8); surface-movement slider (0–1.5); toggles for show-direction, show-status-text, auto-start. (SettingsView, VisualSettings)
• Tap-to-pause/resume, long-press-for-settings, LIVE/PAUSED status pill. (RootView)
• Accessibility value on the surface (quiet / moderate / strong). (SonicDotFieldView)
• On-device feature extraction + hand-written FFT. (AudioFeatureExtractor, Radix2FFT)

Nothing audible is produced — it is sound-in, light-out (no audio playback feature). Dependencies: Apple frameworks only (SwiftUI, AVFoundation, CoreGraphics, Foundation). No third-party runtime libraries.

Technical foundations & local packages

• Language/UI: Swift 5.0, SwiftUI (Canvas + TimelineView), CoreGraphics drawing.
• Audio: AVFoundation (AVAudioEngine tap; AVCaptureDevice permission, unified across iOS/macOS; AVAudioSession on iOS only, guarded by #if os(iOS)).
• DSP: first-party — envelope-follower onset detection, windowed-FFT band energies, custom Radix2FFT (no Accelerate).
• Local/third-party packages: NONE. No Swift Package Manager dependencies, no CocoaPods/Carthage. The only external tool is XcodeGen (project.yml is the source of truth; supportedDestinations: [iOS, macOS], SDKROOT=auto).
• Signing/privacy: macOS SonicReassurance.entitlements (app-sandbox + audio-input) via CODE_SIGN_ENTITLEMENTS[sdk=macosx*]; NSMicrophoneUsageDescription set; PrivacyInfo.xcprivacy clean.

Maturity by feature (dated evidence)

• Source implemented: YES — 13 Swift files (~900 LOC); core authored ~2026-07-13 (file dates + merged export 2026-07-13T23:36).
• macOS multiplatform port: YES — present on disk in project.yml/App (supportedDestinations, entitlements), dated 2026-07-13.
• Tests passed: NONE — no test target or test files exist in the project.
• Built: YES (prior agent session, 2026-07-13) — xcodebuild succeeded for both platform=macOS and generic/platform=iOS (BUILD SUCCEEDED). NOT re-run in this read-only pass.
• Installed / runtime-on-device verified: UNKNOWN — no evidence the app was launched/run; README notes mic testing needs a physical device.
• Released / deployed (App Store / public): UNKNOWN / none found — no release notes, CHANGELOG, TestFlight or store link in the repo. MARKETING_VERSION 1.0 in build settings only.

Roadmap, limitations, FAQs

Roadmap (planned, not built): additional visual modes beyond the dot field (e.g. brush-swell, ping-pong); persist VisualSettings across launches; a reduce-motion / high-contrast accessibility profile; update README to document macOS; capture real screenshots/demos.

Limitations: not zero-latency ; direction only meaningful with ≥2 input channels, else centred ; one visual mode only; settings not persisted; not a medical/diagnostic device ; no app icon/asset catalog; README still says iOS/iPad-only despite the macOS target.

FAQs:

• Does my audio leave the device? No — all on-device; no networking code; PrivacyInfo.xcprivacy declares no collection/tracking.
• Is it a medical or therapy tool? No — an accessibility-oriented interface prototype .
• Which devices? iPhone, iPad (iOS 17+) and Mac (macOS 14+), one multiplatform build.
• Do I need a special microphone? No; any built-in mic. Stereo direction needs ≥2 channels.
• Can I make it calmer or livelier? Yes — choose a profile (Still→Expressive) and adjust sensitivity / surface movement.