Science

Why the concept is plausible, and why it is still not magic.

SONIC VEIL AB-01 is framed around published ingredients: ultrasonic audible enclaves, multi-directional parametric arrays, head-tracked personal sound zones, target-speaker extraction, and explicit ultrasound safety envelopes.
NODE ANODE BLOCAL AUDIBLE ENCLAVEUltrasound beams stay narrow, intersect, then form the useful speech region at the overlap.Outside the overlap, the system aims for weaker, less intelligible leakage instead of claiming zero sound.

Perceptual privacy bubble

Speech inside becomes easier to understand; speech outside becomes weaker and harder to parse.

No claim of perfect silence, zero leakage, or universal isolation in any room.

AI-assisted selective listening

The stack can privilege enrolled voices in a crowd and improve the useful signal for the pair.

No claim that acoustics alone solve the cocktail-party problem.

Safe operating envelope

Tracking loss, calibration drift, and uncertain exposure states must trigger lower output and transparent UI warnings.

No claim that futuristic visuals waive the need for exposure limits, measurement, or certification.

Layer 01 · Directional ultrasonic emitters

prototype

Four ceiling or desktop nodes launch steerable ultrasonic carriers that aim toward a shared conversation volume.

The system does not claim a magical invisible wall. It claims controlled sound delivery: narrow aim, lower spill, and a stable geometric envelope around the two speakers.

Layer 02 · Active field shaping

productization

A beam-control stack continuously trims side lobes and reduces unintended audible spill outside the bubble.

This is where product discipline matters: array calibration, phase alignment, room profiling, and a lower-power idle mode when certainty drops.

Layer 03 · Target-speaker extraction

prototype

Microphone arrays capture the mixed scene, then focus the speech model on the two enrolled voices instead of every nearby talker.

This is the answer to the real problem. In a crowd, acoustics alone cannot perfectly erase interference, so selective listening has to become part of the stack.

Layer 04 · Tracking and fail-safe control

research

Head and mouth tracking update the control geometry in real time and rapidly reduce output when the system loses confidence.

Perfect privacy is not promised. A safe, measurable degradation path is. When lock quality drops, the bubble shrinks, the level falls, and the UI surfaces uncertainty.

Primary sources

Every core marketing claim points back to a paper, a safety statement, or a named engineering boundary.

The site avoids “black box future tech” language. If a capability is not yet demonstrated as a product, it is framed as a prototype target or a research route.

Audible enclaves crafted by nonlinear self-bending ultrasonic beams

PNAS · 2025

Crossing ultrasonic beams can form a localized audible spot instead of flooding the whole room with audible sound.

This is the closest published proof-point for a headphone-free sound enclave, but it is still a research prototype rather than a consumer product.

Open source

MuDiS: An Audio-independent, Wide-angle, and Leak-free Multi-directional Speaker

MobiCom / Microsoft Research · 2024

Parametric arrays can be steered in multiple directions while suppressing unintended leakage better than traditional single-beam systems.

Useful for the outer shell of the product: how the system aims sound, trims spillover, and builds steering logic.

Open source

SANN-PSZ: Spatially Adaptive Neural Network for Head-Tracked Personal Sound Zones

arXiv · 2024

Head-tracked personal sound zones can adapt filter coefficients in response to listener motion, improving robustness in real rooms.

This supports the system's re-lock behavior and explains why tracking is part of the architecture instead of a cosmetic add-on.

Open source

Look Once to Hear: Target Speech Hearing with Noisy Examples

arXiv · 2024

A short, noisy enrollment clip plus visual focus can extract a target speaker in crowded scenes with embedded-friendly latency.

This is the strongest justification for pairing directional acoustics with target-speaker extraction instead of pretending acoustics alone solve the cocktail-party problem.

Open source

ICNIRP Statement on Ultrasonics

Health Physics · 2024

Ultrasonic systems need explicit exposure envelopes, measurement discipline, and conservative operating limits.

This shapes the fail-safe story: power down on tracking loss, cap drive levels, and verify exposure rather than hand-wave it.

Open source

Guidelines for the Safe Use of Ultrasound: Part II - Industrial & Commercial Applications

Health Canada Safety Code 24 · 2020

Industrial and commercial airborne ultrasound use is governed by exposure limits, engineering controls, and hearing-protection logic rather than visual hype.

Useful for translating the concept from lab demo language into a credible product safety discipline.

Open source