The Core Spatial
Audio Engine.
Spatial audio wherever sound is created, processed, or played. NeiSa is Nsync's spatial audio engine, designed to operate wherever audio is created, processed or played back.
NeiSa turns a recording into a
three-dimensional listening field.
Two ears, and a brain that reads a room from what reaches them. That is the whole mechanism a headphone has to work with, and it is why spatial audio is a modelling problem before it is a hardware problem. Reconstruct what the ears would have received, and the sound arrives from a place rather than from a side.
NeiSa does that reconstruction in real time, through Nsync's spatial modeling, neural optimization and rendering architecture. It is software first. The application, the device and the embedded processor are hosts it adapts to — not the thing it is.
One engine. Wherever audio happens.
A software core, first.
NeiSa's binaural renderer is designed through deep neural network optimization. Neural networks derive optimal transfer functions (NRIR) from a large-scale hyper-BRIR database, and the result is distilled so the same engine runs in real time across a wide range of compute budgets.
The same engine, adapted to the compute available to it.
One architecture. Different compute budgets.
NeiSa uses a portable C/C++ architecture and has been ported across multiple DSP and application-processor families. The core rendering architecture is preserved across target-specific ports and optimizations. Given a higher-performance DSP or a mobile application processor, the same engine separates sources with AI and renders a full sphere — around you, above you and below.
What the engine does.
Supported channel configurations depend on the compute, memory and integration constraints of the target platform.
Supported spatial configurations scale with the compute and memory available on the target platform. On resource-constrained embedded targets, the core renderer can be configured within the memory and compute available to the platform.
Higher-compute application environments can support longer filters, more channels and more complex spatial processing than constrained embedded targets. Nsync has also developed personalized spatial rendering approaches that adapt the listening model beyond a single generic spatial profile.
Xphere. Any source becomes spatial.
Conventional discrete-channel binaural rendering typically begins with multichannel content. Xphere extends that pipeline to mono and stereo sources through source separation. A deep learning model separates a recording into its parts in real time, and the renderer places each part in space — so material that was never mixed for spatial audio can be rendered as a full-sphere mix.
hyper-BRIR: proprietary patented technology (KR 10-2620761 and others)
Proven across software,
content and industry.
NeiSa has been developed across software applications, media and content workflows, device prototypes and multiple processor architectures.
Renderer applications
Working renderer implementations exist across Windows, Android and iPhone / iOS application environments.
Immersive content production
Past media collaboration with Kakao Entertainment included approximately 100 immersive K-pop music videos and related content productions.
Direct project work
The technology behind Nsync has been developed through direct PoC, prototype, demonstration and project work across global consumer electronics, automotive, software and connected-device environments.
Processor portability
The renderer has been ported and evaluated across multiple DSP and application-processor architectures, including HiFi3, HiFi4, ARM Cortex-A and RISC DSP families.
Architecture families listed are evaluation targets and do not indicate a commercial relationship or endorsement.
A Grammy Award winner
advises us on sound.