MedCloud Minute

Blog archive

Applied Brain Research Releases SDK for On-Device Voice AI

Applied Brain Research, a developer of AI technology for real-time on-device and edge computing, has released the ABR SDK, a production toolkit for developers building real-time voice interfaces that run entirely on edge processors.

The SDK combines the company's Niagara automatic speech recognition and Nith text-to-speech models through a single API. Processing occurs locally, allowing applications to continue operating without network connectivity while keeping voice data on the device.


ABR reports that Niagara can begin producing text within 115 milliseconds, while Nith can generate its first audio within 147 milliseconds on embedded application-class CPUs. Initial language support includes English, Spanish, Mandarin, Japanese, and Korean.

Developers receive a Python library built over a stable C ABI, with C and Java bindings planned. The SDK also supports custom vocabulary, while an optional capability lets developers create synthetic voices from reference recordings using consent-based training.

Current platforms include Linux x86-64, Linux ARM64 and Android ARM64, with acceleration available on some integrated NPUs and DSPs. ABR plans additional silicon support and RTOS deployments on Cortex-M microcontrollers with NPUs before year-end.

For edge developers, local speech processing can reduce latency and cloud dependencies while providing another option for applications where connectivity, bandwidth or data privacy makes cloud-based voice processing less practical.


Posted by MedCloudInsider Editors on 09/24/2026