When it comes to speech recognition, Sensory always seems to be one step ahead of the game. It’s a pretty low-key company, but they’ve made a name for themselves in their ability to be just a little bit better than competitive offerings. Now, the company is adding machine learning to its product, with the release of TrulyNatural, an embedded large vocabulary speech recognition platform, with natural language understanding. The software will find a home in IoT devices, home appliances, set-top boxes, and cars, among other places.

According to the company, “TrulyNatural will help companies move beyond the cloud to create products capable of natural language interaction without compromising their customers’ privacy and without the high memory cost of open source-based solutions.”

TrulyNatural addresses consumers’ privacy concerns, as it provides the devices with an intelligent natural-language user interface, while keeping voice data private and secure; voice requests never leave the device, nor are they ever stored. TrulyNatural can provide a natural language voice UI on just about any device, and can be deployed for domain-specific applications. It can require less than 10 Mbytes, or it can be scaled to support broad-domain applications like virtual assistants and call center chatbots with a virtually unlimited vocabulary.

Designed to run on an apps processor, TrulyNatural doesn’t require an internet connection, as all of the speech processing is done natively on the device. This results in a safe, secure, and consistent offering. It currently supports U.S. English. Later this year and next, Sensory will roll out U.K. English, French, German, Italian, Japanese, Korean, Mandarin Chinese, Portuguese, Russian, and Spanish. SDKs are available for Android, iOS, Windows and Linux.