Sensory Unveils TrulyNatural Speech-To-Text on the Edge

TrulyNatural Speech-To-Text technology offers unrivaled accuracy and multilingual support in a compact package.

Santa Clara, October 26 – Sensory, a trailblazer in advanced speech and vision technologies, proudly announces TrulyNatural Speech-To-Text (STT). This innovative technology redefines the landscape of voice AI speech recognition for small compute platforms by setting new standards for natural language accuracy, language diversity, and power and energy efficiency – ON DEVICE!

TrulyNatural STT is a remarkable achievement in the realm of AI-powered speech recognition. Designed to operate on edge devices, this technology offers an unparalleled combination of accuracy and versatility, catering to a wide array of applications and products. Its compact size, requiring as little as  20MB of memory, ensures optimal performance without compromising device resources. Additionally, models can be run both with and without a GPU, enabling CPU-based execution on embedded platforms that do not support GPU inference like automotive or medical devices.

One of the crowning features of TrulyNatural STT is its astonishing accuracy. Preliminary testing demonstrates that TNL STT surpasses well known players across a wide spectrum of languages.

TrulyNatural STT is currently supported in 34 languages with a current roadmap set to support 40 or more languages by the end of this year.  This includes both well-established languages and those with complex phonetics and dialects, a testament to Sensory’s commitment to traditionally under-resourced or under-represented languages like Bengali, which is the 7th most widely spoken language in the world.  Languages currently supported: Afrikaans, Arabic, Belarusian, Bengali, Bulgarian, Catalan, Croatian, Czech, Danish, Dutch, English, Finnish, French, German, Greek, Hindi, Hungarian, Indonesian, Italian, Japanese, Korean, Mandarin, Norwegian, Polish, Portuguese, Romanian, Russian, Spanish, Swahili, Swedish, Thai, Turkish, Ukrainian, Vietnamese. Languages under development: Cantonese, Australian & British English, Hebrew, Indonesian, Malay, and Taiwanese Mandarin.

“At Sensory, our mission is to push the boundaries of what’s possible in AI-driven technologies. With TrulyNatural STT, we are redefining the benchmarks for accuracy and efficiency in speech recognition,” said Todd Mozer, CEO at Sensory. “Our technology’s performance showcases our dedication to providing a superior user experience while operating in a private setting without needing to send personal data to the Cloud.”

The TrulyNatural STT embedded SDK, scheduled for release later this year, will empower developers to integrate this cutting-edge technology into products seamlessly. Custom language models and vocabularies can be added to the TrulyNatural end to end model to improve performance in targeted domains. Developed in C, the SDK is compatible with a wide range of operating systems, including Android and Linux and seamlessly integrates with Sensory’s widely adopted TrulyHandsfree.

About Sensory

Sensory is a leading provider of cutting-edge speech and vision AI technologies. With a strong focus on innovation, privacy and user experience, Sensory develops solutions that redefine human-device interaction. The company’s extensive portfolio of technologies includes TrulyHandsfree, TrulyNatural, and TrulySecure, which together have shipped in over 3 billion products by companies such as Amazon, Google, Microsoft, Sony, Samsung and hundreds of more leaders in automotive and consumer electronics.