Sensory Releases TrulyNatural™ Embedded Speech-to-Text 2.0

Highest Accuracy to Size Ratio of ANY Speech-to-Text Technology

Santa Clara, CA – Sensory Inc., a leader in voice AI and speech recognition technology, proudly announces TrulyNatural™ Embedded Speech-to-Text (STT) 2.0 solution with a new SDK offering smaller models with more accuracy, new language coverage, and enhanced compatibility for Windows platforms.

TrulyNatural™: Optimizing Voice User Interfaces for Embedded Solutions

TrulyNatural™ STT and Truly Natural “lite” are Sensory’s flagship speech-to-text technologies, renowned for delivering high accuracy, real-time responsiveness, and robust performance without relying on cloud connectivity. This makes it ideal for applications and products where network availability is unreliable or where data privacy is a paramount concern.

New Features and Enhancements:

• Windows Compatibility: In response to growing demand, Sensory has extended the compatibility of TrulyNatural STT to Windows-based platforms. Windows compatibility opens new possibilities for developers and integrators working on a variety of applications, including consumer electronics, medical, industrial, automotive, enterprise, government and more!
• Enhanced Accuracy and Speed: The updated TrulyNatural STT incorporates advanced acoustic and language models, providing even greater accuracy and faster response times. These models utilize state-of-the-art transformer-like architectures to deliver improved word error rates (WER) and robustness in noisy environments. In benchmark comparisons, TrulyNatural STT is an order of magnitude smaller yet more accurate than industry leading and public domain speech engines.
• Flexible Deployment Options: TrulyNatural STT now offers even more flexibility in deployment, supporting a wider range of hardware configurations. Whether running on GPUs, high-performance multicore CPUs, or utilizing accelerators like Arm Neon and Helium technologies, TrulyNatural is optimized to deliver the best possible voice user experience.
• Voice Assistant Capabilities: Combining Sensory’s awards winning and industry leading wake word technology, with Sensory’s domain specific Small Language Models (SLMs) Sensory can provide a domain specific assistant in as little as 35MB! Or connect Sensory’s wake word and STT to a Large Language Model (LLM) running on device or in the cloud to save money and improve privacy!

Language Coverage: TrulyNatural™ STT now supports nearly 40 languages and regional variations including:

Afrikaans, Arabic, Belarusian. Bengali, Bulgarian, Catalan, Croatian, Czech, Danish, Dutch, English (US, UK, Kids), Finnish, French, German, Greek, Hebrew, Hindi, Hungarian, Indonesian, Italian, Japanese, Korean, Malay, Mandarin, Norwegian, Polish, Portuguese, Romanian, Russian, Spanish, Swahili, Swedish, Thai, Turkish, Ukrainian, & Vietnamese

Why TrulyNatural™ Speech-To-Text?

Sensory’s TrulyNatural STT technology stands out for its ability to operate independently of the cloud, ensuring user and data privacy. Embedded models eliminate latency issues associated with cloud-based solutions, providing a seamless and responsive user experience that is always available and free from hallucinations.

TrulyNatural STT is also highly customizable, supporting multiple languages, dialects, and domain-specific vocabulary. This makes it an ideal choice for applications in automotive, healthcare, retail, media, IoT, and more, where tailored solutions are often required.

For more information contact Sensory or request the TrulyNatural Product Brief.

About Sensory

Sensory Inc. is at the forefront of voice AI innovation, delivering cutting-edge speech recognition and biometric solutions. With over 20 years of experience, Sensory’s technologies are integrated into millions of devices worldwide, powering applications that range from mobile phones and consumer electronics to automotive and industrial systems.