Our voice
technology

SoapBox has a fundamentally different approach to voice AI

Most voice-enabled learning tools have been built using consumer-focused and generic LLMs. These pre-trained systems perform with low accuracy for education use cases such as foundational literacy and the younger the student, the poorer their performance. Many of these systems also suffer from bias. 

In a high stakes environment like education, voice AI systems must deliver teacher-level accuracy, work in noisy classrooms and understand every student’s accent and dialect equally.

Voice-enabled tools must also cater to students at every stage of the literacy journey.

SoapBox has gone narrow and deep to build a voice engine that is safe and accurate for K-12 learning tools.

Our voice engine is built on our large and proprietary database of tens of thousands of hours of natural child speech from children in 193 countries. In independent evaluations, we have been proven to deliver teacher-level accuracy and to show no bias across the wide range of accents and dialects present in classrooms across the US.