This was an R&D project exploring browser-based audio processing using the Web Audio API. The goal was to detect and segment speech directly in the browser without sending anything to a server—all processing happens locally.
We built a JavaScript library that could identify when someone was speaking, segment that speech into useful chunks, and handle it all in real-time. The tricky part was making it performant enough to run smoothly across different browsers and devices.
The technology worked well enough that it caught the attention of a U.S. company doing commercial voice products. They acquired the project and integrated it into their platform. It's one of those projects where the research phase turned into something actually useful.