No reviews yetBe the first to leave a review for Humsana Voice Firwall
Maker
📌
Hi everyone.I built Humsana Voice to solve the critical privacy and latency problem with voice AI.
Current Voice AI redaction relies on Speech-to-Text (STT). You have to transcribe the audio, find the credit card number, and redact it. This adds ~300ms latency, which kills the conversational flow.
I realized we don't need to know what was said, just that a number was read. I trained a lightweight CNN to detect the visual texture of sensitive data. You can test it out on the website (it will take a few seconds initially to stabilize volume etc.). If its something you or someone you know would be interested in please drop a message and I can share more information!
