Yadong Xie / 谢亚东
ProjectsLabWritingSaved|GithubLinkedInX

AI Recorder

Client-side AI

AI
Recorder

A voice activity detector listens, and only the stretches of speech it hears go on to Whisper. Both models run in this browser. No audio leaves the page.

0:00:00

Downloaded from Hugging Face once, then kept in the browser cache. On a phone, better over Wi-Fi.

A remake of ai-recorder, which sent the speech to Whisper on a server. Models: Silero VAD and Whisper base, run with transformers.js.

How it was made →← Lab
by yadong