from Hacker News

I was able to run Silero VAD directly into the browser. It is very accurate and can detect speech irrespective of any language. I tried various libraries but none of them was user friendly. so, I wrote one myself. Under the hood, it uses onnx-web-runtime to run the model.

I was able to do it in 2 days. I used Claude 3.5 to help me port code from Python to JavaScript. That helped me speed up the whole process.

I am publishing the Next.js template code. Going forward, I will convert it into an npm package for easy install.

Github Code https://github.com/DictationDaddy/VAD_WEB_DEMO