whisper-web

Why we included this project

Web developers who want speech-to-text without standing up a backend or paying per-minute API fees will find a working blueprint here. The repo runs OpenAI's Whisper models entirely in the browser via Transformers.js, so audio never leaves the user's device, which is a real privacy win for transcription features. It is more of a working demo and starting point than a drop-in library: clone it, start the dev server, and it transcribes in multiple languages with optional timestamps. An experimental branch moves the heavy lifting to WebGPU for GPU acceleration, and a live hosted demo lets you test the approach before adopting it.