Whisper, packed
This is OpenAI's Whisper speech-to-text model running entirely in this browser tab on WebGPU, from a quantized weight pack that is downloaded once and cached by your browser. Nothing you record or load is uploaded anywhere.
WebGPU is not available in this browser, so this page cannot run Whisper. Try a recent Chrome, Edge or Safari on a device with WebGPU.
Download: about 47 MB, first use only.