In-browser AI

Real models, downloaded once and run on your GPU (or on WASM when WebGPU is missing). Nothing is pre-computed.

Speech-to-text

Whisper tiny.en (40 MB) transcribes English speech. Record, upload a file, or use the hello.wav fixture. Audio is resampled to 16 kHz mono before it reaches the model.

onnx-community/whisper-tiny.en

~40 MB

Model not loaded yet.

The microphone is only used to capture a short clip that is transcribed locally. Nothing leaves your browser.

Any format the browser can decode works; it is converted to 16 kHz mono Float32 on the fly.

Devicenot loaded