Speech to Text

This runs Whisper tiny, an OpenAI speech-recognition model compiled to WebAssembly (WASM) and executed in your browser. The first run downloads roughly 40 MB of model data (cached afterwards); transcription is offline and your audio is never uploaded.