transcribe
Whisper runs on your device — a 2-hour podcast never leaves your laptop. Verify: DevTools → Network.
transcript
segments — click a timestamp to seek
Audio is normalized with FFmpeg.wasm to 16 kHz mono, then transcribed by OpenAI's Whisper
(base) running on-device via Transformers.js — WebGPU when available, WebAssembly otherwise.
The first run downloads the ~150 MB model once and caches it. Nothing is uploaded; export to
SRT, VTT or plain text.