A minimal single-page app for voice recording and transcription. No server required — runs entirely in the browser.
Transcribe audio files entirely in your browser using OpenAI Whisper via WebGPU. No data leaves your device.
- Record from your microphone or upload audio files
- Supported formats: MP3, WAV, WebM, OGG, FLAC, AAC, M4A, MP4, Opus
- Multilingual — auto-detects language
- WebGPU accelerated when available, falls back to WASM
- Three model sizes: Tiny (~40 MB), Base (~145 MB), Small (~244 MB)
- Models download once and are cached locally in your browser
Open index.html in a modern browser (or visit the link above). Allow microphone access, record, and get a transcription.
For local AI transcription, open transcribe.html — load a Whisper model, then record or upload an audio file.
Apache 2.0 — see LICENSE.