Repository navigation
Live demo: https://daily-english-mission-ten.vercel.app
Highlights
🎙️ Speak, and the browser listens — no cloud (#9)
- On-device transcription: Whisper runs inside your browser (WebAssembly, Web Worker). Record your spoken English and it is transcribed in seconds, ready to review before correction.
- Pronunciation feedback: read the corrected text aloud; each word is checked against what was heard, with a score and the unrecognised words highlighted.
- Private by design: off by default and enabled with one click. Your voice never leaves the device; the only network access is a one-time ~77 MB download of the open-source model (
Xenova/whisper-base.en, Apache-2.0) from Hugging Face, cached afterwards (works offline). The ONNX runtime is served by the app itself, no CDN code.
💾 Keep your progress (#10)
- Export / import your Daily English progress from the Données page. Imports are validated strictly and merged, never overwritten.
🛠️ Quality
- CI on every pull request (#8): lint, typecheck and the full test suite, including real-Chrome integration tests.
- Local correction now also catches "I am agreed", a frequent transcription of "I am agree" (#11).
- Hardened Vercel allowlist (#6) and a more robust Chrome test harness (#12).
Quality: 565 tests · typecheck · lint · production build · verified end to end on production.
🤖 Generated with Claude Code