Real-time audio translation for any browser tab.
Transcribes locally with Whisper, translates, and reads it back aloud — no API key, no account.

---

📥 Install

Add AuraLang from the Chrome Web Store →

Prefer to run it from source? See Run it locally.

✨ What it does

Watching a talk, a podcast, or a video in a language you don't speak? AuraLang listens to the tab, translates on the fly, and speaks the result — so you can watch the screen instead of reading subtitles.

• 🎧 Any tab with audio — videos, calls, podcasts, live streams.
• 🧠 On-device transcription — Whisper runs locally; audio never leaves your machine.
• 🗣️ Spoken translation — the original tab audio is muted; you only hear the translation.
• 📊 Live transcript with karaoke — read the original + translation in a side panel; the line being spoken is highlighted.
• 🎚️ Model modes — Auto / Light (whisper-tiny) / Balanced (whisper-base): trade speed for accuracy per your device.
• 🔒 No key, no account, no backend — settings live only in your browser.
• 🌗 Light & dark themes, interface in English and Spanish.

⚙️ How it works

Tab audio ─▶ Whisper (local transcription) ─▶ Google Translate ─▶ Web Speech API (spoken output)

Audio is buffered and cut on natural speech pauses, transcribed on-device, translated, and read aloud through the browser's built-in voice. The original tab audio is muted while capturing.

🚀 Run it locally

Requirements: Google Chrome 116+

npm install
npm run build

Then in Chrome:

1. Open chrome://extensions
2. Enable Developer mode (top-right)
3. Click Load unpacked → select the dist/ folder
4. On the tab you want to translate, click the AuraLang icon to open the side panel
5. Pick source & target languages, then hit Start translation

The Whisper model downloads on first Start and is cached afterwards. Size depends on the model mode: Light ~150 MB, Balanced ~290 MB. Auto picks one from your device.

🛠️ Development

npm run dev         # Vite dev server (reload the extension after changes)
npm run type-check  # TypeScript, no emit
npm run lint        # ESLint
npm run test        # Jest + React Testing Library
npm run zip         # Build + package dist.zip for the Web Store

🧱 Tech stack

Layer	Tool
Extension API	Chrome MV3 — tabCapture, offscreen, sidePanel, storage
Build	Vite + @crxjs/vite-plugin
UI	React 18 + TypeScript (strict)
Styles	Tailwind CSS
Transcription	Whisper via @huggingface/transformers (local, offscreen document)
Translation	Google Translate (free, unofficial endpoint)
Speech	Web Speech API (speechSynthesis)

📁 Project structure

src/
  asr/          # Model tiers/registry, model manager (probe ladder), inference queue
  background/   # Service worker — tabCapture + offscreen + side panel lifecycle
  offscreen/    # AudioContext + transcription / translation / TTS pipeline
  popup/        # React UI (rendered in the side panel)
    components/ # Header, StatusHero, WaveformIndicator, LanguageSelect, SettingsPanel, TranscriptFeed…
    hooks/      # useApiConfig, useTranslation, useI18n, useTheme
  welcome/      # First-install onboarding page
  services/     # transcriptionService, translationService, ttsService
  types/        # Shared TypeScript interfaces

🔐 Privacy

No backend, no telemetry, no account. Transcription happens on your device; only the transcribed text is sent to Google Translate to get the translation back. Settings (languages, theme) are stored in chrome.storage.local on your machine. Full details in PRIVACY.md.

📚 Docs

• AGENTS.md — architecture, TypeScript, services, testing & accessibility conventions (source of truth for contributors).
• CONTRIBUTING.md — setup, branching, commit conventions.
• PRIVACY.md — privacy policy.
• DISCLAIMER.md — accuracy limitations, third-party services, no affiliation.
• SECURITY.md — how to report a vulnerability.
• LICENSE — MIT.

📦 Publishing

npm run zip

Upload dist.zip to the Chrome Developer Dashboard.

---

_{Built with care — concepts before code.}