ReadMe

A lightweight Chrome/Edge extension that records audio and transcribes it in real time.

Installation • Features • Providers • Development • License

---

Installation

For Users

1. Download the latest readme-extension.zip from the Releases page
2. Unzip the file
3. Open chrome://extensions (or edge://extensions)
4. Enable Developer mode (toggle in the top-right corner)
5. Click Load unpacked and select the unzipped folder
6. Click the ReadMe icon in your toolbar to get started

Initial Setup

1. Open the extension popup → Settings → Manage API key
2. Choose an STT provider and enter your API key
3. Save — you're ready to record

No API key? Select the Mock provider to test the full pipeline with placeholder transcriptions.

---

Features

Feature	Description
Multi-source recording	Capture from microphone, browser tab audio, or both simultaneously
Real-time transcription	Audio is chunked and transcribed on the fly as you record
Batch fallback	If live transcription misses anything, a batch pass with overlap deduplication fills the gaps
AI summary	One-click summary powered by GPT-4o-mini — key points and action items extracted automatically
Session management	Full recording history stored locally with search and playback
Export	Download transcripts as .txt, .md, or .srt (subtitles with timestamps)
Audio playback	Replay any recording directly in the popup with seek controls
i18n	English and Chinese (中文) UI
Themes	Light and dark mode
Privacy-first	All data stays in your browser; only audio is sent to your configured STT provider

---

Supported Providers

Provider	Model	Notes
OpenAI Whisper	whisper-1	Multilingual, widely supported
Deepgram	nova-2	Fast, smart formatting
SiliconFlow	SenseVoice	Good for Chinese audio
Mock	—	Offline testing, no API key needed

AI summaries require an OpenAI API key (uses gpt-4o-mini).

---

How It Works

Microphone / Tab Audio / Mix
        │
        ▼
  MediaRecorder (30-second chunks)
        │
        ├──▶ IndexedDB (persistent storage)
        │
        ▼
  Live Transcription Queue
  (every ~60 seconds → STT API)
        │
        ▼
  Overlap Deduplication
        │
        ▼
  Transcript + AI Summary

• Audio is recorded in 30-second WebM chunks and persisted to IndexedDB immediately
• Every 2 chunks (~60s), a batch is sent to your STT provider for transcription
• Adjacent batches overlap slightly; a deduplication algorithm removes repeated text
• If any batch fails during recording, it is retried automatically when you stop
• Recordings up to 4 hours / 500 MB are supported

---

Development

Prerequisites

• Node.js (v18+)
• npm

Setup

cd extension
npm install
npm run build

Load in Browser

1. Open chrome://extensions (or edge://extensions)
2. Enable Developer mode
3. Click Load unpacked → select the extension folder
4. After each rebuild, click Reload on the extension card

Scripts

Command	Description
npm run dev	Vite dev server with hot reload
npm run build	Type-check (tsc) + production build
npm test	Run unit tests (Vitest)
npm run test:watch	Run tests in watch mode
npm run lint	Lint with ESLint
npm run format	Format with Prettier

Tech Stack

• UI: React 18 + TypeScript
• Build: Vite 5
• Extension: Chrome Manifest V3
• Storage: IndexedDB (sessions & audio) + chrome.storage.local (settings)
• Audio: Offscreen document with MediaRecorder API
• Testing: Vitest + Testing Library

Project Structure

extension/
├── src/
│   ├── App.tsx                    # Popup UI (3-tab layout)
│   ├── options.tsx                # Options page (API key management)
│   ├── service_worker.ts          # MV3 background service worker
│   ├── offscreen/
│   │   ├── recording.ts           # Audio capture pipeline
│   │   ├── live-transcribe.ts     # Real-time transcription queue
│   │   ├── segmentation.ts        # WebM segmentation for batch mode
│   │   ├── transcription.ts       # Batch transcription processor
│   │   └── state.ts               # Offscreen state management
│   ├── stt/
│   │   ├── whisper.ts             # OpenAI Whisper client
│   │   ├── deepgram.ts            # Deepgram client
│   │   └── llm.ts                 # GPT-4o-mini summary generation
│   ├── components/
│   │   ├── TranscriptionView.tsx  # Recording controls & live transcript
│   │   ├── NotesView.tsx          # Session list, export, AI summary
│   │   ├── SettingsView.tsx       # Theme, language, provider settings
│   │   └── AudioPlayer.tsx        # Playback with seek controls
│   ├── db/
│   │   └── indexeddb.ts           # IndexedDB schema & CRUD
│   ├── utils/
│   │   ├── dedup.ts               # Overlap deduplication (CJK-aware)
│   │   ├── export.ts              # TXT / Markdown / SRT formatters
│   │   └── webm.ts                # WebM binary parsing
│   └── i18n.ts                    # Translations (en / zh)
├── public/
│   ├── manifest.json              # Extension manifest
│   └── icons/                     # Extension icons
└── dist/                          # Build output

---

License

MIT © 2026 Temp1258