Transcripter.ai

AI-powered meeting transcription & summarization platform with dual-pipeline architecture.

Transcripter.ai is a full-stack application that joins your Google Meet and Microsoft Teams meetings as a bot, captures live audio/captions, transcribes speech locally using WhisperX, generates professional Minutes of Meeting using Gemini AI, and delivers the summary via email — complete with steganographic watermarking for authenticity verification.

---

Features

• 🎵 Real-Time Audio Capture (GMeet) — Intercepts WebRTC audio streams via monkey-patched RTCPeerConnection, no screen recording needed
• 🗣️ Local Speech-to-Text — WhisperX runs entirely on-device (CPU/GPU) — your audio never leaves your machine
• 🤖 Caption Scraping (Teams) — Joins Teams meetings via Playwright and captures closed captions in real-time
• 📝 AI Summarization — Generates structured MoM (Minutes of Meeting) using Gemini AI
• 🔒 Steganographic Watermarking — AES-256-CBC encrypted summaries embedded into PNG images via LSB steganography
• 📧 Email Delivery — Sends transcript, summary, and stego image via SendGrid
• 🖥️ Premium Dashboard — Glassmorphism dark-themed UI with real-time meeting status tracking
• ☁️ Deployable — Headless Chromium for VPS deployment (no display needed)

---

Architecture

┌──────────────────────────────────────────────────────┐
│                    Frontend (Next.js 16)              │
│         Landing Page │ Auth │ Dashboard               │
│              Vercel / localhost:3000                   │
└──────────┬───────────────────────┬────────────────────┘
           │  HTTP (fetch)         │  HTTP (fetch)
           ▼                      ▼
┌────────────────────────┐  ┌────────────────────────────┐
│   GMeet Pipeline ⭐     │  │   Teams Pipeline            │
│   (Primary / Advanced) │  │   (Express API :4000)       │
│                        │  │                             │
│  WebRTC Audio Capture  │  │  Playwright Headless Bot    │
│  ↓                     │  │  Caption Scraping (DOM)     │
│  FFmpeg → WAV (16kHz)  │  │  ↓                         │
│  ↓                     │  │  Gemini → Stego → Email    │
│  WhisperX Local STT    │  └────────────────────────────┘
│  ↓                     │
│  Gemini → Stego → Email│
└────────────────────────┘

Dual Pipeline Strategy

Pipeline	Method	Capture	Best For
GMeet Pipeline ⭐	WebRTC interception → WhisperX (local STT) → Gemini	Real audio stream	Full transcription with timestamps, speaker diarization, offline-capable
Teams Pipeline	Playwright bot → caption DOM scraping → Gemini	Live captions	Lightweight, no GPU needed, deployable to any VPS

---

Tech Stack

Frontend

• Next.js 16 — React framework with App Router
• NextAuth.js v5 — Authentication (credentials provider)
• better-sqlite3 — Lightweight embedded database
• bcryptjs — Password hashing

GMeet Pipeline (Primary)

• Playwright — Headless browser with persistent Chrome profile for Google login
• WebRTC Interception — Monkey-patches RTCPeerConnection to capture raw audio tracks
• MediaRecorder — Records intercepted audio in WebM/Opus format
• FFmpeg — Converts WebM → WAV (16kHz mono) for WhisperX
• WhisperX — Local speech-to-text with word-level timestamps (runs on CPU or GPU)
• Gemini AI — Generates professional Minutes of Meeting from transcript
• SendGrid — Email delivery with transcript + summary + stego image

Teams Pipeline

• Express.js 5 — API server to spawn bot processes
• Playwright — Headless browser automation (joins as "Transcript Bot")
• MutationObserver — Captures closed captions from the DOM with stability detection
• Gemini AI — AI summarization
• SendGrid — Email delivery
• Sharp + pngjs — Image processing & LSB steganography

---

Quick Start

Prerequisites

• Node.js ≥ 18
• npm ≥ 9
• FFmpeg installed and on PATH (for GMeet audio conversion)
• Python 3.10+ with WhisperX installed (for GMeet local STT)
• API keys for Google Gemini and SendGrid

1. Clone the Repository

git clone https://github.com/iamanimeshdev/Transcripter.ai.git
cd Transcripter.ai

2. Set Up the Frontend

cd frontend
npm install

Create frontend/.env:

NEXTAUTH_SECRET=your-secret-key
NEXTAUTH_URL=http://localhost:3000
TEAMS_API_URL=http://localhost:4000

3. Set Up the GMeet Pipeline

cd gmeet-pipeline
npm install
npx playwright install chromium

Create gmeet-pipeline/.env:

GEMINI_API_KEY=your-gemini-api-key
SENDGRID_API_KEY=your-sendgrid-api-key
EMAIL_FROM=your-verified-sender@email.com
EMAIL_TO=default-recipient@email.com
GOOGLE_EMAIL=your-google-email
GOOGLE_PASSWORD=your-google-password
PYTHON_PATH=path/to/whisperx/python

4. Set Up the Teams Pipeline

cd teams-pipeline
npm install
npx playwright install chromium

Create teams-pipeline/.env:

GEMINI_API_KEY=your-gemini-api-key
SENDGRID_API_KEY=your-sendgrid-api-key
EMAIL_FROM=your-verified-sender@email.com
EMAIL_TO=default-recipient@email.com
FRONTEND_URL=http://localhost:3000

5. Run Locally

# Terminal 1 — Teams Pipeline API
cd teams-pipeline
npm start
# → 🎯 Teams pipeline server running on http://localhost:4000

# Terminal 2 — Frontend
cd frontend
npm run dev
# → http://localhost:3000

6. Use It

Via Dashboard (Teams):

1. Open http://localhost:3000 → Sign up → Login
2. Click "+ Start New Meeting"
3. Paste your Microsoft Teams meeting URL
4. The bot joins headlessly, captures captions, and emails you the summary when the meeting ends

Via CLI (GMeet):

cd gmeet-pipeline
node gmeet.js --url "https://meet.google.com/xxx-xxxx-xxx"

The bot logs into Google, joins the meeting, captures audio via WebRTC, transcribes locally with WhisperX, and emails the summary when the meeting ends.

Via CLI (Teams — standalone):

cd teams-pipeline
node teams.js --url "https://teams.live.com/meet/..." --email recipient@example.com

---

Security Features

• bcrypt password hashing with salt rounds
• NextAuth.js session management with JWT
• LSB Steganography — Meeting summaries are invisibly embedded into PNG images, providing a cryptographic watermark to verify transcript authenticity
• Environment variables for all API keys (never hardcoded)

---

License

This project is built for educational and portfolio purposes.

---

Built with ❤️ using Next.js, Playwright, Gemini AI, and a lot of coffee ☕