Live Meeting Helper

Real-time meeting transcription with AI-powered insights. Captures live audio from your microphone, transcribes it in real-time using Google Cloud Speech-to-Text, and provides AI-generated summaries and answers using Google Gemini.

Features

• Real-Time Transcription: Live speech-to-text as you speak, with interim results and automatic punctuation
• AI Meeting Assistant: Ask questions about the meeting or get automatic summaries powered by Gemini
• Streaming Responses: AI answers stream in real-time for immediate feedback
• WebSocket Architecture: Low-latency audio streaming for responsive transcription
• Audio Level Monitoring: Visual feedback showing microphone input levels
• Microphone Testing: Built-in mic test to verify audio setup before recording
• Transcript Download: Export your meeting transcripts as text files
• Dark Theme: Easy-on-the-eyes dark interface designed for extended use

Tech Stack

Layer	Technology
Frontend	Next.js 16, React 19, TypeScript, Tailwind CSS 4
Backend	Node.js custom server with WebSocket support
Speech Recognition	Google Cloud Speech-to-Text (streaming API)
AI Assistant	Google Gemini via LangChain
Package Manager	pnpm (monorepo workspace)

Requirements

• Node.js 20+ (Node.js 25+ recommended)
• pnpm 10+ - Fast, disk space efficient package manager
• Google Cloud Account - For Speech-to-Text API
• Google AI API Key - For Gemini AI

Installation

# Clone the repository
git clone https://github.com/WSzP/live-meeting-helper.git
cd live-meeting-helper

# Install dependencies
pnpm install

Configuration

Create a .env.local file in the project root:

# Google Cloud Speech-to-Text credentials
GOOGLE_APPLICATION_CREDENTIALS=.gcp/your-service-account.json

# Google AI (Gemini) API key
GOOGLE_API_KEY=your-google-ai-api-key

# Optional: Custom port (default: 3000)
PORT=3000

Google Cloud Setup

1. Create a Google Cloud project
2. Enable the Speech-to-Text API
3. Create a service account with Speech-to-Text permissions
4. Download the JSON key file and place it in .gcp/ (do NOT commit it)

See docs/GOOGLE_CLOUD_SETUP.md for detailed instructions.

Google AI (Gemini) Setup

1. Go to Google AI Studio
2. Create an API key
3. Add it to .env.local as GOOGLE_API_KEY

Usage

# Start development server
pnpm dev

Open http://localhost:3000 in your browser.

How to Use

1. Click Start Recording to begin
2. Allow microphone access when prompted
3. Speak clearly into your microphone
4. Watch real-time transcription appear
5. Click Ask AI to get insights about the meeting
6. Click Stop Recording when finished
7. Download your transcript

Project Structure

live-meeting-helper/
├── frontend/                 # Next.js application
│   ├── src/
│   │   ├── app/             # App Router pages and layouts
│   │   ├── components/      # React components
│   │   └── hooks/           # Custom React hooks
│   ├── server.js            # Custom Node.js server with WebSocket
│   ├── next.config.ts       # Next.js configuration
│   ├── eslint.config.mjs    # ESLint flat config
│   └── postcss.config.js    # PostCSS/Tailwind configuration
├── docs/                    # Documentation
├── .gcp/                    # Google Cloud credentials (gitignored)
├── package.json             # Root package.json (workspace)
└── pnpm-workspace.yaml      # pnpm workspace configuration

How It Works

1. Audio Capture: Browser captures microphone audio using MediaRecorder API (WebM/Opus format)
2. WebSocket Streaming: Audio chunks are streamed to the server via WebSocket every 250ms
3. Speech Recognition: Server pipes audio to Google Cloud Speech-to-Text streaming API
4. Real-Time Results: Interim and final transcription results are sent back to the client
5. AI Processing: When requested, transcript is sent to Gemini (gemini-3-flash-preview) via LangChain, with responses streaming back through the same WebSocket connection

Architecture

┌─────────────┐     WebSocket      ┌─────────────────┐
│   Browser   │ ◄────────────────► │  Node.js Server │
│  (React)    │   Audio + Text     │                 │
└─────────────┘                    └────────┬────────┘
                                            │
                         ┌──────────────────┼──────────────────┐
                         │                  │                  │
                         ▼                  ▼                  ▼
                 ┌───────────────┐  ┌───────────────┐  ┌──────────────┐
                 │ Google Cloud  │  │   Google AI   │  │    Next.js   │
                 │ Speech-to-Text│  │   (Gemini)    │  │   (SSR/CSR)  │
                 └───────────────┘  └───────────────┘  └──────────────┘

Development

# Install dependencies
pnpm install

# Start development server
pnpm dev

# Build for production
pnpm build

# Start production server
pnpm start

# Run linting
pnpm lint

Limitations

• 5-minute streaming limit: Google Cloud Speech-to-Text has a ~305 second limit per stream. The app automatically restarts streams to handle longer meetings.
• Browser support: Requires a modern browser with MediaRecorder API support (Chrome, Firefox, Edge)
• English only: Currently configured for en-US. Can be modified in server.js

Brand

Colors

Name	Hex	Usage
Night Navy	#040932	Background
Cloud Lilac	#F3F1FA	Primary text color over dark backgrounds
Pulse Cyan	#0BB2F2	"Live" feeling - realtime indicators, highlights, badges, waveform accents

Logo

In the logo text: Live and Helper use Cloud Lilac, Meeting uses Pulse Cyan.

Asset	Path	Usage
Logo (PNG)	assets/lmh-logo.png	High quality
Logo (WebP)	assets/lmh-logo.webp	Web optimized
Open Graph	assets/lmh-open-graph.png	Social media previews

License

Apache License 2.0

---

Created by Peter W. Szabo, 2026.