ScribeAI - AI-Powered Meeting Transcription

Real-time audio transcription and AI-powered meeting summaries using Next.js, Socket.io, and Google Gemini

Demo Video

---

Overview

ScribeAI transforms live audio into searchable, summarized transcripts in real-time. Capture meeting audio from your microphone or browser tabs (Google Meet, Zoom, YouTube) and receive instant AI-powered transcriptions with automatic summaries.

Built for: AttackCapital Technical Assignment
Timeline: 4 days
Status: ✅ Production-ready prototype

---

Features

Core Functionality

• Real-time Transcription - Live streaming audio chunks (5s intervals) to Gemini API
• Dual Input Sources
• Microphone recording
• Browser tab audio capture (Google Meet, Zoom, Spotify, YouTube)
• Pause/Resume - Control recording flow with state preservation
• AI-Powered Summaries - Automatic generation of key points, action items, and decisions
• Session Management - Complete history with search, filter, and export capabilities
• Live UI Updates - Real-time transcript display via Socket.io

---

Tech Stack

Frontend

• Next.js 16.0.1
• Tailwind CSS + shadcn/ui
• Zustand (State Management)
• Socket.io Client

Backend

• Node.js 20+ with TypeScript
• Socket.io Server
• Express
• Prisma ORM

AI & APIs

• Google Gemini 2.5 Flash (Transcription)
• Gemini via Vercel AI SDK (Structured Summaries)

Database

• PostgreSQL

Authentication

• Better Auth

DevOps

• pnpm (Package Manager)
• Turbopack (Next.js Dev Server)

---

Installation

Prerequisites

• Node.js 20+
• PostgreSQL 15+ (local or cloud)
• pnpm installed (npm install -g pnpm)
• Google Gemini API Key (Get one free)

Setup Instructions

1. Clone Repository

git clone https://github.com/RitamPal26/ScribeAI.git
cd ScribeAI

2. Install Dependencies

pnpm install

3. Setup Database

   a. Create Supabase Project:

   • Go to Supabase Dashboard
   • Click "New Project"
   • Choose a name and set a secure database password
   • Wait for project initialization (~2 minutes)

   b. Get Database URL:

   • Go to Project Settings → Database
   • Scroll to Connection String section
   • Copy the Connection pooling URI (recommended for production)
   • Format: postgresql://postgres.[project-ref]:[password]@aws-0-[region].pooler.supabase.com:6543/postgres

4. Configure Environment Variables

cp .env.example .env

And put in all the required varibles

5. Start Development Server

pnpm dev

---

Project Structure

ScribeAI/
├── src/
│   ├── app/                      # Next.js App Router
│   │   ├── (dashboard)/
│   │   │   └── dashboard/
│   │   │       ├── page.tsx      # Dashboard home
│   │   │       ├── record/       # Recording interface
│   │   │       └── sessions/     # Session history
│   │   ├── api/                  # API routes
│   │   ├── login/                # Auth pages
│   │   └── signup/
│   │
│   ├── components/
│   │   ├── auth/                 # Auth forms
│   │   ├── dashboard/            # Dashboard components
│   │   ├── recording/            # Recording UI
│   │   ├── sessions/             # Session cards
│   │   └── ui/                   # shadcn/ui components
│   │
│   ├── hooks/
│   │   ├── useRecording.ts       # Core recording logic
│   │   └── useSocket.ts          # Socket.io client hook
│   │
│   ├── lib/
│   │   ├── auth.ts               # Better Auth config
│   │   ├── prisma.ts             # Database client
│   │   └── socket-client.ts      # Socket.io setup
│   │
│   └── stores/
│       └── recordingStore.ts     # Zustand state
│
├── server/
│   ├── services/
│   │   ├── audioProcessor.ts     # Gemini integration
│   │   └── sessionService.ts     # Database operations
│   │
│   ├── socket/
│   │   └── recording.ts          # Socket.io handlers
│   │
│   └── index.ts                  # Server entry point
│
├── prisma/
│   └── schema.prisma             # Database schema
│
└── server.js                     # Socket.io server

---

Screenshots

Session Details

View complete transcript and download options

AI Summary

Automatic summary with key points and action items

Session History

Browse and manage all recorded sessions

Live Transcription

Real-time text appears as you speak

Dashboard Overview

Main dashboard showing session statistics and recent recordings

Landing Page

Simple landing page

---

Architecture & Design

For complete system architecture, data flow diagrams, and technical decisions:

👉 View Architecture Documentation

---

Demo Video

📹 Watch Full Walkthrough

Demonstration includes:

• ✅ Microphone recording with live transcription
• ✅ Tab audio capture from YouTube video
• ✅ Pause/Resume functionality
• ✅ Stop recording & AI summary generation
• ✅ Session management & transcript export

---

Testing & Quality

Tested Scenarios

• ✅ 5-minute continuous microphone recording
• ✅ Tab audio from Google Meet call
• ✅ Pause/Resume mid-recording
• ✅ Network interruption recovery
• ✅ 1-hour marathon session

Code Quality

• TypeScript strict mode enabled
• ESLint + Prettier formatting
• JSDoc comments on key functions
• Prisma type safety throughout

---

License

MIT License - See LICENSE file

---

Author

Ritam Pal

• GitHub: @RitamPal26
• Email: ritamjunior26@gmail.com

Built as part of AttackCapital technical assignment (November 2025)

---

Acknowledgments

• Google Gemini API for powerful audio transcription
• Vercel AI SDK for structured output generation
• Better Auth for authentication
• shadcn/ui for beautiful components
• AttackCapital for giving this idea