TEDMaster is an AI-driven TED speech study platform built with Next.js. It integrates Google Gemini AI and Whisper transcription technologie to provide users with an immersive bilingual learning experience, smart vocabulary analysis, and deep content insights.
- 🤖 AI Deep Insight: Uses Google Gemini 1.5 Flash for topic analysis, difficulty identification, and in-depth vocabulary extraction from speeches.
- 🌍 Bilingual Subtitle Sync: Precision-synced English and Chinese subtitles with real-time word lookup support.
- 🎙️ Smart Transcription: Integrated Whisper/Moonshine technology to extract and transcribe audio directly from YouTube URLs.
- 🔍 Intelligent Search: Built-in efficient video search functionality to quickly find relevant study materials.
- 💳 Credit System: A complete user credit and quota management system covering word lookups, AI analysis, and membership levels.
- 📄 Multi-Format Export: Export analyzed content and subtitles to PDF or Markdown for offline study.
- 📱 Responsive Design: optimized for both desktop and mobile devices for learning anytime, anywhere.
- Frontend: Next.js 15+ (App Router), TypeScript, Tailwind CSS
- Backend Logic: Next.js Server Actions & API Routes
- Database: PostgreSQL (Prisma ORM)
- AI Engine: Google Gemini 1.5 Flash
- Containerization: Docker & Docker Compose
- Authentication: JWT (JSON Web Tokens)
- Node.js 18+ or 20+
- PostgreSQL Database
- Google AI Studio Gemini API Key
Copy and edit .env.local:
cp .env.example .env.localFill in GEMINI_API_KEY, JWT_SECRET, and DATABASE_URL.
# Install dependencies
npm install
# Initialize database
npx prisma generate
npx prisma migrate dev
# Start development server
npm run devAccess: http://localhost:3000
We provide detailed production deployment guides:
Default Port: The production image maps to port 3005 by default.
src/app: Core page routes & API endpointssrc/components: UI componentssrc/lib: Core utilities (AI integration, i18n, rate-limiting, etc.)prisma: Database schema & migration filespublic: Static assets and local model weights
This project is licensed under the MIT License.
Note
The first time you perform a browser-based transcription, the Whisper model weights (~150MB) will be downloaded. A stable internet connection is recommended.