SightMate is an AI-driven, voice-first assistant designed to empower blind and visually impaired individuals. It combines vision, audio, and language models into an accessible AI-first platform to assist with real-world challenges.
Features
- Real-Time Road Guidance: Live camera stream interpreted with LLaVA, providing obstacle and road condition alerts through audio feedback.
- Daily News Summarizer: Fetches real-time news, summarizes with Mixtral LLM, and reads out top headlines.
- Document & Handwriting Reader: Extracts, summarizes, and reads aloud scanned documents or handwritten notes.
- Indian Currency Recognition: Detects INR denominations, calculates totals, and reads out values.
- Artistic Scene Description: Generates creative or poetic descriptions of surroundings to enhance joyful interaction.
- Personalized Voice Interaction: Natural and expressive TTS for engaging conversations.
Tech Stack
- Frontend: Next.js, TailwindCSS, ShadCN, Framer Motion
- Backend: FastAPI (Python)
- Database: SQLite
- Hosting: Vercel (Frontend), Render (Backend)
- Groq API:
- LLaVA for image-based understanding
- TTS for expressive voice generation
Demo Video:
SightMate isn’t just a project — it’s a mission.
A mission to make the world more inclusive, one intelligent voice at a time.