BrailleVision

Inspiration

Over 253 million people worldwide live with visual impairment. While Braille is their primary gateway to literacy and independence, there is a massive communication barrier: 90% of sighted teachers, parents, and caregivers cannot read Braille dots.

Existing assistive devices are bulky, hard to source, and cost upward of ₹50,000—making them inaccessible to those who need them most. We built BrailleVision to break this barrier, turning any standard smartphone into an instant, free, AI-powered Braille reader.

What it does

BrailleVision is a fully accessible web platform designed to empower visually impaired individuals and their supporters in real-time.

• Live Braille Scanner: Decodes Grade 1 Braille instantly using the phone's camera, displaying clear English translations on the screen.
• Scene Vision (Real-time Spatial Awareness): Scans the environment to identify objects, structures, and potential safety hazards (obstacles, stairs, etc.) and speaks them aloud.
• Automated Text-to-Speech (TTS): Automatically reads aloud translations and environmental descriptions for complete hands-free operation.
• Voice Control & Commands: Users can control scanning, text synthesis, and navigation entirely via simple voice commands.
• History & Analytics: Keeps track of previous scans, allowing users to save favorites and track their Braille learning journey with beautiful progress graphs.

How we built it

We engineered a highly responsive, mobile-first frontend using React + Vite and Tailwind CSS.

• AI OCR & Vision: Integrated OpenAI's GPT Vision API via secure backend endpoints to analyze dot coordinates and transcribe printed text alongside Braille.
• Voice Integration: Leveraged the native Web Speech API for ultra-fast, low-latency audio readouts and offline-capable text-to-speech.
• Data & Subscriptions: Used the Base44 Platform SDK to manage user scan histories, real-time database state, and user authentication seamlessly.
• Animations & Design: Built beautiful, highly interactive transitions and progress charts using Framer Motion and Recharts.

Challenges we ran into

• Lighting and Angle Variations: Camera captures of raised Braille dots depend heavily on shadow contrast. We overcame this by designing a robust system prompt for the Vision model that acts as an expert interpreter, filtering out noise, angle skews, and lighting issues.
• Accessibility Design: Making an app designed for the visually impaired required precise keyboard/screen-reader compatibility. We implemented clear audio cues, auto-speak actions on scan completions, and large, high-contrast, screen-reader friendly buttons.

Accomplishments that we're proud of

• Zero-Hardware Translation: We proved that a free web-browser solution can match the utility of expensive ₹50,000+ hardware readers.
• Dual Mode Capability: Successfully combined Braille decoding with general object/hazard detection (Scene Vision) under a single lightweight web app.
• Polished UX: Designed a beautiful interface with a scanning line animation, progress streaks, and interactive historical trends that both sighted guides and blind users can enjoy.

What we learned

• We learned how crucial immediate audio feedback is for visual accessibility tools.
• We realized the immense power of combining LLMs with vision inputs to translate tactile languages like Braille that previously required specialized hardware.

What's next for BrailleVision

• Multilingual Support: Adding support for Hindi Braille (Bharati Braille) and other regional scripts.
• Offline Mode: Porting translation logic locally using lightweight on-device models so it works in areas with poor internet connection.
• Companion Mobile App: Launching native iOS and Android versions with deeper haptic vibration feedback for physical guiding alerts.
• Built with

Built With

• base44-backend
• framer-motion
• icons
• lucide
• openai-vision-api
• react
• recharts
• tailwind-css
• typescript
• vite
• web-speech-api