About BrailleVision AI

Why Accessibility Matters

285 million people worldwide are visually impaired. BrailleVision AI exists to remove one more barrier — making written Braille instantly accessible to everyone.

What is Braille?

A tactile writing system used by millions

Braille was invented in 1824 by Louis Braille, a Frenchman who lost his sight at age three. The system uses a grid of six raised dots — arranged in two columns of three — to represent letters, numbers, punctuation, and even musical notation.

Each unique combination of raised and flat dots within the six-dot cell represents a different character. Grade 1 Braille maps directly to the alphabet; Grade 2 introduces contractions and shorthand forms for common words to speed up reading.

Despite digital advances, physical Braille remains essential in signage, educational materials, medication labels, and everyday documents — places where screens are absent or impractical.

6

Dots per cell

2 columns × 3 rows

64

Unique combinations

Including blank cell

200+

Languages supported

Worldwide Braille codes

1824

Year invented

Louis Braille, France

Braille Alphabet — A to L

A
B
C
D
E
F
G
H
I
J
K
L

Filled dots represent raised embossed points. BrailleVision AI detects these dot patterns using computer vision and maps them to characters.

The Problem

Accessibility gaps that technology can close

Millions of people interact with Braille documents daily — but the gap between sighted and non-sighted access to written information remains vast.

285M

Visually impaired worldwide

43M

Completely blind individuals

90%

Of blind children lack Braille access

1 in 5

Caregivers can read Braille

Our Mission

BrailleVision AI believes that every written document — whether a school textbook, a pharmacy label, a restaurant menu, or a legal notice — should be instantly accessible to anyone who holds it, regardless of whether they can read Braille.

We build for caregivers who want to understand their loved ones' written notes. For teachers who want to read back what students have written. For volunteers in NGOs who work daily with visually impaired communities. For the visually impaired themselves — giving them a tool to verify what they've written.

Our Commitment to Accessibility

  • WCAG 2.1 AA compliance across every screen
  • Screen reader compatibility with ARIA roles throughout
  • Full keyboard navigation — no mouse required
  • High contrast mode for low-vision users
  • Large text mode for enhanced readability
  • Voice guidance status messages during scanning
  • Free forever for individual users and NGOs
Computer Vision Pipeline

How the technology works

Five sequential processing stages transform a raw image of embossed Braille into readable English text and natural speech.

Stage 1

Image Capture

WebRTC / File API

Webcam stream or uploaded JPEG/PNG. Frame extraction at 1080p for optimal dot resolution. Minimum 300 DPI recommended for embossed Braille.

Stage 2

Preprocessing

OpenCV.js

Grayscale conversion → Gaussian blur for noise reduction → Adaptive thresholding → Morphological operations to enhance dot boundaries.

Stage 3

Cell Detection

YOLO / CNN Model

Convolutional neural network identifies individual Braille cell boundaries. Each 2×3 dot grid is extracted with a bounding box and confidence score.

Stage 4

Translation

Braille Decoder

Dot pattern within each cell is mapped to the Unified English Braille code table. Grade 1 and Grade 2 contractions are resolved in context.

Stage 5

Speech Output

Web Speech API

Translated English text is passed to the browser TTS engine. Server-side fallback available for environments without Web Speech API support.

Roadmap

Where we're headed

A transparent look at what's been built, what's in progress, and what's coming next.

Q1 2026

Core Braille Scanner MVP

Complete

Camera capture, image upload, Grade 1 Braille detection, basic TTS output.

Q2 2026

Grade 2 Braille Support

Complete

Full Grade 2 contraction library, improved CNN model accuracy to 96%+.

Q3 2026

Mobile App (iOS & Android)

In Progress

Native camera access, offline processing, haptic feedback on detection.

Q4 2026

Multi-language Braille

Planned

Support for Hindi Braille, Arabic Braille, Spanish Braille, and 10 additional scripts.

Q1 2027

Real-time Video Scanning

Planned

Continuous frame analysis without manual capture — point and read like a live translator.

Q2 2027

Braille Authoring Tool

Planned

Type English text and generate a printable Braille document — bidirectional translation.

Partner Organizations

Working alongside these organizations to reach the communities that need this most.

N

National Association for the Blind

India

NGO
R

Royal National Institute of Blind People

UK

NGO
P

Perkins School for the Blind

USA

Education
L

Lighthouse for the Blind

USA

Education
B

Braille Without Borders

Global

NGO
S

Sense International

Global

NGO