Gubli A.S

Hackathon · Winner

SignBridge Ghana

An app that turns Ghanaian Sign Language into English and Twi, as text and speech, in real time. No sign-up and no uploads: the camera never leaves your device.

Event
Cursor Hackathon Ghana, 2025
Result
Winner
Type
Web app with computer vision
Role
Developer, team of three

Tools used

  • Next.js
  • React
  • TypeScript
  • Tailwind CSS
  • MediaPipe
A community left without language
A community left without language
Three steps, no setup
Three steps, no setup
Three ways to use SignBridge
Three ways to use SignBridge
Built for real situations
Built for real situations

Project overview

More than 470,000 Ghanaians live with hearing loss, and there are fewer than 50 certified sign language interpreters for 33 million people. SignBridge lets a deaf person sign to a camera and be understood, in English and in Twi.

  • Sign to the camera and see and hear the words in English and Twi
  • Learn signs with simple lesson cards
  • Type a word and see how to sign it
  • Works in a web browser, nothing to install

The challenge

Build something genuinely useful for a real community in just a few hours. It had to work on an ordinary phone or laptop, protect people's privacy, and speak Twi as well as English.

  • Nothing to install
  • Keep video private on the device
  • Support Twi, not just English

Approach

The app watches the hand through the camera and recognises the shape of each sign. Only the recognised word is ever sent out, to be translated and spoken in Twi; the video never leaves the device.

  • Private by design
  • Twi speech plays instantly
SignBridge on three phones
On a phone
Receiving the prize at the Cursor Hackathon
Receiving the prize
Two of the team deciding what to build next during the hackathon
Deciding fast, together
SignBridge recognising the sign for hello from the camera
Live demo: the sign for hello, recognised
Building SignBridge during the hackathon
Building it in the room

What this shows

We won the Cursor Hackathon in Ghana with it. A working app built by a team of three in a few hours, recognising everyday signs such as hello, help, water and school, and speaking them in English and Twi.

I can build a working, useful product under pressure, using camera recognition and local languages, and win with it.

  • Winner, Cursor Hackathon Ghana 2025
  • Camera recognition that stays private
  • English and Twi, text and speech