AI and agentsMobile appLumyeye2025
Lumyeye, an AI app that describes the world out loud
Lumyeye helps blind and visually impaired people every day. We built the fully voice-driven app, its AI server and its back office.
- The need
- For a blind person, reading a letter, recognising an object or finding information often means asking for help. The founder of Lumyeye wanted an app that understands what you ask out loud and answers out loud, with no screen to look at.
- What we built
- An iOS and Android app driven entirely by voice, compatible with VoiceOver and TalkBack
- Photo descriptions, text reading and live object detection with the camera
- Voice web search
- Fifteen languages in the app and about fifty synthetic voices
- Accounts, in-app subscriptions and activation codes
- A back office for users, campaigns and the AI instructions in each language
- How we worked
- Most of the work was architecture. A spoken request had to become a precise action. A first call to Gemini acts as a dispatcher. It decides whether to take a photo, run a search, start object detection or ask for more detail, and the app triggers the right action. The answer is split into sentences and each one is read as soon as it is ready, so the app speaks without waiting for the full answer.
- Technologies
- React Native
- Node.js
- Socket.IO
- Neo4j
- Gemini
- Google Vision
- Google Text-to-Speech
- React
- The result
- The app was released on the App Store in May 2025. It is rated 4.3 out of 5 and is on its fourth major version.
