Glasses for dyslexic readers
- Idea
- Design
- Build
- Shipped
- In use
Glasses with a button. Press it and a camera on the frame photographs the printed page in front of you, sends it to my server, and a reading screen reads it out loud with word-by-word highlighting. The reading screen also lets you change letter spacing, word spacing, size and font (Lexend, Atkinson Hyperlegible or OpenDyslexic), which is what actually helps a lot of dyslexic readers.
The glasses only take the photo. The server cleans it up with OpenCV and reads it with Tesseract in English, Russian and Uzbek. Parts cost about 190,000 sum.
1Key Specifications
- Camera
- AI Thinker ESP32-CAM with flash
- Power
- LiPo, TP4056 charger, MT3608 boost to 5 V
- Frame
- 3D printed in PETG, designed in Fusion 360
- Server
- Python, Flask + waitress, OpenCV, Tesseract
- Languages
- English, Russian, Uzbek
- Speech
- The browser's Web Speech API
2Functional Block Diagram
- inputbutton
- cameraESP32-CAM + flash
- networkHTTPS via hotspot
- clean-upOpenCV
- readTesseract (uz, ru, en)
- outputreading screen + speech
3How it works
- Press the button. The ESP32-CAM on the frame takes a photo with the flash on.
- It sends the photo over HTTPS, through a phone hotspot, to my server.
- The server straightens and cleans the image with OpenCV and reads it with Tesseract.
- The reading screen, on any phone or laptop, shows the text with your spacing and font, and reads it out loud word by word, into Bluetooth earbuds if you want.
4Backup plans
If the venue has no mobile data, a laptop runs the same server and the glasses switch to it. If the glasses fail on stage, the reading screen has a Take photo button that sends a phone photo through the same path.
5Status
The server is live and has read its first test photo. The frame and the pitch are being finished for demo day on Friday, 9 October 2026. The code is in a private repo for now because it has deploy details for the server in it.
6Technical Documents
7Design Files
- INOFirmware (ESP32-CAM sketch)8 KB
- Hsecrets.example.hTemplate for Wi-Fi and server settings.1 KB
- XLSXParts and costs (smeta)5.2 MB
- PYOCR pipeline (Python)7 KB
- JSReading screen (JavaScript)20 KB
The code is in a private repository for now.