Dialekt

• 

  Basic landing page as this is a prototype version

• 

  Here you can select any image from gallery or you can click from camera not added yet

• 

  Here is the translation

• 

  Same here is the translation

Inspiration

While building for a hackathon, we wanted to solve something real. In Tunisia, most public signs and forms are in French or Arabic, but many locals don’t fully understand them. So we thought — why not build something that reads those signs and shows them in Tunisian dialect? A simple idea that can actually help.

What It Does

Dialekt is a small tool that:

• Takes an image of any sign, board, or document
  
• Reads the text in French or Arabic using OCR
  
• Translates it to Tunisian dialect
  
• Shows it in a clean and simple interface

It’s built to be lightweight, fast, and easy to use.

How We Built It

• Used Tesseract OCR to get text from images
  
• Made a custom rule-based translator (French to Tunisian)
  
• Ran the backend on Google Colab, exposed with ngrok
  
• Built the frontend using Streamlit
  
• Styled it using basic CSS to keep the layout clean and modern

Challenges We Faced

• OCR didn’t work well with poor quality or stylized fonts
  
• Translating to a dialect (not a full language) took manual work
  
• Streamlit UI was limited — had to tweak it using CSS
  
• Had to keep everything minimal due to time and resource limits

Accomplishments

• Fully working flow: upload → extract → translate → show
  
• Runs completely on Colab — no setup needed
  
• Solves a real local problem with a working MVP
  
• Made a clean UI inside Streamlit without using extra tools
• Sometimes the link (streamlit) may not work because i am using ngrok to connect both backend and the frontend so it may change.

What We Learned

• How OCR tools like Tesseract actually work
  
• Why keeping things local and simple makes a difference
  
• How to customize Streamlit beyond its defaults
  
• That even small, focused ideas can create real value

What’s Next

• Add voice output in Tunisian
  
• Improve translation using ML
  
• Support more content like legal forms and admin papers
  
• Make the UI mobile-friendly
  
• Package it as a PWA or native mobile app

Built With

• google-colab
• html/css-(custom-styling)
• markdown
• pil-(pillow)
• pyngrok
• python
• requests
• streamlit
• tesseract-ocr