Soundify - Arabic OCR & Text-to-Speech

Soundify is a Python desktop application that converts Arabic text from images into spoken audio. It offers two OCR recognition modes in a single unified interface:

1. Full Text (Tesseract): For recognizing sentences, paragraphs, or full pages of Arabic text
2. Character (CNN): For recognizing single Arabic characters using a custom-trained neural network

Features

• Dual OCR Modes: Switch between Tesseract (full text) and CNN (single character) recognition
• Image Input Options:
  • Live camera feed capture
  • File upload from local storage
• Text-to-Speech: Automatically converts recognized text to speech (MP3)
• Modern GUI: Dark/Light theme toggle with customtkinter
• Robust Error Handling: Gracefully handles missing dependencies

Project Structure

• soundify.py: Main application with unified Tesseract & CNN modes
• utils_tesseract.py: Tesseract OCR integration functions
• utils_cnn.py: CNN model prediction and character mapping
• model_training_notebook.ipynb: Jupyter notebook for CNN model training
• model/arabic-OCR.h5: Pre-trained CNN model (72MB)
• assets/: GUI icons and images
• requirements.txt: Python dependencies

Installation

Prerequisites

• Python 3.7+
• Tesseract OCR (Required for Full Text mode)
  • Windows: Download from UB-Mannheim/tesseract
  • Add Tesseract to your system PATH
  • Install Arabic language data during setup

Installation Steps

1. Clone the repository:

   git clone https://github.com/3bsalam-1/Arabic-OCR.git
   cd Arabic-OCR/Code

2. Install Python dependencies:

   pip install -r requirements.txt

   This will install:

   • customtkinter - Modern GUI framework
   • opencv-python - Camera and image processing
   • Pillow - Image manipulation
   • pygame - Audio playback
   • pytesseract - Tesseract OCR wrapper
   • gTTS - Google Text-to-Speech
   • tensorflow - CNN model (large package ~330MB)

3. Download the CNN Model (Required for Character mode):

   • The pre-trained model file arabic-OCR.h5 (72MB) is not included in the repository
   • Download it from: Release Page or contact the author
   • Place it in the model/ directory: Code/model/arabic-OCR.h5
   • If you skip this step, only Tesseract mode will work

Usage

Running the Application

python soundify.py

Mode Selection

Use the mode switcher at the bottom-left of the window:

• Full Text (Tesseract): Best for sentences, paragraphs, or documents

  • Requires: Tesseract OCR installed on your system
  • Use case: Reading full pages, signs, documents

• Character (CNN): Best for single Arabic characters

  • Requires: Tensorflow (installed via requirements.txt)
  • Use case: Individual letter recognition

Controls

Button	Function
Browse	Select an image file from your computer
Camera	Toggle live camera feed on/off
📷 Icon	Capture current camera frame
🔊 Icon	Play audio of recognized text
📝 Icon	Display recognized text on screen
Theme	Switch between Light/Dark mode

Workflow

1. Select your preferred mode (Tesseract or CNN)
2. Choose input method:
   • Click Browse to select an image file, OR
   • Click Camera to enable live feed, then click 📷 to capture
3. Click 🔊 to hear the text read aloud
4. Click 📝 to see the recognized text

Troubleshooting

"Tesseract is not installed or not found in PATH"

• Install Tesseract OCR from the link above
• Ensure it's added to your system PATH
• Restart your terminal/IDE after installation

"CNN Model or Tensorflow not found"

• Run: pip install tensorflow
• Note: This is a large package (~330MB) and may take time to download

Camera not working

• Ensure your camera is not being used by another application
• Check camera permissions in your system settings
• The app will show "Camera not accessible" if no camera is detected

No audio playback

• Ensure test.mp3 is being generated in the project directory
• Check your system audio settings
• Verify pygame is installed: pip install pygame

Dependencies

All dependencies are listed in requirements.txt:

customtkinter
opencv-python
Pillow
pygame
pytesseract
gTTS
tensorflow

Credits

Author: 3bsalam-1
Last Updated: December 2025

---