app.py

import gradio as gr
import easyocr
from PIL import Image
import numpy as np
import google.generativeai as genai
import time  # For retry delays

# Gemini API key - Hardcoded internal variable (replace with your actual key!)
api_key = "AIzaSyAKI92YawOKQ1-HRLmvaryMEWk_y4alJgA"

# Global reader - initialize once
reader = None

def initialize_reader():
    """Initialize EasyOCR reader"""
    global reader
    if reader is None:
        print("Loading EasyOCR model...")
        reader = easyocr.Reader(['en'], gpu=False, verbose=False)
        print("EasyOCR model loaded successfully!")
    return reader

def extract_text_from_quote(image):
    """Extract text from quote image using EasyOCR"""
    if image is None:
        return "Please upload an image first.", "Words: 0"
    
    try:
        # Initialize reader
        reader = initialize_reader()
        
        # Convert PIL to numpy array
        img_array = np.array(image)
        
        # Extract text with EasyOCR
        results = reader.readtext(img_array, paragraph=True)
        
        # Extract just the text from results
        if results:
            # EasyOCR returns list of (bbox, text, confidence)
            text_parts = []
            for result in results:
                if len(result) >= 2 and result[1].strip():  # Check if text exists
                    text_parts.append(result[1].strip())
            
            if text_parts:
                extracted_text = ' '.join(text_parts)
                word_count = f"Words: {len(extracted_text.split())}"
                return extracted_text, word_count
        
        return "No text detected in the image.", "Words: 0"
            
    except Exception as e:
        return f"Error processing image: {str(e)}", "Words: 0"

def translate_extracted(text, lang):
    """Translate the extracted English text using Gemini API with retry"""
    if not text or "No text" in text or "Error" in text or "Please upload" in text:
        return "No valid text to translate."
    
    try:
        print(f"DEBUG: API Key loaded (first 5 chars: {api_key[:5]}...). Starting translation to {lang}")
        genai.configure(api_key=api_key)
        
        # Retry logic for HF network issues (up to 3 attempts)
        for attempt in range(3):
            try:
                model = genai.GenerativeModel('gemini-1.5-flash')  # Primary model
                prompt = f"Translate the following English text accurately to persian,You are a cool, chill translator with a fun and warm personality, inspired by Persian Twitter (توویتر فارسی) style. Your translations should be natural, slangy, touching, and relatable—like casual chats with a friend: short, heartfelt, with colloquial Persian words (e.g. for cultural fit, “فک می‌کنی” for “think,” contractions like “اونی‌ام”). No emojis, keep it RTL-friendly for Persian text. Be concise but preserve the emotional depth and humor.Translate the following English fact/quote into Persian in this style , Maintain correct grammar, even when using slang. Avoid direct, literal translations that sound unnatural in Persian.. Make it feel like a real tweet: warm, fun, and human.English fact: [{text}]Output only the Persian whit Format for image overlay: Break the Persian text into short, visually appealing lines.donot use any emojies at all Ensure the output is RTL-friendly and the text should be out any self represent or emojies just translated text "
                response = model.generate_content(prompt)
                translated = response.text.strip()
                print(f"DEBUG: Translation successful on attempt {attempt+1}: {translated[:50]}...")
                return translated
            except Exception as inner_e:
                print(f"DEBUG: Attempt {attempt+1} failed: {str(inner_e)}. Retrying...")
                if 'flash' in str(inner_e):  # Fallback to pro if flash unavailable
                    model = genai.GenerativeModel('gemini-pro')
                time.sleep(2)  # Delay between retries
        
        raise Exception("All retries failed.")
    
    except Exception as e:
        error_msg = f"Error translating: {str(e)}. Check HF logs for details. (If on HF, ensure network allows Google API access.)"
        print(f"DEBUG: Translation failed: {str(e)}")
        return error_msg

def clear_all():
    """Clear all inputs and outputs"""
    return None, "Your extracted quote will appear here...", "Words: 0", "Translation will appear here...", False  # Include auto-translate checkbox

# Create Gradio interface
with gr.Blocks(title="Quote OCR Extractor", theme=gr.themes.Soft()) as demo:
    
    gr.Markdown("# 📝 Quote Text Extractor")
    gr.Markdown("Upload a quote image and extract the text instantly!")
    
    with gr.Row():
        with gr.Column(scale=1):
            image_input = gr.Image(
                label="Upload Quote Image",
                type="pil",
                sources=["upload", "clipboard"]
            )
            
            with gr.Row():
                clear_btn = gr.Button("Clear", variant="secondary")
                extract_btn = gr.Button("Extract Text", variant="primary")
        
        with gr.Column(scale=1):
            text_output = gr.Textbox(
                label="Extracted Quote Text (English)",
                placeholder="Your extracted quote will appear here...",
                lines=8,
                max_lines=15,
                show_copy_button=True
            )
            
            word_count = gr.Textbox(
                label="Word Count",
                interactive=False,
                max_lines=1
            )
            
            target_lang = gr.Dropdown(
                label="Select Target Language for Translation",
                choices=["Spanish", "French", "German", "Chinese", "Arabic", "persian(farsi)"],
                value="persian(farsi)"  # Default
            )
            
            auto_translate = gr.Checkbox(
                label="Auto-Translate After Extraction",
                value=True
            )
            
            translate_btn = gr.Button("Translate Extracted Text", variant="primary")
            
            translated_output = gr.Textbox(
                label="Translated Quote",
                placeholder="Translation will appear here...",
                lines=8,
                max_lines=15,
                show_copy_button=True
            )
    
    # Event handlers
    def extract_and_translate(image, lang, auto):
        text, wc = extract_text_from_quote(image)
        trans = "Translation will appear here..."  # Default
        if auto and "No text" not in text and "Error" not in text and "Please upload" not in text:
            trans = translate_extracted(text, lang)
        return text, wc, trans
    
    extract_btn.click(
        fn=extract_and_translate,
        inputs=[image_input, target_lang, auto_translate],
        outputs=[text_output, word_count, translated_output]
    )
    
    image_input.change(
        fn=extract_and_translate,
        inputs=[image_input, target_lang, auto_translate],
        outputs=[text_output, word_count, translated_output]
    )
    
    translate_btn.click(
        fn=translate_extracted,
        inputs=[text_output, target_lang],
        outputs=translated_output
    )
    
    clear_btn.click(
        fn=clear_all,
        outputs=[image_input, text_output, word_count, translated_output, auto_translate]
    )
    
    gr.Markdown("""
    ### 💡 Tips for Best Results:
    - Upload clear images with good contrast
    - Works best with simple backgrounds
    - Supports PNG, JPG, and other common formats
    - Text will be automatically extracted and formatted
    - Translation assumes extracted text is English; select a target language and click Translate
    - If translation fails on HF, check your API key and console logs!
    """)

if __name__ == "__main__":
    demo.launch()