Update app.py

app.py

@@ -1,59 +1,63 @@
 import gradio as gr
 import easyocr
-from deep_translator import GoogleTranslator
-from PIL import Image, ImageDraw, ImageFont
 import numpy as np
+from transformers import pipeline
+import urllib.parse
 
-# Initialize reader with allowed lang_list only
+# Initialize OCR
 reader = easyocr.Reader(['en', 'hi', 'mr', 'ne'])
 
-# Font path — works on HuggingFace
-FONT_PATH = "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf"
+# Initialize translation model
+translator = pipeline("translation", model="facebook/nllb-200-distilled-600M", src_lang="eng_Latn", tgt_lang="hin_Deva")
 
-def process_image(image, target_lang):
-    img_array = np.array(image)
+# Cloudinary Info
+CLOUD_NAME = "deoux7285"  # No need for image name, this will be generated dynamically
+FONT = "Arial"
+SIZE = "40"
+COLOR = "black"
+POSITION = "south"
 
-    # Run OCR safely
-    results = reader.readtext(img_array)
+# Function to generate Cloudinary overlay URL
+def generate_cloudinary_url(translated_text):
+    encoded = urllib.parse.quote(translated_text)  # Ensure text is URL encoded
+    url = f"https://res.cloudinary.com/{deoux7285}/image/upload/l_text:{FONT}_{SIZE}:{encoded},co_{COLOR},g_{POSITION}/sample.jpg"  # Image URL dynamically handled
+    return url
 
-    if not results:
-        return image, "No text detected"
+# Function to process image: OCR + Translation + Overlay
+def process_image(image, target_lang):
+    img_array = np.array(image)  # Convert PIL image to numpy array for OCR
+    results = reader.readtext(img_array)  # Run OCR on the image
 
-    final_translations = []
-    img_pil = image.convert("RGB")
-    draw = ImageDraw.Draw(img_pil)
-    font = ImageFont.truetype(FONT_PATH, size=20)
+    if not results:
+        return None, "No text detected"
 
-    for (bbox, text, prob) in results:
-        if text.strip():
+    full_text = ""
+    for _, text, _ in results:
+        if text.strip():  # Process non-empty text
             try:
-                translated = GoogleTranslator(source='auto', target=target_lang).translate(text)
+                translation = translator(text, max_length=200)[0]["translation_text"]
             except:
-                translated = "[Translation Error]"
-
-            final_translations.append(translated)
-
-            # Draw translation at top-left of bounding box
-            top_left = bbox[0]
-            draw.text((top_left[0], top_left[1] - 20), translated, fill=(255, 0, 0), font=font)
+                translation = "[Translation Error]"
+            full_text += translation + " "
 
-    full_text = "\n".join(final_translations)
-    return img_pil, full_text
+    # Generate Cloudinary URL for translated text overlay
+    cloud_url = generate_cloudinary_url(full_text.strip())
+    return cloud_url, full_text.strip()
 
-# Gradio UI
+# Gradio UI setup
 iface = gr.Interface(
     fn=process_image,
     inputs=[
-        gr.Image(type="pil", label="Upload Image"),
-        gr.Textbox(label="Translate To (hi / en / mr / ne)")
+        gr.Image(type="pil", label="Upload Image"),  # Upload image input
+        gr.Textbox(label="Target Language (e.g., hin_Deva, mar_Deva, nep_Deva)")  # Textbox for target language
     ],
     outputs=[
-        gr.Image(type="pil", label="Image with Translated Overlay"),
-        gr.Textbox(label="Translated Text Output")
+        gr.Image(type="url", label="Image with Cloudinary Text"),  # Image with overlay from Cloudinary
+        gr.Textbox(label="Translated Text Output")  # Output of translated text
     ],
-    title="Image Translator with Overlay",
+    title="Advanced Image Translator",
     description="Extract & translate text from an image (OCR + translation overlay)"
 )
 
 if __name__ == "__main__":
-    iface.launch()
+    iface.launch()  # Launch Gradio interface