Update app.py

app.py

@@ -1,64 +1,58 @@
 import gradio as gr
+import cloudinary
+import cloudinary.uploader
 import easyocr
+from PIL import Image
 import numpy as np
-from transformers import pipeline
-import urllib.parse
+from transformers import MarianMTModel, MarianTokenizer
 
-# Initialize OCR
-reader = easyocr.Reader(['en', 'hi', 'mr', 'ne'])
+# Cloudinary configuration
+cloudinary.config(
+    cloud_name="your_cloud_name",
+    api_key="your_api_key",
+    api_secret="your_api_secret"
+)
+
+# Initialize OCR reader
+reader = easyocr.Reader(['en'])
 
 # Initialize translation model
-translator = pipeline("translation", model="facebook/nllb-200-distilled-600M", src_lang="eng_Latn", tgt_lang="hin_Deva")
+model_name = 'Helsinki-NLP/opus-mt-en-hin'
+tokenizer = MarianTokenizer.from_pretrained(model_name)
+model = MarianMTModel.from_pretrained(model_name)
 
-# Cloudinary Info
-CLOUD_NAME = "deoux7285"  # No need for image name, this will be generated dynamically
-FONT = "Arial"
-SIZE = "40"
-COLOR = "black"
-POSITION = "south"
+def translate_text(text, target_lang='hin'):
+    translated = model.generate(**tokenizer.prepare_seq2seq_batch(text, return_tensors="pt"))
+    return tokenizer.decode(translated[0], skip_special_tokens=True)
 
-# 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
+def process_image(image, target_lang='hin'):
+    # OCR to extract text
+    result = reader.readtext(np.array(image))
+    extracted_text = " ".join([item[1] for item in result])
 
-# 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
+    # Translate extracted text
+    translated_text = translate_text(extracted_text, target_lang)
 
-    if not results:
-        return None, "No text detected"
+    # Upload image to Cloudinary
+    buffered = io.BytesIO()
+    image.save(buffered, format="JPEG")
+    buffered.seek(0)
+    upload_result = cloudinary.uploader.upload(buffered)
+    cloud_url = upload_result['secure_url']
 
-    full_text = ""
-    for _, text, _ in results:
-        if text.strip():  # Process non-empty text
-            try:
-                translation = translator(text, max_length=200)[0]["translation_text"]
-            except:
-                translation = "[Translation Error]"
-            full_text += translation + " "
+    # Apply text overlay using Cloudinary
+    overlay_url = f"{cloud_url}?overlay_text={translated_text}&font_size=30&font_color=white&gravity=center"
 
-    # Generate Cloudinary URL for translated text overlay
-    cloud_url = generate_cloudinary_url(full_text.strip())
-    return cloud_url, full_text.strip()
+    return overlay_url, translated_text
 
-# Inside Gradio interface
+# Gradio interface
 iface = gr.Interface(
     fn=process_image,
-    inputs=[
-        gr.Image(type="pil", label="Upload Image"),
-        gr.Textbox(label="Target Language (e.g., hin_Deva, mar_Deva, nep_Deva)")
-    ],
-    outputs=[
-        gr.HTML(label="Image with Cloudinary Text"),  # ✅ changed from gr.Image to gr.HTML
-        gr.Textbox(label="Translated Text Output")
-    ],
-    title="Advanced Image Translator",
-    description="Extract & translate text from an image (OCR + translation overlay)"
+    inputs=[gr.Image(type="pil", label="Upload Image"), gr.Textbox(label="Target Language (e.g., hin)"), gr.Slider(minimum=10, maximum=100, label="Font Size")],
+    outputs=[gr.Image(label="Image with Overlay"), gr.Textbox(label="Translated Text")],
+    title="Image Translator with Overlay",
+    description="Upload an image, extract text using OCR, translate it, and overlay the translated text on the image."
 )
 
-
 if __name__ == "__main__":
-    iface.launch(share=True)  # Launch Gradio interface
+    iface.launch(share=True)