Update app.py

app.py

@@ -359,83 +359,79 @@ def generate_all_scenes(prompts_text):
     
 @spaces.GPU(duration=60)
 def add_text_to_scenes(gallery_images, prompts_text):
-    """Add text overlay to scene images"""
+    """Add text overlays to all scenes"""
+    print(f"Received gallery_images type: {type(gallery_images)}")
+    print(f"Number of images in gallery: {len(gallery_images) if isinstance(gallery_images, list) else 0}")
+    
     if not isinstance(gallery_images, list):
-        print("Error: gallery_images is not a list")
+        print("Gallery images must be a list")
         return [], []
-    
+
     clear_memory()
     
+    # Process text sections
     sections = prompts_text.split('='*50)
     overlaid_images = []
     output_files = []
     
+    # Create temporary directory for saving files
     temp_dir = "temp_book_pages"
     os.makedirs(temp_dir, exist_ok=True)
     
-    for i, (image_data, section) in enumerate(zip(gallery_images, sections)):
+    for i, (img_data, section) in enumerate(zip(gallery_images, sections)):
         if not section.strip():
             continue
             
-        lines = [line.strip() for line in section.split('\n') if line.strip()]
-        paragraph = None
-        for j, line in enumerate(lines):
-            if line.startswith('Paragraph'):
-                if j + 1 < len(lines):
-                    paragraph = lines[j + 1]
-                    break
-        
-        if paragraph and image_data is not None:
-            try:
-                print(f"Processing image {i+1}, type: {type(image_data)}")
-                print(f"Image data: {str(image_data)[:100]}...")  # Print first 100 chars
+        print(f"\nProcessing image {i+1}:")
+        print(f"Image data type: {type(img_data)}")
+        
+        try:
+            # Handle tuple from Gradio gallery
+            if isinstance(img_data, tuple):
+                filepath = img_data[0] if isinstance(img_data[0], str) else None
+                print(f"Found filepath: {filepath}")
                 
-                # Handle different types of image data
-                if isinstance(image_data, str):
-                    # If it's a string, try to load it as a numpy array
-                    try:
-                        import ast
-                        # Convert string representation of array to actual array
-                        array_data = ast.literal_eval(image_data)
-                        image = Image.fromarray(np.array(array_data, dtype=np.uint8))
-                    except:
-                        print(f"Failed to convert string to array for image {i+1}")
-                        continue
-                elif isinstance(image_data, tuple):
-                    # Handle gallery tuple format (image, label)
-                    image_data = image_data[0]
-                    if isinstance(image_data, np.ndarray):
-                        image = Image.fromarray(image_data)
-                    else:
-                        print(f"Unexpected tuple data type: {type(image_data)}")
-                        continue
-                elif isinstance(image_data, np.ndarray):
-                    image = Image.fromarray(image_data)
-                elif isinstance(image_data, Image.Image):
-                    image = image_data
+                if filepath and os.path.exists(filepath):
+                    print(f"Loading image from: {filepath}")
+                    image = Image.open(filepath).convert('RGB')
                 else:
-                    print(f"Unsupported image data type: {type(image_data)}")
+                    print(f"Invalid filepath: {filepath}")
                     continue
-                
-                # Ensure we have a valid RGB image
-                if image.mode != 'RGB':
-                    image = image.convert('RGB')
-                
+            else:
+                print(f"Unexpected image data type: {type(img_data)}")
+                continue
+            
+            # Extract paragraph text
+            lines = [line.strip() for line in section.split('\n') if line.strip()]
+            paragraph = None
+            for j, line in enumerate(lines):
+                if line.startswith('Paragraph'):
+                    if j + 1 < len(lines):
+                        paragraph = lines[j + 1]
+                        print(f"Found paragraph text for image {i+1}")
+                        break
+            
+            if paragraph and image:
+                # Add text overlay
                 overlaid_img = overlay_text_on_image(image, paragraph)
                 if overlaid_img is not None:
+                    # Convert to numpy array for gallery display
                     overlaid_array = np.array(overlaid_img)
                     overlaid_images.append(overlaid_array)
                     
+                    # Save file for download
                     output_path = os.path.join(temp_dir, f"panel_{i+1}.png")
                     overlaid_img.save(output_path)
                     output_files.append(output_path)
                     print(f"Successfully processed image {i+1}")
-                
-            except Exception as e:
-                print(f"Error processing image {i+1}: {str(e)}")
-                import traceback
-                print(traceback.format_exc())  # Print full error trace
-                continue
+                else:
+                    print(f"Failed to overlay text on image {i+1}")
+            
+        except Exception as e:
+            print(f"Error processing image {i+1}: {str(e)}")
+            import traceback
+            print(traceback.format_exc())
+            continue
     
     if not overlaid_images:
         print("No images were successfully processed")
@@ -446,7 +442,7 @@ def add_text_to_scenes(gallery_images, prompts_text):
     return overlaid_images, output_files
 
 def overlay_text_on_image(image, text):
-    """Helper function to overlay text on an image"""
+    """Add black text with white outline for better visibility."""
     if image is None:
         return None
     
@@ -460,6 +456,7 @@ def overlay_text_on_image(image, text):
         try:
             font = ImageFont.truetype("/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf", font_size)
         except:
+            print("Using default font as DejaVuSans-Bold.ttf not found")
             font = ImageFont.load_default()
 
         # Calculate text positioning