Update app.py

app.py

@@ -21,6 +21,12 @@ from diffusers import FluxPipeline
 from transformers import pipeline, AutoTokenizer, AutoModelForSeq2SeqLM
 import gc
 
+from PIL import Image, ImageDraw, ImageFont
+from PIL import Image
+from gradio_client import Client, handle_file
+import uuid
+
+
 def clear_memory():
     """메모리 정리 함수"""
     gc.collect()
@@ -113,6 +119,8 @@ try:
 except Exception as e:
     print(f"Warning: Could not move pipeline to CUDA: {str(e)}")
 
+client = Client("NabeelShar/BiRefNet_for_text_writing")
+
 class timer:
     def __init__(self, method_name="timed process"):
         self.method = method_name
@@ -482,9 +490,117 @@ button.primary:hover {
 }
 """
 
-# UI 구성
-# UI 구성 부분에서 process_btn을 위로 이동하고 position_grid.click 부분 제거
 
+
+def add_text_with_stroke(draw, text, x, y, font, text_color, stroke_width):
+    """Helper function to draw text with stroke"""
+    # Draw the stroke/outline
+    for adj_x in range(-stroke_width, stroke_width + 1):
+        for adj_y in range(-stroke_width, stroke_width + 1):
+            draw.text((x + adj_x, y + adj_y), text, font=font, fill=text_color)
+
+def remove_background(image):
+    # Save the image to a specific location
+    filename = f"image_{uuid.uuid4()}.png"  # Generates a universally unique identifier (UUID) for the filename
+    image.save(filename)
+    # Call gradio client for background removal
+    result = client.predict(images=handle_file(filename), api_name="/image")
+    return Image.open(result[0])
+
+def superimpose(image_with_text, overlay_image):
+    # Open image as RGBA to handle transparency
+    overlay_image = overlay_image.convert("RGBA")
+    # Paste overlay on the background
+    image_with_text.paste(overlay_image, (0, 0), overlay_image)
+    # Save the final image
+    # image_with_text.save("output_image.png")
+    return image_with_text
+
+def add_text_to_image(
+    input_image,
+    text,
+    font_size,
+    color,
+    opacity,
+    x_position,
+    y_position,
+    thickness
+):
+    """
+    Add text to an image with customizable properties
+    """
+    # Convert gradio image (numpy array) to PIL Image
+    if input_image is None:
+        return None
+    
+    image = Image.fromarray(input_image)
+    # remove background 
+    overlay_image = remove_background(image)
+
+    # Create a transparent overlay for the text
+    txt_overlay = Image.new('RGBA', image.size, (255, 255, 255, 0))
+    draw = ImageDraw.Draw(txt_overlay)
+    
+    # Create a font with specified size
+    try:
+        font = ImageFont.truetype("DejaVuSans.ttf", int(font_size))
+    except:
+        # If DejaVu font is not found, try to use Arial or default
+        try:
+            font = ImageFont.truetype("arial.ttf", int(font_size))
+        except:
+            print("Using default font as system fonts not found")
+            font = ImageFont.load_default()
+    
+    # Convert color name to RGB
+    color_map = {
+        'White': (255, 255, 255),
+        'Black': (0, 0, 0),
+        'Red': (255, 0, 0),
+        'Green': (0, 255, 0),
+        'Blue': (0, 0, 255),
+        'Yellow': (255, 255, 0),
+        'Purple': (128, 0, 128)
+    }
+    rgb_color = color_map.get(color, (255, 255, 255))
+    
+    # Get text size for positioning
+    text_bbox = draw.textbbox((0, 0), text, font=font)
+    text_width = text_bbox[2] - text_bbox[0]
+    text_height = text_bbox[3] - text_bbox[1]
+
+    # Calculate actual x and y positions based on percentages
+    actual_x = int((image.width - text_width) * (x_position / 100))
+    actual_y = int((image.height - text_height) * (y_position / 100))
+
+    # Create final color with opacity
+    text_color = (*rgb_color, int(opacity))
+    
+    # Draw the text with stroke for thickness
+    add_text_with_stroke(
+        draw, 
+        text, 
+        actual_x, 
+        actual_y, 
+        font, 
+        text_color, 
+        int(thickness)
+    )
+
+    # Combine the original image with the text overlay
+    if image.mode != 'RGBA':
+        image = image.convert('RGBA')
+    output_image = Image.alpha_composite(image, txt_overlay)
+    
+    # Convert back to RGB for display
+    output_image = output_image.convert('RGB')
+
+    # superimpose images
+    output_image = superimpose(output_image, overlay_image)
+    
+    # Convert PIL image back to numpy array for Gradio
+    return np.array(output_image)
+    
 # UI 구성
 with gr.Blocks(theme=gr.themes.Soft(), css=css) as demo:
     gr.HTML("""