Update app.py

app.py

@@ -19,10 +19,11 @@ def remove_background(img: Image.Image) -> Image.Image:
     r, g, b, _ = rgba.split()
     return Image.merge("RGBA", (r, g, b, alpha))
 
-def blur_face_region(image: Image.Image) -> Image.Image:
+def blur_faces_smooth(image: Image.Image) -> tuple[Image.Image, list]:
     """
-    Uses OpenCV Haar Cascade to automatically detect all faces in the image
-    and blurs only those bounding boxes, keeping the rest of the image details intact.
+    Detects faces using Haar Cascades and applies a smooth, feathered Gaussian blur
+    only to the face interiors. This avoids any box/crop lines in Canny and provides
+    a premium faceless look.
     """
     rgba = image.convert("RGBA")
     width, height = rgba.size
@@ -34,8 +35,6 @@ def blur_face_region(image: Image.Image) -> Image.Image:
     
     # Load frontal face cascade
     face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')
-    
-    # Detect faces
     faces = face_cascade.detectMultiScale(
         gray, 
         scaleFactor=1.1, 
@@ -43,31 +42,44 @@ def blur_face_region(image: Image.Image) -> Image.Image:
         minSize=(25, 25)
     )
     
-    result = rgba.copy()
-    
-    if len(faces) > 0:
-        for (x, y, w, h) in faces:
-            # Expand box slightly to cover ears, eyebrows, forehead and chin
-            pad_x = int(w * 0.1)
-            pad_y = int(h * 0.15)
-            x1 = max(0, x - pad_x)
-            y1 = max(0, y - pad_y * 2) # extend upwards for forehead
-            x2 = min(width, x + w + pad_x)
-            y2 = min(height, y + h + pad_y)
-            
-            # Crop, blur, and paste back
-            region = result.crop((x1, y1, x2, y2))
-            blurred_region = region.filter(ImageFilter.GaussianBlur(radius=16))
-            result.paste(blurred_region, (x1, y1))
-    else:
-        # Fallback: soft general blur only on upper 30% center if no faces detected
-        top = max(0, int(height * 0.1))
-        bottom = min(height, int(height * 0.4))
-        region = result.crop((0, top, width, bottom))
-        blurred_region = region.filter(ImageFilter.GaussianBlur(radius=10))
-        result.paste(blurred_region, (0, top))
+    if len(faces) == 0:
+        # Fallback: soft feathered blur on upper center if no faces detected
+        mask = Image.new("L", (width, height), 0)
+        mask_np = np.array(mask)
+        # generic head ellipse at top center
+        cx, cy = width // 2, int(height * 0.25)
+        rx, ry = int(width * 0.25), int(height * 0.15)
+        cv2.ellipse(mask_np, (cx, cy), (rx, ry), 0, 0, 360, 255, -1)
+        mask = Image.fromarray(mask_np).filter(ImageFilter.GaussianBlur(radius=20))
+        
+        blurred_img = rgba.filter(ImageFilter.GaussianBlur(radius=14))
+        return Image.composite(blurred_img, rgba, mask), []
+
+    # Create a smooth, feathered mask (0 = original, 255 = blurred)
+    mask = Image.new("L", (width, height), 0)
+    mask_np = np.array(mask)
+    
+    for (x, y, w, h) in faces:
+        cx = x + w // 2
+        cy = y + h // 2
         
-    return result
+        # Erase eyes, nose, mouth but keep eyebrows & ears by shaping the ellipse
+        # Height radius (ry) covers center of face. Shift center down slightly.
+        rx = int(w * 0.36)
+        ry = int(h * 0.38)
+        cy_shifted = cy + int(h * 0.06) # shift down to preserve eyebrows
+        
+        cv2.ellipse(mask_np, (cx, cy_shifted), (rx, ry), 0, 0, 360, 255, -1)
+        
+    mask = Image.fromarray(mask_np)
+    # Heavy blur on mask to feather edges completely smoothly
+    mask = mask.filter(ImageFilter.GaussianBlur(radius=12))
+    
+    # Generate the blurred version of the face areas
+    blurred_img = rgba.filter(ImageFilter.GaussianBlur(radius=18))
+    
+    # Composite the blurred face areas onto the original image
+    return Image.composite(blurred_img, rgba, mask), list(faces)
 
 def extract_fine_details_cv2(img: Image.Image, detail_profile="balanced") -> Image.Image:
     """
@@ -87,9 +99,9 @@ def extract_fine_details_cv2(img: Image.Image, detail_profile="balanced") -> Ima
     if detail_profile == "soft":
         t1, t2 = 50, 150
     elif detail_profile == "balanced":
-        t1, t2 = 25, 90
+        t1, t2 = 22, 85
     else:  # raw / detailed
-        t1, t2 = 12, 60
+        t1, t2 = 12, 55
         
     # Run Canny edge detector
     edges = cv2.Canny(smoothed, t1, t2)
@@ -139,8 +151,8 @@ def process_faceless(input_image, prompt="", negative_prompt=""):
     # 1. Remove background
     subject = remove_background(input_image)
     
-    # 2. Blur face region to keep it faceless (no eyes, eyebrows, mouth)
-    preprocessed = blur_face_region(subject)
+    # 2. Smoothly blur face regions (removes eyes/mouth organically, keeps eyebrows/ears/jawline)
+    preprocessed, faces = blur_faces_smooth(subject)
     
     # 3. Extract line details using OpenCV Canny (highly superior thin lines)
     details_alpha = extract_fine_details_cv2(preprocessed, detail_profile="balanced")