Update app.py

app.py

@@ -1,848 +1,820 @@
-import spaces
-import logging
-from datetime import datetime
-from pathlib import Path
+"""
+ANSIM BLUR - Face Privacy Protection
+=====================================
+Advanced AI-Powered Face Detection & Privacy Protection
+Using YOLOv8 for face detection with Gaussian/Mosaic blur options
+"""
+
 import gradio as gr
-import torch
-import torchaudio
-import os
-import requests
-from transformers import pipeline
-import tempfile
-import numpy as np
-from einops import rearrange
 import cv2
-from scipy.io import wavfile
-import librosa
-import json
+import numpy as np
+import tempfile
+import os
+from pathlib import Path
 from typing import Optional, Tuple, List
-import atexit
-
-# 환경 변수 설정으로 torch.load 체크 우회 (임시 해결책)
-os.environ["TRANSFORMERS_ALLOW_UNSAFE_DESERIALIZATION"] = "1"
-
-try:
-    import mmaudio
-except ImportError:
-    os.system("pip install -e .")
-    import mmaudio
-
-from mmaudio.eval_utils import (ModelConfig, all_model_cfg, generate, load_video, make_video,
-                                setup_eval_logging)
-from mmaudio.model.flow_matching import FlowMatching
-from mmaudio.model.networks import MMAudio, get_my_mmaudio
-from mmaudio.model.sequence_config import SequenceConfig
-from mmaudio.model.utils.features_utils import FeaturesUtils
-
-# 로깅 설정
-logging.basicConfig(
-    level=logging.INFO,
-    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
-)
-log = logging.getLogger()
-
-# CUDA 설정
-if torch.cuda.is_available():
-    device = torch.device("cuda")
-    torch.backends.cuda.matmul.allow_tf32 = True
-    torch.backends.cudnn.allow_tf32 = True
-    torch.backends.cudnn.benchmark = True
-else:
-    device = torch.device("cpu")
-
-dtype = torch.bfloat16
-
-# 모델 설정
-model: ModelConfig = all_model_cfg['large_44k_v2']
-model.download_if_needed()
-output_dir = Path('./output/gradio')
-
-setup_eval_logging()
-
-# 번역기 설정
-try:
-    translator = pipeline("translation", 
-                         model="Helsinki-NLP/opus-mt-ko-en", 
-                         device="cpu",
-                         use_fast=True,
-                         trust_remote_code=False)
-except Exception as e:
-    log.warning(f"Failed to load translation model with safetensors: {e}")
-    try:
-        translator = pipeline("translation", 
-                             model="Helsinki-NLP/opus-mt-ko-en", 
-                             device="cpu")
-    except Exception as e2:
-        log.error(f"Failed to load translation model: {e2}")
-        translator = None
-
-PIXABAY_API_KEY = "33492762-a28a596ec4f286f84cd328b17"
-
-def cleanup_temp_files():
-    temp_dir = tempfile.gettempdir()
-    for file in os.listdir(temp_dir):
-        if file.endswith(('.mp4', '.flac')):
-            try:
-                os.remove(os.path.join(temp_dir, file))
-            except:
-                pass
-
-atexit.register(cleanup_temp_files)
-
-def get_model() -> tuple[MMAudio, FeaturesUtils, SequenceConfig]:
-    with torch.cuda.device(device):
-        seq_cfg = model.seq_cfg
-        net: MMAudio = get_my_mmaudio(model.model_name).to(device, dtype).eval()
-        net.load_weights(torch.load(model.model_path, map_location=device, weights_only=True))
-        log.info(f'Loaded weights from {model.model_path}')
-
-        feature_utils = FeaturesUtils(
-            tod_vae_ckpt=model.vae_path,
-            synchformer_ckpt=model.synchformer_ckpt,
-            enable_conditions=True,
-            mode=model.mode,
-            bigvgan_vocoder_ckpt=model.bigvgan_16k_path,
-            need_vae_encoder=False
-        ).to(device, dtype).eval()
-
-        return net, feature_utils, seq_cfg
-
-net, feature_utils, seq_cfg = get_model()
-
-def translate_prompt(text):
-    try:
-        if translator is None:
-            return text
-            
-        if text and any(ord(char) >= 0x3131 and ord(char) <= 0xD7A3 for char in text):
-            with torch.no_grad():
-                translation = translator(text)[0]['translation_text']
-            return translation
-        return text
-    except Exception as e:
-        logging.error(f"Translation error: {e}")
-        return text
+import torch
+from PIL import Image
+
+# ============================================
+# Constants & Configuration
+# ============================================
+BLUR_MODES = ["Gaussian Blur", "Mosaic Effect"]
+
+DEFAULT_CONFIG = {
+    "confidence": 0.25,
+    "iou": 0.45,
+    "expand_ratio": 0.05,
+    "blur_intensity": 51,
+    "mosaic_size": 15,
+}
+
+SLIDER_CONFIG = {
+    "confidence": {"minimum": 0.05, "maximum": 0.9, "step": 0.01},
+    "iou": {"minimum": 0.1, "maximum": 0.9, "step": 0.01},
+    "expand": {"minimum": 0.0, "maximum": 0.5, "step": 0.01},
+    "blur": {"minimum": 15, "maximum": 151, "step": 2},
+    "mosaic": {"minimum": 5, "maximum": 40, "step": 1},
+}
 
-@torch.no_grad()
-def search_videos(query):
-    try:
-        query = translate_prompt(query)
-        return search_pixabay_videos(query, PIXABAY_API_KEY)
-    except Exception as e:
-        logging.error(f"Video search error: {e}")
-        return []
 
-def search_pixabay_videos(query, api_key):
+# ============================================
+# Model Manager
+# ============================================
+class FaceDetector:
+    """YOLOv8 Face Detection Model Manager"""
+    
+    def __init__(self, model_path: str = "yolov8-face-hf.pt"):
+        self.model = None
+        self.device = self._get_device()
+        self._load_model(model_path)
+    
+    def _get_device(self) -> str:
+        """Determine the best available device"""
+        if torch.cuda.is_available():
+            return "cuda"
+        elif torch.backends.mps.is_available():
+            return "mps"
+        return "cpu"
+    
+    def _load_model(self, model_path: str) -> None:
+        """Load YOLO model"""
+        from ultralytics import YOLO
+        self.model = YOLO(model_path)
+        self.model.to(self.device)
+    
+    def detect(self, image: np.ndarray, conf: float, iou: float) -> List:
+        """Run face detection on image"""
+        with torch.no_grad():
+            results = self.model.predict(
+                image, 
+                conf=conf, 
+                iou=iou, 
+                verbose=False, 
+                device=self.device
+            )
+        return results
+
+
+# Initialize global detector
+detector = FaceDetector()
+
+
+# ============================================
+# Image Processing Functions
+# ============================================
+def ensure_odd(x: int) -> int:
+    """Ensure kernel size is odd for OpenCV"""
+    return x if x % 2 == 1 else x + 1
+
+
+def get_even_dimensions(w: int, h: int) -> Tuple[int, int]:
+    """Ensure video dimensions are even for codec compatibility"""
+    return (w if w % 2 == 0 else w - 1, h if h % 2 == 0 else h - 1)
+
+
+def apply_blur(
+    face_roi: np.ndarray, 
+    mode: str, 
+    blur_kernel: int, 
+    mosaic_size: int = 15
+) -> np.ndarray:
+    """Apply blur or mosaic effect to face region"""
+    if face_roi.size == 0:
+        return face_roi
+    
+    if mode == "Gaussian Blur":
+        k = ensure_odd(max(blur_kernel, 15))
+        return cv2.GaussianBlur(face_roi, (k, k), 0)
+    else:  # Mosaic Effect
+        m = max(2, mosaic_size)
+        h, w = face_roi.shape[:2]
+        small = cv2.resize(
+            face_roi, 
+            (max(1, w // m), max(1, h // m)), 
+            interpolation=cv2.INTER_LINEAR
+        )
+        return cv2.resize(small, (w, h), interpolation=cv2.INTER_NEAREST)
+
+
+def expand_bbox(
+    x1: int, y1: int, x2: int, y2: int,
+    expand_ratio: float,
+    img_w: int, img_h: int
+) -> Tuple[int, int, int, int]:
+    """Expand bounding box by ratio and clip to image bounds"""
+    if expand_ratio > 0:
+        bw, bh = x2 - x1, y2 - y1
+        dx, dy = int(bw * expand_ratio), int(bh * expand_ratio)
+        x1, y1 = x1 - dx, y1 - dy
+        x2, y2 = x2 + dx, y2 + dy
+    
+    # Clip to image bounds
+    x1 = max(0, min(img_w, x1))
+    x2 = max(0, min(img_w, x2))
+    y1 = max(0, min(img_h, y1))
+    y2 = max(0, min(img_h, y2))
+    
+    return x1, y1, x2, y2
+
+
+def blur_faces_in_image(
+    image_bgr: np.ndarray,
+    conf: float,
+    iou: float,
+    expand_ratio: float,
+    mode: str,
+    blur_kernel: int,
+    mosaic_size: int
+) -> Tuple[np.ndarray, int]:
+    """Detect and blur faces in a single image"""
+    h, w = image_bgr.shape[:2]
+    face_count = 0
+    
+    results = detector.detect(image_bgr, conf, iou)
+    
+    for r in results:
+        boxes = r.boxes.xyxy.cpu().numpy() if hasattr(r.boxes, "xyxy") else []
+        face_count = len(boxes)
+        
+        for box in boxes:
+            x1, y1, x2, y2 = map(int, box[:4])
+            x1, y1, x2, y2 = expand_bbox(x1, y1, x2, y2, expand_ratio, w, h)
+            
+            if x2 <= x1 or y2 <= y1:
+                continue
+            
+            roi = image_bgr[y1:y2, x1:x2]
+            image_bgr[y1:y2, x1:x2] = apply_blur(roi, mode, blur_kernel, mosaic_size)
+    
+    return image_bgr, face_count
+
+
+def blur_faces_in_video(
+    input_path: str,
+    conf: float,
+    iou: float,
+    expand_ratio: float,
+    mode: str,
+    blur_kernel: int,
+    mosaic_size: int,
+    progress: gr.Progress
+) -> Tuple[str, int, int]:
+    """Process video file and blur all detected faces"""
+    from moviepy.editor import VideoFileClip
+    
+    cap = cv2.VideoCapture(input_path)
+    if not cap.isOpened():
+        raise IOError("Cannot open video file")
+    
+    # Get video properties
+    in_w = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
+    in_h = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
+    fps = cap.get(cv2.CAP_PROP_FPS) or 25.0
+    total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT)) or 0
+    
+    out_w, out_h = get_even_dimensions(in_w, in_h)
+    
+    # Create temp files
+    temp_video = tempfile.NamedTemporaryFile(delete=False, suffix="_temp.mp4")
+    output_video = tempfile.NamedTemporaryFile(delete=False, suffix="_blurred.mp4")
+    temp_path = temp_video.name
+    output_path = output_video.name
+    temp_video.close()
+    output_video.close()
+    
+    # Setup video writer
+    fourcc = cv2.VideoWriter_fourcc(*"mp4v")
+    writer = cv2.VideoWriter(temp_path, fourcc, fps, (out_w, out_h))
+    
+    frame_idx = 0
+    total_faces = 0
+    
     try:
-        base_url = "https://pixabay.com/api/videos/"
-        params = {
-            "key": api_key,
-            "q": query,
-            "per_page": 40
-        }
+        while True:
+            ret, frame = cap.read()
+            if not ret:
+                break
+            
+            frame = cv2.resize(frame, (out_w, out_h))
+            h, w = frame.shape[:2]
+            
+            # Detect faces
+            results = detector.detect(frame, conf, iou)
+            
+            if results:
+                r = results[0]
+                boxes = r.boxes.xyxy.cpu().numpy() if hasattr(r.boxes, "xyxy") else []
+                total_faces += len(boxes)
+                
+                for box in boxes:
+                    x1, y1, x2, y2 = map(int, box[:4])
+                    x1, y1, x2, y2 = expand_bbox(x1, y1, x2, y2, expand_ratio, w, h)
+                    
+                    if x2 <= x1 or y2 <= y1:
+                        continue
+                    
+                    roi = frame[y1:y2, x1:x2]
+                    frame[y1:y2, x1:x2] = apply_blur(roi, mode, blur_kernel, mosaic_size)
+            
+            writer.write(frame)
+            frame_idx += 1
+            
+            if total_frames > 0:
+                progress(frame_idx / total_frames, desc=f"Processing frame {frame_idx}/{total_frames}")
+    
+    finally:
+        cap.release()
+        writer.release()
+    
+    # Merge audio from original video
+    try:
+        progress(0.95, desc="Merging audio...")
+        original = VideoFileClip(input_path)
+        processed = VideoFileClip(temp_path).set_audio(original.audio)
+        processed.write_videofile(
+            output_path,
+            codec="libx264",
+            audio_codec="aac",
+            threads=1,
+            logger=None
+        )
+        original.close()
+        processed.close()
+        
+        # Clean up temp file
+        if os.path.exists(temp_path):
+            os.remove(temp_path)
         
-        response = requests.get(base_url, params=params)
-        if response.status_code == 200:
-            data = response.json()
-            return [video['videos']['large']['url'] for video in data.get('hits', [])]
-        return []
+        return output_path, total_faces, total_frames
+    
     except Exception as e:
-        logging.error(f"Pixabay API error: {e}")
-        return []
-
-@spaces.GPU
-@torch.inference_mode()
-def video_to_audio(video: gr.Video, prompt: str, negative_prompt: str, seed: int, num_steps: int,
-                   cfg_strength: float, duration: float):
-    prompt = translate_prompt(prompt)
-    negative_prompt = translate_prompt(negative_prompt)
-
-    rng = torch.Generator(device=device)
-    rng.manual_seed(seed)
-    fm = FlowMatching(min_sigma=0, inference_mode='euler', num_steps=num_steps)
-
-    clip_frames, sync_frames, duration = load_video(video, duration)
-    clip_frames = clip_frames.unsqueeze(0)
-    sync_frames = sync_frames.unsqueeze(0)
-    seq_cfg.duration = duration
-    net.update_seq_lengths(seq_cfg.latent_seq_len, seq_cfg.clip_seq_len, seq_cfg.sync_seq_len)
-
-    audios = generate(clip_frames,
-                      sync_frames, [prompt],
-                      negative_text=[negative_prompt],
-                      feature_utils=feature_utils,
-                      net=net,
-                      fm=fm,
-                      rng=rng,
-                      cfg_strength=cfg_strength)
-    audio = audios.float().cpu()[0]
-
-    video_save_path = tempfile.NamedTemporaryFile(delete=False, suffix='.mp4').name
-    make_video(video,
-               video_save_path,
-               audio,
-               sampling_rate=seq_cfg.sampling_rate,
-               duration_sec=seq_cfg.duration)
-    
-    # 정보 로그 생성
-    info_log = f"""✅ VIDEO TO AUDIO COMPLETE!
+        print(f"Audio merging failed: {e}")
+        return temp_path, total_faces, total_frames
+
+
+# ============================================
+# Gradio Processing Handlers
+# ============================================
+def process_image(
+    image: Optional[Image.Image],
+    conf: float,
+    iou: float,
+    expand_ratio: float,
+    mode_choice: str,
+    blur_intensity: int,
+    mosaic_size: int
+) -> Tuple[Optional[Image.Image], str]:
+    """Main image processing handler"""
+    if image is None:
+        return None, "⚠️ Please upload an image first!"
+    
+    # Convert PIL to BGR
+    image_bgr = cv2.cvtColor(np.array(image), cv2.COLOR_RGB2BGR)
+    h, w = image_bgr.shape[:2]
+    
+    # Set blur parameters based on mode
+    blur_kernel = blur_intensity if mode_choice == "Gaussian Blur" else 51
+    mosaic = mosaic_size if mode_choice == "Mosaic Effect" else 15
+    
+    # Process image
+    result_bgr, face_count = blur_faces_in_image(
+        image_bgr.copy(), conf, iou, expand_ratio,
+        mode_choice, blur_kernel, mosaic
+    )
+    
+    # Convert back to RGB
+    result_rgb = cv2.cvtColor(result_bgr, cv2.COLOR_BGR2RGB)
+    result_pil = Image.fromarray(result_rgb)
+    
+    # Generate log
+    intensity_value = blur_intensity if mode_choice == "Gaussian Blur" else mosaic_size
+    info_log = f"""✅ IMAGE PROCESSING COMPLETE!
 {'=' * 50}
-🎬 Video Info:
-   • Duration: {duration:.2f} seconds
+🖼️ Image Info:
+   • Size: {w} x {h} pixels
+   • Format: RGB
 {'=' * 50}
-⚙️ Generation Settings:
-   • Seed: {seed}
-   • Steps: {num_steps}
-   • CFG Strength: {cfg_strength}
+🔍 Detection Settings:
+   • Confidence: {conf}
+   • IoU Threshold: {iou}
+   • Box Expansion: {expand_ratio}
 {'=' * 50}
-📝 Prompts:
-   • Prompt: {prompt[:40]}{'...' if len(prompt) > 40 else ''}
-   • Negative: {negative_prompt[:30]}{'...' if len(negative_prompt) > 30 else ''}
+🎨 Blur Settings:
+   • Style: {mode_choice}
+   • Intensity: {intensity_value}
 {'=' * 50}
-💾 Video with audio ready!"""
-    
-    return video_save_path, info_log
-
-@spaces.GPU
-@torch.inference_mode()
-def text_to_audio(prompt: str, negative_prompt: str, seed: int, num_steps: int, cfg_strength: float,
-                  duration: float):
-    prompt = translate_prompt(prompt)
-    negative_prompt = translate_prompt(negative_prompt)
-
-    rng = torch.Generator(device=device)
-    rng.manual_seed(seed)
-    fm = FlowMatching(min_sigma=0, inference_mode='euler', num_steps=num_steps)
-
-    clip_frames = sync_frames = None
-    seq_cfg.duration = duration
-    net.update_seq_lengths(seq_cfg.latent_seq_len, seq_cfg.clip_seq_len, seq_cfg.sync_seq_len)
-
-    audios = generate(clip_frames,
-                      sync_frames, [prompt],
-                      negative_text=[negative_prompt],
-                      feature_utils=feature_utils,
-                      net=net,
-                      fm=fm,
-                      rng=rng,
-                      cfg_strength=cfg_strength)
-    audio = audios.float().cpu()[0]
-
-    audio_save_path = tempfile.NamedTemporaryFile(delete=False, suffix='.flac').name
-    torchaudio.save(audio_save_path, audio, seq_cfg.sampling_rate)
-    
-    # 정보 로그 생성
-    info_log = f"""✅ TEXT TO AUDIO COMPLETE!
+👤 Results:
+   • Faces Detected: {face_count}
+   • Faces Blurred: {face_count}
 {'=' * 50}
-🎵 Audio Info:
-   • Duration: {duration:.2f} seconds
-   • Sample Rate: {seq_cfg.sampling_rate} Hz
+💾 Ready to download!"""
+    
+    return result_pil, info_log
+
+
+def process_video(
+    video: Optional[str],
+    conf: float,
+    iou: float,
+    expand_ratio: float,
+    mode_choice: str,
+    blur_intensity: int,
+    mosaic_size: int,
+    progress: gr.Progress = gr.Progress()
+) -> Tuple[Optional[str], str]:
+    """Main video processing handler"""
+    if video is None:
+        return None, "⚠️ Please upload a video first!"
+    
+    # Set blur parameters based on mode
+    blur_kernel = blur_intensity if mode_choice == "Gaussian Blur" else 51
+    mosaic = mosaic_size if mode_choice == "Mosaic Effect" else 15
+    
+    try:
+        output_path, total_faces, total_frames = blur_faces_in_video(
+            video, conf, iou, expand_ratio,
+            mode_choice, blur_kernel, mosaic, progress
+        )
+        
+        intensity_value = blur_intensity if mode_choice == "Gaussian Blur" else mosaic_size
+        info_log = f"""✅ VIDEO PROCESSING COMPLETE!
 {'=' * 50}
-⚙️ Generation Settings:
-   • Seed: {seed}
-   • Steps: {num_steps}
-   • CFG Strength: {cfg_strength}
+🎥 Video Info:
+   • Total Frames: {total_frames}
+   • Output Path: {os.path.basename(output_path)}
 {'=' * 50}
-📝 Prompts:
-   • Prompt: {prompt[:40]}{'...' if len(prompt) > 40 else ''}
-   • Negative: {negative_prompt[:30]}{'...' if len(negative_prompt) > 30 else ''}
+🔍 Detection Settings:
+   • Confidence: {conf}
+   • IoU Threshold: {iou}
+   • Box Expansion: {expand_ratio}
 {'=' * 50}
-💾 Audio ready to download!"""
+🎨 Blur Settings:
+   • Style: {mode_choice}
+   • Intensity: {intensity_value}
+{'=' * 50}
+👤 Results:
+   • Total Faces Detected: {total_faces}
+   • Frames Processed: {total_frames}
+{'=' * 50}
+💾 Ready to download!"""
+        
+        return output_path, info_log
     
-    return audio_save_path, info_log
+    except Exception as e:
+        return None, f"❌ Error: {str(e)}"
 
 
 # ============================================
-# 🎨 Comic Classic Theme - Toon Playground
+# CSS Styling - Comic Classic Theme
 # ============================================
-
-css = """
-/* ===== 🎨 Google Fonts Import ===== */
+CSS = """
+/* Google Fonts */
 @import url('https://fonts.googleapis.com/css2?family=Bangers&family=Comic+Neue:wght@400;700&display=swap');
 
-/* ===== 🎨 Comic Classic 배경 - 빈티지 페이퍼 + 도트 패턴 ===== */
+/* Background */
 .gradio-container {
     background-color: #FEF9C3 !important;
-    background-image: 
-        radial-gradient(#1F2937 1px, transparent 1px) !important;
+    background-image: radial-gradient(#1F2937 1px, transparent 1px) !important;
     background-size: 20px 20px !important;
     min-height: 100vh !important;
     font-family: 'Comic Neue', cursive, sans-serif !important;
 }
 
-/* ===== 허깅페이스 상단 요소 숨김 ===== */
-.huggingface-space-header,
-#space-header,
-.space-header,
-[class*="space-header"],
-.svelte-1ed2p3z,
-.space-header-badge,
-.header-badge,
-[data-testid="space-header"],
-.svelte-kqij2n,
-.svelte-1ax1toq,
-.embed-container > div:first-child {
+/* Hide HuggingFace header */
+.huggingface-space-header, #space-header, .space-header,
+[class*="space-header"], .svelte-1ed2p3z, .space-header-badge {
     display: none !important;
-    visibility: hidden !important;
-    height: 0 !important;
-    width: 0 !important;
-    overflow: hidden !important;
-    opacity: 0 !important;
-    pointer-events: none !important;
 }
 
-/* ===== Footer 완전 숨김 ===== */
-footer,
-.footer,
-.gradio-container footer,
-.built-with,
-[class*="footer"],
-.gradio-footer,
-.main-footer,
-div[class*="footer"],
-.show-api,
-.built-with-gradio,
-a[href*="gradio.app"],
-a[href*="huggingface.co/spaces"] {
+/* Hide footer */
+footer, .footer, .gradio-container footer, .built-with,
+[class*="footer"], .show-api, .built-with-gradio {
     display: none !important;
-    visibility: hidden !important;
-    height: 0 !important;
-    padding: 0 !important;
-    margin: 0 !important;
 }
 
-/* ===== 메인 컨테이너 ===== */
+/* Main container */
 #col-container { 
-    max-width: 1200px; 
+    max-width: 1400px; 
     margin: 0 auto; 
 }
 
-/* ===== 🎨 헤더 타이틀 - 코믹 스타일 ===== */
+/* Header */
 .header-text h1 {
     font-family: 'Bangers', cursive !important;
     color: #1F2937 !important;
     font-size: 3.5rem !important;
-    font-weight: 400 !important;
     text-align: center !important;
-    margin-bottom: 0.5rem !important;
-    text-shadow: 
-        4px 4px 0px #FACC15,
-        6px 6px 0px #1F2937 !important;
+    text-shadow: 4px 4px 0px #FACC15, 6px 6px 0px #1F2937 !important;
     letter-spacing: 3px !important;
     -webkit-text-stroke: 2px #1F2937 !important;
 }
 
-/* ===== 🎨 서브타이틀 ===== */
 .subtitle {
     text-align: center !important;
     font-family: 'Comic Neue', cursive !important;
     font-size: 1.2rem !important;
     color: #1F2937 !important;
-    margin-bottom: 1.5rem !important;
     font-weight: 700 !important;
 }
 
-/* ===== 🎨 탭 스타일 ===== */
-.tabs {
-    background: #FFFFFF !important;
+/* Stats cards */
+.stats-row {
+    display: flex !important;
+    justify-content: center !important;
+    gap: 1rem !important;
+    margin: 1.5rem 0 !important;
+    flex-wrap: wrap !important;
+}
+
+.stat-card {
+    background: linear-gradient(135deg, #3B82F6 0%, #8B5CF6 100%) !important;
     border: 3px solid #1F2937 !important;
     border-radius: 12px !important;
-    box-shadow: 6px 6px 0px #1F2937 !important;
-    padding: 10px !important;
+    padding: 1rem 1.5rem !important;
+    text-align: center !important;
+    box-shadow: 4px 4px 0px #1F2937 !important;
+    min-width: 120px !important;
 }
 
-.tab-nav {
-    background: #FACC15 !important;
+.stat-card .emoji { font-size: 2rem !important; display: block !important; }
+.stat-card .label { color: #FFFFFF !important; font-family: 'Comic Neue', cursive !important; font-weight: 700 !important; }
+
+/* Panels */
+.gr-panel, .gr-box, .gr-form, .block, .gr-group {
+    background: #FFFFFF !important;
+    border: 3px solid #1F2937 !important;
     border-radius: 8px !important;
-    padding: 5px !important;
-    border: 2px solid #1F2937 !important;
+    box-shadow: 6px 6px 0px #1F2937 !important;
+    transition: all 0.2s ease !important;
 }
 
-.tab-nav button {
-    font-family: 'Bangers', cursive !important;
-    font-size: 1.1rem !important;
-    letter-spacing: 1px !important;
-    color: #1F2937 !important;
-    background: transparent !important;
-    border: none !important;
-    padding: 10px 20px !important;
-    border-radius: 6px !important;
-    transition: all 0.2s ease !important;
+.gr-panel:hover, .block:hover {
+    transform: translate(-2px, -2px) !important;
+    box-shadow: 8px 8px 0px #1F2937 !important;
 }
 
-.tab-nav button:hover {
-    background: #FEF3C7 !important;
+/* Tabs */
+.gr-tabs {
+    border: 3px solid #1F2937 !important;
+    border-radius: 12px !important;
+    box-shadow: 6px 6px 0px #1F2937 !important;
 }
 
-.tab-nav button.selected {
-    background: #3B82F6 !important;
-    color: #FFFFFF !important;
-    box-shadow: 3px 3px 0px #1F2937 !important;
+.gr-tab-nav {
+    background: #FACC15 !important;
+    border-bottom: 3px solid #1F2937 !important;
 }
 
-/* ===== 🎨 카드/패널 - 만화 프레임 스타일 ===== */
-.gr-panel,
-.gr-box,
-.gr-form,
-.block,
-.gr-group {
-    background: #FFFFFF !important;
-    border: 3px solid #1F2937 !important;
-    border-radius: 8px !important;
-    box-shadow: 6px 6px 0px #1F2937 !important;
-    transition: all 0.2s ease !important;
+.gr-tab-nav button {
+    font-family: 'Bangers', cursive !important;
+    font-size: 1.2rem !important;
+    color: #1F2937 !important;
+    padding: 12px 24px !important;
 }
 
-.gr-panel:hover,
-.block:hover {
-    transform: translate(-2px, -2px) !important;
-    box-shadow: 8px 8px 0px #1F2937 !important;
+.gr-tab-nav button.selected {
+    background: #3B82F6 !important;
+    color: #FFFFFF !important;
 }
 
-/* ===== 🎨 입력 필드 (Textbox) ===== */
-textarea, 
-input[type="text"], 
-input[type="number"] {
+/* Inputs */
+textarea, input[type="text"], input[type="number"] {
     background: #FFFFFF !important;
     border: 3px solid #1F2937 !important;
     border-radius: 8px !important;
-    color: #1F2937 !important;
     font-family: 'Comic Neue', cursive !important;
-    font-size: 1rem !important;
     font-weight: 700 !important;
-    transition: all 0.2s ease !important;
 }
 
-textarea:focus, 
-input[type="text"]:focus, 
-input[type="number"]:focus {
+textarea:focus, input:focus {
     border-color: #3B82F6 !important;
     box-shadow: 4px 4px 0px #3B82F6 !important;
-    outline: none !important;
 }
 
-textarea::placeholder {
-    color: #9CA3AF !important;
-    font-weight: 400 !important;
+/* Dropdown */
+.gr-dropdown {
+    background: #FFFFFF !important;
+    border: 3px solid #1F2937 !important;
+    border-radius: 8px !important;
+    box-shadow: 3px 3px 0px #1F2937 !important;
 }
 
-/* ===== 🎨 Primary 버튼 - 코믹 블루 ===== */
-.gr-button-primary,
-button.primary,
-.gr-button.primary {
+/* Primary button */
+.gr-button-primary, button.primary, .process-btn {
     background: #3B82F6 !important;
     border: 3px solid #1F2937 !important;
     border-radius: 8px !important;
     color: #FFFFFF !important;
     font-family: 'Bangers', cursive !important;
-    font-weight: 400 !important;
     font-size: 1.3rem !important;
     letter-spacing: 2px !important;
     padding: 14px 28px !important;
     box-shadow: 5px 5px 0px #1F2937 !important;
-    transition: all 0.1s ease !important;
-    text-shadow: 1px 1px 0px #1F2937 !important;
 }
 
-.gr-button-primary:hover,
-button.primary:hover,
-.gr-button.primary:hover {
+.gr-button-primary:hover, button.primary:hover {
     background: #2563EB !important;
     transform: translate(-2px, -2px) !important;
     box-shadow: 7px 7px 0px #1F2937 !important;
 }
 
-.gr-button-primary:active,
-button.primary:active,
-.gr-button.primary:active {
+.gr-button-primary:active, button.primary:active {
     transform: translate(3px, 3px) !important;
     box-shadow: 2px 2px 0px #1F2937 !important;
 }
 
-/* ===== 🎨 Secondary 버튼 - 코믹 레드 ===== */
-.gr-button-secondary,
-button.secondary {
-    background: #EF4444 !important;
-    border: 3px solid #1F2937 !important;
-    border-radius: 8px !important;
-    color: #FFFFFF !important;
-    font-family: 'Bangers', cursive !important;
-    font-weight: 400 !important;
-    font-size: 1.1rem !important;
-    letter-spacing: 1px !important;
-    box-shadow: 4px 4px 0px #1F2937 !important;
-    transition: all 0.1s ease !important;
-    text-shadow: 1px 1px 0px #1F2937 !important;
-}
-
-.gr-button-secondary:hover,
-button.secondary:hover {
-    background: #DC2626 !important;
-    transform: translate(-2px, -2px) !important;
-    box-shadow: 6px 6px 0px #1F2937 !important;
-}
-
-/* ===== 🎨 로그 출력 영역 ===== */
+/* Log output */
 .info-log textarea {
     background: #1F2937 !important;
     color: #10B981 !important;
     font-family: 'Courier New', monospace !important;
-    font-size: 0.9rem !important;
-    font-weight: 400 !important;
     border: 3px solid #10B981 !important;
-    border-radius: 8px !important;
     box-shadow: 4px 4px 0px #10B981 !important;
 }
 
-/* ===== 🎨 비디오/오디오 영역 ===== */
-.gr-video,
-.gr-audio,
-video,
-audio {
+/* Image/Video containers */
+.gr-image, .gr-video {
     border: 4px solid #1F2937 !important;
     border-radius: 8px !important;
     box-shadow: 8px 8px 0px #1F2937 !important;
-    overflow: hidden !important;
-    background: #FFFFFF !important;
 }
 
-/* ===== 🎨 갤러리 스타일 ===== */
-.gr-gallery {
-    background: #FFFFFF !important;
+/* Accordion */
+.gr-accordion {
+    background: #FACC15 !important;
     border: 3px solid #1F2937 !important;
     border-radius: 8px !important;
-    box-shadow: 6px 6px 0px #1F2937 !important;
-    padding: 10px !important;
-}
-
-.gr-gallery .thumbnail-item {
-    border: 3px solid #1F2937 !important;
-    border-radius: 6px !important;
-    transition: all 0.2s ease !important;
-    overflow: hidden !important;
-}
-
-.gr-gallery .thumbnail-item:hover {
-    transform: scale(1.05) !important;
-    box-shadow: 4px 4px 0px #3B82F6 !important;
-}
-
-/* ===== 🎨 슬라이더 스타일 ===== */
-input[type="range"] {
-    accent-color: #3B82F6 !important;
-}
-
-.gr-slider {
-    background: #FFFFFF !important;
-}
-
-/* ===== 🎨 라벨 스타일 ===== */
-label,
-.gr-input-label,
-.gr-block-label {
-    color: #1F2937 !important;
-    font-family: 'Comic Neue', cursive !important;
-    font-weight: 700 !important;
-    font-size: 1rem !important;
-}
-
-span.gr-label {
-    color: #1F2937 !important;
-}
-
-/* ===== 🎨 정보 텍스트 ===== */
-.gr-info,
-.info {
-    color: #6B7280 !important;
-    font-family: 'Comic Neue', cursive !important;
-    font-size: 0.9rem !important;
+    box-shadow: 4px 4px 0px #1F2937 !important;
 }
 
-/* ===== 🎨 Number Input 스타일 ===== */
-.gr-number input {
-    background: #FFFFFF !important;
-    border: 3px solid #1F2937 !important;
-    border-radius: 8px !important;
+/* Labels */
+label, .gr-input-label, .gr-block-label {
     color: #1F2937 !important;
     font-family: 'Comic Neue', cursive !important;
     font-weight: 700 !important;
-    box-shadow: 3px 3px 0px #1F2937 !important;
 }
 
-/* ===== 🎨 스크롤바 - 코믹 스타일 ===== */
-::-webkit-scrollbar {
-    width: 12px;
-    height: 12px;
-}
+/* Slider */
+input[type="range"] { accent-color: #3B82F6 !important; }
 
-::-webkit-scrollbar-track {
-    background: #FEF9C3;
-    border: 2px solid #1F2937;
-}
+/* Scrollbar */
+::-webkit-scrollbar { width: 12px; }
+::-webkit-scrollbar-track { background: #FEF9C3; border: 2px solid #1F2937; }
+::-webkit-scrollbar-thumb { background: #3B82F6; border: 2px solid #1F2937; }
 
-::-webkit-scrollbar-thumb {
-    background: #3B82F6;
-    border: 2px solid #1F2937;
-    border-radius: 0px;
-}
-
-::-webkit-scrollbar-thumb:hover {
-    background: #EF4444;
-}
+/* Selection */
+::selection { background: #FACC15; color: #1F2937; }
 
-/* ===== 🎨 선택 하이라이트 ===== */
-::selection {
-    background: #FACC15;
-    color: #1F2937;
-}
-
-/* ===== 🎨 링크 스타일 ===== */
-a {
-    color: #3B82F6 !important;
-    text-decoration: none !important;
-    font-weight: 700 !important;
-}
-
-a:hover {
-    color: #EF4444 !important;
-}
-
-/* ===== 🎨 Row/Column 간격 ===== */
-.gr-row {
-    gap: 1.5rem !important;
-}
-
-.gr-column {
-    gap: 1rem !important;
-}
-
-/* ===== 반응형 조정 ===== */
+/* Responsive */
 @media (max-width: 768px) {
-    .header-text h1 {
-        font-size: 2.2rem !important;
-        text-shadow: 
-            3px 3px 0px #FACC15,
-            4px 4px 0px #1F2937 !important;
-    }
-    
-    .gr-button-primary,
-    button.primary {
-        padding: 12px 20px !important;
-        font-size: 1.1rem !important;
-    }
-    
-    .gr-panel,
-    .block {
-        box-shadow: 4px 4px 0px #1F2937 !important;
-    }
-    
-    .tab-nav button {
-        font-size: 0.9rem !important;
-        padding: 8px 12px !important;
-    }
-}
-
-/* ===== 🎨 다크모드 비활성화 ===== */
-@media (prefers-color-scheme: dark) {
-    .gradio-container {
-        background-color: #FEF9C3 !important;
-    }
+    .header-text h1 { font-size: 2.2rem !important; }
+    .gr-button-primary { padding: 12px 20px !important; font-size: 1.1rem !important; }
 }
 """
 
 
-# Gradio Blocks 인터페이스 생성
-with gr.Blocks(fill_height=True, css=css, title="MMAudio Studio") as demo:
-    
-    # HOME Badge
+# ============================================
+# UI Component Builders
+# ============================================
+def create_header() -> None:
+    """Create header section"""
     gr.HTML("""
         <div style="text-align: center; margin: 20px 0 10px 0;">
-            <a href="https://www.humangen.ai" target="_blank" style="text-decoration: none;">
+            <a href="https://www.humangen.ai" target="_blank">
                 <img src="https://img.shields.io/static/v1?label=🏠 HOME&message=HUMANGEN.AI&color=0000ff&labelColor=ffcc00&style=for-the-badge" alt="HOME">
             </a>
+            <a href="https://discord.gg/openfreeai" target="_blank" style="margin-left: 10px;">
+                <img src="https://img.shields.io/static/v1?label=Discord&message=OpenFree%20AI&color=5865F2&labelColor=1F2937&logo=discord&logoColor=white&style=for-the-badge" alt="Discord">
+            </a>
         </div>
     """)
     
-    # Header Title
-    gr.Markdown(
-        """
-        # 🎵 MMAUDIO STUDIO 🎬
-        """,
-        elem_classes="header-text"
-    )
+    gr.Markdown("# 🔒 ANSIM BLUR - FACE PRIVACY 🛡️", elem_classes="header-text")
+    gr.Markdown('<p class="subtitle">🎭 Advanced AI-Powered Face Detection & Privacy Protection! ✨</p>')
     
-    gr.Markdown(
-        """
-        <p class="subtitle">🔊 Generate Audio from Text or Video • Korean Supported! 한글지원 🇰🇷</p>
-        """,
-    )
+    gr.HTML("""
+        <div class="stats-row">
+            <div class="stat-card"><span class="emoji">🖼️</span><span class="label">Image Support</span></div>
+            <div class="stat-card"><span class="emoji">🎥</span><span class="label">Video Processing</span></div>
+            <div class="stat-card"><span class="emoji">⚡</span><span class="label">Real-time AI</span></div>
+            <div class="stat-card"><span class="emoji">🛡️</span><span class="label">Privacy First</span></div>
+        </div>
+    """)
     
-    with gr.Tabs():
-        # Tab 1: Video Search
-        with gr.TabItem("🔍 Video Search"):
-            gr.Markdown(
-                """
-                <p style="text-align: center; font-family: 'Comic Neue', cursive; font-weight: 700; color: #1F2937; margin-bottom: 1rem;">
-                    📹 Search for videos from Pixabay to use as input!
-                </p>
-                """
-            )
-            
-            with gr.Row():
-                with gr.Column(scale=1):
-                    search_query = gr.Textbox(
-                        label="🔎 Search Query (한글지원)" if translator else "🔎 Search Query",
-                        placeholder="Enter search keywords...",
-                        lines=1
-                    )
-                    search_btn = gr.Button(
-                        "🔍 SEARCH VIDEOS!",
-                        variant="primary",
-                        size="lg"
-                    )
-                
-            search_gallery = gr.Gallery(
-                label="📺 Search Results",
-                columns=4,
-                rows=5,
-                height=500
-            )
-            
-            search_btn.click(
-                fn=search_videos,
-                inputs=[search_query],
-                outputs=[search_gallery]
-            )
+    gr.Markdown(f"""
+        <p style="text-align: center; font-family: 'Comic Neue', cursive; font-weight: 700; color: #1F2937;">
+            🖥️ Running on: <span style="color: #3B82F6;">{detector.device.upper()}</span>
+        </p>
+    """)
+
+
+def create_detection_settings(suffix: str = "") -> Tuple[gr.Slider, gr.Slider, gr.Slider]:
+    """Create detection settings accordion"""
+    with gr.Accordion("⚙️ Detection Settings", open=True):
+        conf = gr.Slider(
+            **SLIDER_CONFIG["confidence"],
+            value=DEFAULT_CONFIG["confidence"],
+            label="🎯 Confidence Threshold"
+        )
+        iou = gr.Slider(
+            **SLIDER_CONFIG["iou"],
+            value=DEFAULT_CONFIG["iou"],
+            label="📐 NMS IoU"
+        )
+        expand = gr.Slider(
+            **SLIDER_CONFIG["expand"],
+            value=DEFAULT_CONFIG["expand_ratio"],
+            label="🔲 Box Expansion"
+        )
+    return conf, iou, expand
+
+
+def create_blur_settings(suffix: str = "") -> Tuple[gr.Dropdown, gr.Slider, gr.Slider]:
+    """Create blur settings accordion"""
+    with gr.Accordion("🎨 Blur Settings", open=True):
+        mode = gr.Dropdown(
+            choices=BLUR_MODES,
+            value=BLUR_MODES[0],
+            label="🖌️ Style"
+        )
+        blur_intensity = gr.Slider(
+            **SLIDER_CONFIG["blur"],
+            value=DEFAULT_CONFIG["blur_intensity"],
+            label="💨 Blur Intensity"
+        )
+        mosaic_size = gr.Slider(
+            **SLIDER_CONFIG["mosaic"],
+            value=DEFAULT_CONFIG["mosaic_size"],
+            label="🧩 Mosaic Size"
+        )
+    return mode, blur_intensity, mosaic_size
+
+
+def create_footer() -> None:
+    """Create footer with instructions"""
+    gr.Markdown("""
+        <div style="background: linear-gradient(135deg, #EFF6FF 0%, #DBEAFE 100%); border: 3px solid #3B82F6; border-radius: 12px; padding: 1.5rem; box-shadow: 6px 6px 0px #1F2937; margin-top: 2rem;">
+            <h3 style="font-family: 'Bangers', cursive; color: #1F2937; font-size: 1.3rem;">📝 HOW TO USE</h3>
+            <ol style="font-family: 'Comic Neue', cursive; color: #1F2937; font-weight: 700;">
+                <li>Upload an image or video containing faces</li>
+                <li>Adjust detection settings (confidence, IoU, expansion)</li>
+                <li>Choose blur style (Gaussian or Mosaic)</li>
+                <li>Click the Process button and wait for results</li>
+                <li>Download your privacy-protected media!</li>
+            </ol>
+        </div>
         
-        # Tab 2: Video to Audio
-        with gr.TabItem("🎬 Video-to-Audio"):
-            gr.Markdown(
-                """
-                <p style="text-align: center; font-family: 'Comic Neue', cursive; font-weight: 700; color: #1F2937; margin-bottom: 1rem;">
-                    🎥 Upload a video and generate matching audio!
-                </p>
-                """
-            )
+        <div style="background: linear-gradient(135deg, #FEF3C7 0%, #FDE68A 100%); border: 3px solid #F59E0B; border-radius: 12px; padding: 1.5rem; box-shadow: 6px 6px 0px #1F2937; margin-top: 1rem;">
+            <h3 style="font-family: 'Bangers', cursive; color: #1F2937; font-size: 1.3rem;">💡 TIPS</h3>
+            <ul style="font-family: 'Comic Neue', cursive; color: #1F2937; font-weight: 700;">
+                <li>Lower confidence = more faces detected (may include false positives)</li>
+                <li>Higher blur intensity = stronger privacy protection</li>
+                <li>Mosaic effect works better for artistic results</li>
+                <li>Video processing may take time depending on length</li>
+            </ul>
+        </div>
+    """)
+
+
+# ============================================
+# Main Application Builder
+# ============================================
+def create_app() -> gr.Blocks:
+    """Build and return the Gradio application"""
+    
+    with gr.Blocks(
+        fill_height=True, 
+        css=CSS, 
+        title="Ansim Blur - Face Privacy Protection",
+        theme=gr.themes.Default()
+    ) as app:
+        
+        # Header
+        create_header()
+        
+        # Main tabs
+        with gr.Tabs():
             
-            with gr.Row(equal_height=False):
-                with gr.Column(scale=1):
-                    v2a_video = gr.Video(label="📹 Input Video")
-                    v2a_prompt = gr.Textbox(
-                        label="✏️ Prompt (한글지원)" if translator else "✏️ Prompt",
-                        placeholder="Describe the audio you want...",
-                        lines=2
-                    )
-                    v2a_negative = gr.Textbox(
-                        label="🚫 Negative Prompt",
-                        value="music",
-                        lines=1
-                    )
+            # === IMAGE TAB ===
+            with gr.Tab("📸 Image Processing"):
+                with gr.Row(equal_height=False):
                     
-                    with gr.Row():
-                        v2a_seed = gr.Number(label="🎲 Seed", value=0)
-                        v2a_steps = gr.Number(label="🔄 Steps", value=25)
-                    
-                    with gr.Row():
-                        v2a_cfg = gr.Number(label="🎯 Guidance Scale", value=4.5)
-                        v2a_duration = gr.Number(label="⏱️ Duration (sec)", value=8)
-                    
-                    v2a_btn = gr.Button(
-                        "🎬 GENERATE AUDIO! 🔊",
-                        variant="primary",
-                        size="lg"
-                    )
+                    # Left column - Input
+                    with gr.Column(scale=1, min_width=400):
+                        input_image = gr.Image(
+                            label="🖼️ Upload Image",
+                            type="pil",
+                            height=350
+                        )
+                        
+                        conf_img, iou_img, expand_img = create_detection_settings("img")
+                        mode_img, blur_img, mosaic_img = create_blur_settings("img")
+                        
+                        process_img_btn = gr.Button(
+                            "🔍 PROCESS IMAGE! 🎭",
+                            variant="primary",
+                            size="lg",
+                            elem_classes="process-btn"
+                        )
                     
-                    with gr.Accordion("📜 Generation Log", open=True):
-                        v2a_log = gr.Textbox(
-                            label="",
-                            placeholder="Upload video and click generate...",
-                            lines=12,
-                            interactive=False,
-                            elem_classes="info-log"
+                    # Right column - Output
+                    with gr.Column(scale=1, min_width=400):
+                        output_image = gr.Image(
+                            label="🖼️ Processed Result",
+                            type="pil",
+                            height=350
                         )
-                
-                with gr.Column(scale=1):
-                    v2a_output = gr.Video(label="🎥 Generated Result", height=400)
-                    gr.Markdown(
-                        """
-                        <p style="text-align: center; margin-top: 15px; font-weight: 700; color: #1F2937;">
-                            💡 Right-click on the video to save!
-                        </p>
-                        """
-                    )
+                        
+                        with gr.Accordion("📜 Processing Log", open=True):
+                            info_log_img = gr.Textbox(
+                                label="",
+                                placeholder="Upload an image and click process...",
+                                lines=12,
+                                max_lines=18,
+                                interactive=False,
+                                elem_classes="info-log"
+                            )
             
-            v2a_btn.click(
-                fn=video_to_audio,
-                inputs=[v2a_video, v2a_prompt, v2a_negative, v2a_seed, v2a_steps, v2a_cfg, v2a_duration],
-                outputs=[v2a_output, v2a_log]
-            )
-        
-        # Tab 3: Text to Audio
-        with gr.TabItem("🎵 Text-to-Audio"):
-            gr.Markdown(
-                """
-                <p style="text-align: center; font-family: 'Comic Neue', cursive; font-weight: 700; color: #1F2937; margin-bottom: 1rem;">
-                    ✨ Generate audio from text description!
-                </p>
-                """
-            )
-            
-            with gr.Row(equal_height=False):
-                with gr.Column(scale=1):
-                    t2a_prompt = gr.Textbox(
-                        label="✏️ Prompt (한글지원)" if translator else "✏️ Prompt",
-                        placeholder="Describe the audio you want to generate...",
-                        lines=3
-                    )
-                    t2a_negative = gr.Textbox(
-                        label="🚫 Negative Prompt",
-                        placeholder="What to avoid...",
-                        lines=1
-                    )
-                    
-                    with gr.Row():
-                        t2a_seed = gr.Number(label="🎲 Seed", value=0)
-                        t2a_steps = gr.Number(label="🔄 Steps", value=25)
+            # === VIDEO TAB ===
+            with gr.Tab("🎬 Video Processing"):
+                with gr.Row(equal_height=False):
                     
-                    with gr.Row():
-                        t2a_cfg = gr.Number(label="🎯 Guidance Scale", value=4.5)
-                        t2a_duration = gr.Number(label="⏱️ Duration (sec)", value=8)
-                    
-                    t2a_btn = gr.Button(
-                        "🎵 GENERATE AUDIO! ✨",
-                        variant="primary",
-                        size="lg"
-                    )
+                    # Left column - Input
+                    with gr.Column(scale=1, min_width=400):
+                        input_video = gr.Video(
+                            label="🎥 Upload Video",
+                            height=350
+                        )
+                        
+                        conf_vid, iou_vid, expand_vid = create_detection_settings("vid")
+                        mode_vid, blur_vid, mosaic_vid = create_blur_settings("vid")
+                        
+                        process_vid_btn = gr.Button(
+                            "🎬 PROCESS VIDEO! 🛡️",
+                            variant="primary",
+                            size="lg",
+                            elem_classes="process-btn"
+                        )
                     
-                    with gr.Accordion("📜 Generation Log", open=True):
-                        t2a_log = gr.Textbox(
-                            label="",
-                            placeholder="Enter prompt and click generate...",
-                            lines=12,
-                            interactive=False,
-                            elem_classes="info-log"
+                    # Right column - Output
+                    with gr.Column(scale=1, min_width=400):
+                        output_video = gr.Video(
+                            label="🎥 Processed Result",
+                            height=350
                         )
-                
-                with gr.Column(scale=1):
-                    t2a_output = gr.Audio(label="🔊 Generated Audio")
-                    gr.Markdown(
-                        """
-                        <p style="text-align: center; margin-top: 15px; font-weight: 700; color: #1F2937;">
-                            💡 Click the download button to save!
-                        </p>
-                        """
-                    )
-            
-            t2a_btn.click(
-                fn=text_to_audio,
-                inputs=[t2a_prompt, t2a_negative, t2a_seed, t2a_steps, t2a_cfg, t2a_duration],
-                outputs=[t2a_output, t2a_log]
-            )
+                        
+                        with gr.Accordion("📜 Processing Log", open=True):
+                            info_log_vid = gr.Textbox(
+                                label="",
+                                placeholder="Upload a video and click process...",
+                                lines=12,
+                                max_lines=18,
+                                interactive=False,
+                                elem_classes="info-log"
+                            )
+        
+        # Footer
+        create_footer()
+        
+        # === EVENT HANDLERS ===
+        process_img_btn.click(
+            fn=process_image,
+            inputs=[input_image, conf_img, iou_img, expand_img, mode_img, blur_img, mosaic_img],
+            outputs=[output_image, info_log_img]
+        )
+        
+        process_vid_btn.click(
+            fn=process_video,
+            inputs=[input_video, conf_vid, iou_vid, expand_vid, mode_vid, blur_vid, mosaic_vid],
+            outputs=[output_video, info_log_vid]
+        )
+    
+    return app
 
 
-# 메인 실행
+# ============================================
+# Entry Point
+# ============================================
 if __name__ == "__main__":
-    if translator is None:
-        log.warning("Translation model failed to load. Korean translation will be disabled.")
-        
-    demo.launch(allowed_paths=[output_dir])
+    app = create_app()
+    app.launch()