Update app_enhanced.py

app_enhanced.py

@@ -24,7 +24,7 @@ def gpu_warmup():
     return True
 
 # ======================================================
-# 💾 PERSISTENT STORAGE
+# 💾 STORAGE SETUP
 # ======================================================
 if os.path.exists('/data'):
     BASE_STORAGE_PATH = '/data'
@@ -39,6 +39,19 @@ SAVED_COMICS_DIR = os.path.join(BASE_STORAGE_PATH, "saved_comics")
 os.makedirs(BASE_USER_DIR, exist_ok=True)
 os.makedirs(SAVED_COMICS_DIR, exist_ok=True)
 
+# ======================================================
+# 🔧 APP CONFIG
+# ======================================================
+app = Flask(__name__)
+app.config['MAX_CONTENT_LENGTH'] = 500 * 1024 * 1024  # 500MB Limit
+
+def generate_save_code(length=8):
+    chars = string.ascii_uppercase + string.digits
+    while True:
+        code = ''.join(random.choices(chars, k=length))
+        if not os.path.exists(os.path.join(SAVED_COMICS_DIR, code)):
+            return code
+
 # ======================================================
 # 🧱 DATA CLASSES
 # ======================================================
@@ -60,32 +73,17 @@ class Page:
         self.panels = panels
         self.bubbles = bubbles
 
-# ======================================================
-# 🔧 APP CONFIG
-# ======================================================
-app = Flask(__name__)
-app.config['MAX_CONTENT_LENGTH'] = 500 * 1024 * 1024
-
-def generate_save_code(length=8):
-    chars = string.ascii_uppercase + string.digits
-    while True:
-        code = ''.join(random.choices(chars, k=length))
-        if not os.path.exists(os.path.join(SAVED_COMICS_DIR, code)):
-            return code
-
 # ======================================================
 # 🧠 GPU GENERATION (FULL TEXT + HD IMAGE)
 # ======================================================
 @spaces.GPU(duration=300)
 def generate_comic_gpu(video_path, user_dir, frames_dir, metadata_path, target_pages):
-    print(f"🚀 Generating Comic with TEXT: {video_path}")
+    print(f"🚀 Generating HD Comic with TEXT: {video_path}")
     
     import cv2
     import srt
     import numpy as np
-    # Import backend modules for text extraction
-    from backend.subtitles.subs_real import get_real_subtitles
-    # Optional: AI placement (can be slow, using center fallback for speed/visibility)
+    from backend.subtitles.subs_real import get_real_subtitles # Ensure this backend file exists
     
     # 1. Video Setup
     cap = cv2.VideoCapture(video_path)
@@ -95,16 +93,14 @@ def generate_comic_gpu(video_path, user_dir, frames_dir, metadata_path, target_p
     duration = total_frames / fps
     cap.release()
 
-    # 2. GENERATE SUBTITLES (The Real Logic)
+    # 2. GENERATE SUBTITLES
     user_srt = os.path.join(user_dir, 'subs.srt')
     try:
         print("🎙️ Extracting subtitles...")
         get_real_subtitles(video_path)
-        # Move the generated SRT to user dir
         if os.path.exists('test1.srt'):
             shutil.move('test1.srt', user_srt)
         elif not os.path.exists(user_srt):
-            # Fallback if extractor failed silently
             with open(user_srt, 'w') as f: f.write("1\n00:00:01,000 --> 00:00:04,000\n(No Text)\n")
     except Exception as e:
         print(f"⚠️ Subtitle error: {e}")
@@ -117,30 +113,26 @@ def generate_comic_gpu(video_path, user_dir, frames_dir, metadata_path, target_p
 
     valid_subs = [s for s in all_subs if s.content and s.content.strip()]
     
-    # Create "Moments" from subtitles
     if valid_subs:
         raw_moments = [{'text': s.content.strip(), 'start': s.start.total_seconds(), 'end': s.end.total_seconds()} for s in valid_subs]
     else:
-        # If no speech, create time-based moments
         raw_moments = []
 
-    # 4. Determine Frames needed
+    # 4. Determine Frames needed (4 per page)
     panels_per_page = 4
     total_panels_needed = int(target_pages) * panels_per_page
 
     selected_moments = []
     if not raw_moments:
-        # Time based distribution if no text
         times = np.linspace(1, max(1, duration-1), total_panels_needed)
         for t in times: selected_moments.append({'text': '', 'start': t, 'end': t+1})
     elif len(raw_moments) <= total_panels_needed:
         selected_moments = raw_moments
     else:
-        # Sample moments evenly
         indices = np.linspace(0, len(raw_moments) - 1, total_panels_needed, dtype=int)
         selected_moments = [raw_moments[i] for i in indices]
         
-    # 5. Extract Frames (HD) & Metadata
+    # 5. Extract Frames (HD 1280x720)
     frame_metadata = {}
     cap = cv2.VideoCapture(video_path)
     count = 0
@@ -153,9 +145,8 @@ def generate_comic_gpu(video_path, user_dir, frames_dir, metadata_path, target_p
         ret, frame = cap.read()
         
         if ret:
-            # ----------------------------------------------------
-            # 🎯 HD EXTRACTION (1280x720) - Preserves quality
-            # ----------------------------------------------------
+            # 🎯 KEEP 16:9 ASPECT RATIO (1280x720)
+            # This ensures no data is lost. Frontend controls crop/zoom.
             frame = cv2.resize(frame, (1280, 720))
             
             fname = f"frame_{count:04d}.png"
@@ -169,20 +160,16 @@ def generate_comic_gpu(video_path, user_dir, frames_dir, metadata_path, target_p
     cap.release()
     with open(metadata_path, 'w') as f: json.dump(frame_metadata, f, indent=2)
         
-    # 6. Generate Bubbles with Text
+    # 6. Generate Bubbles
     bubbles_list = []
     for f in frame_files_ordered:
         dialogue = frame_metadata.get(f, {}).get('dialogue', '')
-        
-        # Determine Type
         b_type = 'speech'
         if '(' in dialogue: b_type = 'narration'
-        elif '!' in dialogue and dialogue.isupper(): b_type = 'reaction'
-        elif '?' in dialogue: b_type = 'speech'
+        elif '!' in dialogue: b_type = 'reaction'
         
-        # Place Bubble (Center Top 50px, 20px) to ensure visibility in Crop
-        # Users can drag it later.
-        bubbles_list.append(bubble(dialog=dialogue, x=50, y=50, type=b_type))
+        # Center bubbles initially
+        bubbles_list.append(bubble(dialog=dialogue, x=50, y=20, type=b_type))
             
     # 7. Construct Pages
     pages = []
@@ -192,7 +179,7 @@ def generate_comic_gpu(video_path, user_dir, frames_dir, metadata_path, target_p
         p_frames = frame_files_ordered[start_idx:end_idx]
         p_bubbles = bubbles_list[start_idx:end_idx]
         
-        # Pad with empty
+        # Pad with empty frames if not enough
         while len(p_frames) < 4:
              fname = f"empty_{i}_{len(p_frames)}.png"
              img = np.zeros((720, 1280, 3), dtype=np.uint8); img[:] = (30,30,30)
@@ -217,7 +204,6 @@ def generate_comic_gpu(video_path, user_dir, frames_dir, metadata_path, target_p
 def regen_frame_gpu(video_path, frames_dir, metadata_path, fname, direction):
     import cv2
     import json
-    # (Same fast regen logic)
     if not os.path.exists(metadata_path): return {"success": False, "message": "No metadata"}
     with open(metadata_path, 'r') as f: meta = json.load(f)
     if fname not in meta: return {"success": False, "message": "Frame not found"}
@@ -243,6 +229,28 @@ def regen_frame_gpu(video_path, frames_dir, metadata_path, fname, direction):
         return {"success": True, "message": f"Time: {new_t:.2f}s"}
     return {"success": False, "message": "End of video"}
 
+@spaces.GPU
+def get_frame_at_ts_gpu(video_path, frames_dir, metadata_path, fname, ts):
+    import cv2
+    import json
+    cap = cv2.VideoCapture(video_path)
+    cap.set(cv2.CAP_PROP_POS_MSEC, float(ts) * 1000)
+    ret, frame = cap.read()
+    cap.release()
+
+    if ret:
+        frame = cv2.resize(frame, (1280, 720)) # Keep HD
+        p = os.path.join(frames_dir, fname)
+        cv2.imwrite(p, frame)
+        if os.path.exists(metadata_path):
+            with open(metadata_path, 'r') as f: meta = json.load(f)
+            if fname in meta:
+                if isinstance(meta[fname], dict): meta[fname]['time'] = float(ts)
+                else: meta[fname] = float(ts)
+                with open(metadata_path, 'w') as f: json.dump(meta, f, indent=2)
+        return {"success": True, "message": f"Jumped to {ts}s"}
+    return {"success": False, "message": "Invalid timestamp"}
+
 # ======================================================
 # 💻 BACKEND CLASS
 # ======================================================
@@ -324,6 +332,8 @@ INDEX_HTML = '''
         width: 100%; height: 100%;
         position: relative;
         background: #000;
+        
+        /* Grid Variables */
         --y: 50%;      
         --t1: 100%; --t2: 100%; /* Hidden Right by default */
         --b1: 100%; --b2: 100%; /* Hidden Right by default */
@@ -331,7 +341,15 @@ INDEX_HTML = '''
     }
 
     .panel { position: absolute; top: 0; left: 0; width: 100%; height: 100%; overflow: hidden; background: #1a1a1a; cursor: grab; }
-    .panel img { width: 100%; height: 100%; object-fit: cover; transform-origin: center; transition: transform 0.05s ease-out; display: block; }
+    
+    /* IMAGE HANDLING: Object-fit cover fills the square. Pan/Zoom reveals hidden parts. */
+    .panel img { 
+        width: 100%; height: 100%; 
+        object-fit: cover; 
+        transform-origin: center; 
+        transition: transform 0.05s ease-out; 
+        display: block; 
+    }
     .panel img.panning { cursor: grabbing; transition: none; }
     .panel.selected { outline: 4px solid #2196F3; z-index: 5; }
 
@@ -445,7 +463,8 @@ INDEX_HTML = '''
         
         <div class="control-group">
             <label>🔍 Zoom (Mouse Wheel):</label>
-            <input type="range" id="zoom-slider" min="100" max="300" value="100" step="5" oninput="handleZoom(this.value)" disabled>
+            <!-- Min zoom 20 allowed to zoom OUT -->
+            <input type="range" id="zoom-slider" min="20" max="300" value="100" step="5" oninput="handleZoom(this.value)" disabled>
             <button onclick="resetPanelTransform()" class="secondary-btn">Reset View</button>
         </div>
         
@@ -489,7 +508,9 @@ INDEX_HTML = '''
             const panels = [];
             grid.querySelectorAll('.panel').forEach(pan => {
                 const img = pan.querySelector('img');
-                panels.push({ zoom: img.dataset.zoom, tx: img.dataset.translateX, ty: img.dataset.translateY });
+                const srcParts = img.src.split('frames/');
+                const fname = srcParts.length > 1 ? srcParts[1].split('?')[0] : '';
+                panels.push({ image: fname, zoom: img.dataset.zoom, tx: img.dataset.translateX, ty: img.dataset.translateY });
             });
             state.push({ layout, bubbles, panels });
         });
@@ -584,7 +605,19 @@ INDEX_HTML = '''
                 img.src = pan.src.includes('?') ? pan.src : pan.src + `?sid=${sid}`;
                 img.dataset.zoom = 100; img.dataset.translateX = 0; img.dataset.translateY = 0;
                 img.onmousedown = (e) => { e.preventDefault(); e.stopPropagation(); selectPanel(pDiv); dragType = 'pan'; activeObj = img; dragStart = {x:e.clientX, y:e.clientY}; img.classList.add('panning'); };
-                img.onwheel = (e) => { e.preventDefault(); let zoom = parseFloat(img.dataset.zoom); zoom += e.deltaY * -0.1; zoom = Math.min(Math.max(100, zoom), 300); img.dataset.zoom = zoom; updateImageTransform(img); if(selectedPanel === pDiv) document.getElementById('zoom-slider').value = zoom; saveState(); };
+                
+                // 🚀 ZOOM WHEEL LOGIC (Min zoom 20%)
+                img.onwheel = (e) => { 
+                    e.preventDefault(); 
+                    let zoom = parseFloat(img.dataset.zoom); 
+                    zoom += e.deltaY * -0.1; 
+                    zoom = Math.min(Math.max(20, zoom), 300); // Allow zoom out to 20%
+                    img.dataset.zoom = zoom; 
+                    updateImageTransform(img); 
+                    if(selectedPanel === pDiv) document.getElementById('zoom-slider').value = zoom; 
+                    saveState(); 
+                };
+                
                 pDiv.appendChild(img); grid.appendChild(pDiv);
             });