Update app.py

app.py

@@ -125,7 +125,7 @@ def generate_blip_name(frame: np.ndarray) -> str:
         print(f"BLIP error: {e}")
         return "video"
 
-# --- 🎨 NEW: Simple Thumbnail Generation Function ---
+# --- 🎨 Thumbnail Generation Functions ---
 
 def create_simple_gradient_thumbnail(rgb_frame, depth_frame):
     """
@@ -160,6 +160,8 @@ def create_simple_gradient_thumbnail(rgb_frame, depth_frame):
         
         # 2. Erstelle Gradient-Bereiche
         height = rgb_thumb.shape[0]
+        print(f"DEBUG: Height for zones: {height}")  # Debug height
+        
         rgb_end = int(height * 0.60)      # 60% RGB
         gradient_start = rgb_end          # Gradient startet bei 60%
         gradient_end = int(height * 0.80) # Gradient endet bei 80%
@@ -498,7 +500,7 @@ current_video_url = None
 blip_generated_name = ""
 original_filename = ""
 
-# --- MAIN INFERENCE FUNCTION - FINAL FIX ---
+# --- MAIN INFERENCE FUNCTION - FINAL THUMBNAIL FIX ---
 def infer_video_depth_from_source(upload_video, video_url, filename, use_blip, create_thumbnail, *args):
     """Process video to generate depth maps and RGBD output - FINAL THUMBNAIL FIX"""
     try:
@@ -576,7 +578,7 @@ def infer_video_depth_from_source(upload_video, video_url, filename, use_blip, c
                 stitched = cv2.hconcat([rgb, depth_resized])
                 stitched_frames.append(stitched)
                 
-                # 🎯 FINAL FIX: Nutze erstes Frame-Pair für Thumbnail (bereits perfekt passend)
+                # 🎯 CREATE THUMBNAIL from first perfectly matched RGB+Depth pair (but don't embed yet)
                 if i == 0 and create_thumbnail:
                     print("Creating thumbnail from first perfectly matched RGB+Depth pair...")
                     try:
@@ -586,7 +588,7 @@ def infer_video_depth_from_source(upload_video, video_url, filename, use_blip, c
                         # Erstelle Thumbnail mit den bereits perfekt passenden Frames
                         thumbnail = create_simple_gradient_thumbnail(rgb, depth_resized)
                         
-                        print("✅ Thumbnail created from first RGBD pair")
+                        print("✅ Thumbnail created from first RGBD pair (not embedded yet)")
                         
                     except Exception as e:
                         print(f"❌ Thumbnail creation failed: {e}")
@@ -598,14 +600,6 @@ def infer_video_depth_from_source(upload_video, video_url, filename, use_blip, c
             save_video(np.array(stitched_frames), rgbd_video_path, fps=fps)
             print("✅ RGBD video created successfully")
 
-            # Embed thumbnail in videos if created
-            if create_thumbnail and thumbnail is not None:
-                embed_thumbnail_in_video(vis_video_path, thumbnail, base_name)
-                embed_thumbnail_in_video(rgbd_video_path, thumbnail, base_name)
-                print("✅ Thumbnail embedded in videos")
-            elif create_thumbnail:
-                print("❌ No thumbnail to embed")
-
             # Add audio from original video if possible
             try:
                 temp_audio_path = rgbd_video_path.replace('.mp4', '_audio.mp4')
@@ -617,10 +611,19 @@ def infer_video_depth_from_source(upload_video, video_url, filename, use_blip, c
                 result = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
                 if result.returncode == 0:
                     os.replace(temp_audio_path, rgbd_video_path)
-                    rgbd_path = rgbd_video_path
+                    print("✅ Audio added successfully")
             except Exception as e:
                 print(f"Audio processing failed: {e}")
-                rgbd_path = rgbd_video_path
+            
+            rgbd_path = rgbd_video_path
+
+            # 🎯 FINAL FIX: Embed thumbnail ONLY in RGBD video AFTER all processing
+            if create_thumbnail and thumbnail is not None:
+                print("Embedding thumbnail in RGBD video only (after all processing)...")
+                embed_thumbnail_in_video(rgbd_video_path, thumbnail, base_name)
+                print("✅ Thumbnail embedded in RGBD video only")
+            elif create_thumbnail:
+                print("❌ No thumbnail to embed")
 
         else:
             # If no RGBD stitching, fallback to old thumbnail method
@@ -631,8 +634,7 @@ def infer_video_depth_from_source(upload_video, video_url, filename, use_blip, c
                     if rgb_frame is not None:
                         # Simple RGB thumbnail with logo
                         thumbnail = add_depth_logo(rgb_frame)
-                        embed_thumbnail_in_video(vis_video_path, thumbnail, base_name)
-                        print("✅ Fallback RGB thumbnail created and embedded")
+                        print("✅ Fallback RGB thumbnail created")
                     else:
                         print("❌ Could not create fallback thumbnail")
                         thumbnail = None