Upload 2 files

main_improved.py

@@ -121,6 +121,7 @@ def main():
     parser.add_argument("--face-model", choices=["insightface", "mediapipe"], default="insightface", help="Face detection model")
     parser.add_argument("--face-mode", choices=["auto", "1", "2"], default="auto", help="Face tracking mode: auto, 1, 2")
     parser.add_argument("--subtitle-config", help="Path to subtitle configuration JSON file")
+    parser.add_argument("--no-face-mode", choices=["padding", "zoom"], default="padding", help="Method to handle segments with no face detected: 'padding' (9:16 frame with black bars) or 'zoom' (Center Crop Zoom)")
     parser.add_argument("--face-detect-interval", type=str, default="0.17,1.0", help="Face detection interval in seconds. Single value or 'interval_1face,interval_2face'")
     parser.add_argument("--face-filter-threshold", type=float, default=0.35, help="Relative area threshold to ignore background faces (default: 0.35)")
     parser.add_argument("--face-two-threshold", type=float, default=0.60, help="Relative area threshold to trigger 2-face mode (default: 0.60)")
@@ -487,6 +488,34 @@ def main():
                 
                 save_json.save_viral_segments(viral_segments, project_folder=project_folder) 
 
+        # 3.5. Fix Raw Segments (missing timestamps)
+        if workflow_choice != "3" and viral_segments and "segments" in viral_segments:
+            segs = viral_segments.get("segments", [])
+            if segs and len(segs) > 0:
+                 # Check first segment for duration 0 but having start_time_ref or just check duration
+                 first = segs[0]
+                 # If duration is effectively 0 and we have a ref tag (or even if we dont, we cant cut 0s video)
+                 # We assume if duration is 0, it is raw.
+                 if first.get("duration", 0) == 0:
+                      print(i18n("Detected raw AI segments without timestamps (Duration 0). Running alignment..."))
+                      try:
+                          # Load transcript
+                          transcript = create_viral_segments.load_transcript(project_folder)
+                          # Process (Align)
+                          # Use None for output_count to keep all found segments
+                          viral_segments = create_viral_segments.process_segments(
+                              segs, 
+                              transcript, 
+                              args.min_duration, 
+                              args.max_duration, 
+                              output_count=None 
+                          )
+                          save_json.save_viral_segments(viral_segments, project_folder=project_folder)
+                          print(i18n("Segments aligned and saved."))
+                      except Exception as e:
+                          print(i18n("Failed to align raw segments: {}").format(e))
+                          # If alignment fails, it might crash later, but we tried. 
+
         # 4. Cut Segments
         # Se workflow for 3, pulamos corte
         if workflow_choice == "3":
@@ -547,7 +576,8 @@ def main():
                 active_speaker_motion_deadzone=args.active_speaker_motion_threshold,
                 active_speaker_motion_sensitivity=args.active_speaker_motion_sensitivity,
                 active_speaker_decay=args.active_speaker_decay,
-                segments_data=viral_segments.get("segments", []) if viral_segments else None
+                segments_data=viral_segments.get("segments", []) if viral_segments else None,
+                no_face_mode=args.no_face_mode
             )
 
 

prompt.txt

@@ -28,7 +28,7 @@ The transcript below is a continuous text stream with embedded **Time Tags** lik
 
 4.  **DURATION MATH:**
     -   Use the `(XXs)` tags to estimate duration.
-    -   Target: {min_duration}s to {max_duration}s.
+    -   CONSTRAINT: Segment MUST be between {min_duration}s and {max_duration}s.
 
 ### YOUR TASK:
 Analyze the transcript below. Find {amount} potential viral segments.