Update preprocessing_router.py

preprocessing_router.py

@@ -25,6 +25,7 @@ from sklearn.cluster import KMeans
 from sklearn.neighbors import KNeighborsClassifier
 
 from svision_client import get_face_embeddings_simple
+from asr_client import get_voice_embedding
 
 
 ROOT = Path("/tmp/veureu")
@@ -269,6 +270,43 @@ async def finalize_casting(
         print("-" * 30)
     print("="*50 + "\n")
 
+    print("\n" + "="*50)
+    print("DEBUG: PROCESANDO VOCES")
+    print("="*50)
+    
+    for v_idx, cluster in enumerate(voice_clusters):
+        v_name = cluster.get("name", f"Voz_{v_idx}")
+        v_folder = cluster.get("folder", "")
+        v_files = cluster.get("files", [])
+    
+        print(f"🔊 Voz {v_idx+1}: {v_name}")
+        print(f"   📂 Carpeta origen: {v_folder}")
+        print(f"   🎵 Audios seleccionados ({len(v_files)}):")
+    
+        for f in v_files:
+            f_name = Path(f).name   # seg_0001.wav
+            f_path = Path(v_folder) / f_name
+    
+            if not f_path.exists():
+                print(f"❌ NO EXISTE: {f_path}")
+                continue
+    
+            # ⬇️ AQUÍ tu función de embeddings de audio
+            v_emb = get_voice_embedding(str(f_path))
+    
+            print(f"      - original: {f}")
+            print(f"      - resolved: {f_path}")
+            print(f"      - emb: {'OK' if v_emb else 'VACÍO'}")
+    
+            if v_emb:
+                casting_json["voice_col"].append({
+                    "nombre": v_name,
+                    "embedding": v_emb,
+                })
+    
+        print("-" * 30)
+
+
     print(casting_json)
 
     MEDIA_ROOT = _P("/data/media")