Upload 3 files

audio_tools.py

@@ -28,10 +28,7 @@ try:
     import torchaudio as ta
     import torchaudio.transforms as T
     HAS_TORCHAUDIO = True
-    try:
-        ta.set_audio_backend("soundfile")
-    except Exception:
-        pass
+    # Note: ta.set_audio_backend is deprecated in newer torchaudio versions
 except Exception:
     HAS_TORCHAUDIO = False
     ta = None  # type: ignore

identity_manager.py

@@ -2,9 +2,10 @@
 # File: identity_manager.py
 # =========================
 from __future__ import annotations
-from typing import Any, Dict, List, Optional, Tuple
+from typing import Any, Dict, List, Optional, Tuple, TYPE_CHECKING
 
-from chromadb.api.types import Collection
+if TYPE_CHECKING:
+    from chromadb.api.models.Collection import Collection
 
 
 class IdentityManager:
@@ -13,7 +14,7 @@ class IdentityManager:
     y su proyección sobre frames, clips y SRT.
     """
 
-    def __init__(self, face_collection: Optional[Collection] = None, voice_collection: Optional[Collection] = None):
+    def __init__(self, face_collection: Optional["Collection"] = None, voice_collection: Optional["Collection"] = None):
         self.face_collection = face_collection
         self.voice_collection = voice_collection
 

vision_tools.py

@@ -30,7 +30,7 @@ import torchaudio
 import torchaudio.transforms as T
 from transformers import WhisperProcessor, WhisperForConditionalGeneration
 from pyannote.audio import Pipeline as PyannotePipeline
-from speechbrain.pretrained import SpeakerRecognition
+from speechbrain.inference.speaker import SpeakerRecognition
 from pydub import AudioSegment
 from sklearn.cluster import KMeans
 from sklearn.metrics import silhouette_score