Make whisper use CUDA if available

app.py

@@ -100,9 +100,11 @@ def get_or_create_speaker(interface, audio_file):
         print(f"✅ Using cached speaker profile for {os.path.basename(audio_file)}")
         return speaker_cache[cache_key]
     
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+
     # Create new speaker profile
     print(f"🔄 Creating new speaker profile for {os.path.basename(audio_file)}")
-    speaker = interface.create_speaker(audio_file, whisper_model="large-v3-turbo")
+    speaker = interface.create_speaker(audio_file, whisper_model="large-v3-turbo", whisper_device=device)
     
     # Cache the speaker profile
     speaker_cache[cache_key] = speaker