Update app.py

app.py

@@ -223,8 +223,8 @@ def save_wav(audio, file_path):
         print(f"❌ Failed to save WAV: {e}")
         return False
 
-def load_tts_model():
-    """Load TTS model - Force Tacotron2 for reliability"""
+def load_tts_model(voice_style="default"):
+    """Load TTS model with different voice options"""
     global tts, model_loaded, current_model, voice_cloning_supported, model_loading, model_load_attempts
     
     if model_loading:
@@ -245,30 +245,81 @@ def load_tts_model():
         sys.stdin = StringIO('y\n')
         
         try:
-            print("🚀 Loading Tacotron2 model (most reliable)...")
+            # Different models with different voice characteristics
+            model_options = {
+                "male_deep": {
+                    "name": "tts_models/en/vctk/vits",
+                    "description": "VITS - Multiple speakers (male/female options)",
+                    "speaker": "p225"  # Male voice
+                },
+                "male_medium": {
+                    "name": "tts_models/en/vctk/vits", 
+                    "description": "VITS - Multiple speakers",
+                    "speaker": "p226"  # Male voice
+                },
+                "female_1": {
+                    "name": "tts_models/en/vctk/vits",
+                    "description": "VITS - Multiple speakers", 
+                    "speaker": "p227"  # Female voice
+                },
+                "female_2": {
+                    "name": "tts_models/en/vctk/vits",
+                    "description": "VITS - Multiple speakers",
+                    "speaker": "p228"  # Female voice  
+                },
+                "default_female": {
+                    "name": "tts_models/en/ljspeech/tacotron2-DDC",
+                    "description": "Tacotron2 - Default female (current)",
+                    "speaker": None
+                },
+                "clear_male": {
+                    "name": "tts_models/en/ek1/tacotron2",
+                    "description": "Tacotron2 - Clear male voice",
+                    "speaker": None
+                }
+            }
             
-            # Force Tacotron2 - it's the most reliable model
-            tts = TTS("tts_models/en/ljspeech/tacotron2-DDC").to(DEVICE)
+            selected_model = model_options.get(voice_style, model_options["default_female"])
+            
+            print(f"🚀 Loading {selected_model['description']}...")
+            
+            # Load the selected model
+            tts = TTS(selected_model["name"]).to(DEVICE)
             
             # Test the model
             test_path = "/tmp/test_output.wav"
-            tts.tts_to_file(text="Test", file_path=test_path)
+            
+            if selected_model["speaker"]:
+                # For VITS model with speaker selection
+                tts.tts_to_file(
+                    text="Test voice", 
+                    file_path=test_path,
+                    speaker=selected_model["speaker"]
+                )
+            else:
+                # For standard models
+                tts.tts_to_file(text="Test voice", file_path=test_path)
             
             if os.path.exists(test_path):
                 os.remove(test_path)
-                print("✅ Tacotron2 model tested and working!")
+                print(f"✅ {selected_model['description']} loaded successfully!")
             else:
                 raise Exception("Test failed - no file created")
             
             model_loaded = True
-            current_model = "tts_models/en/ljspeech/tacotron2-DDC"
-            voice_cloning_supported = False  # Tacotron2 doesn't support voice cloning
-            print("✅ Tacotron2 model loaded successfully!")
+            current_model = selected_model["name"]
+            voice_cloning_supported = False
             return True
             
         except Exception as e:
-            print(f"❌ Tacotron2 model failed: {e}")
-            return False
+            print(f"❌ Model loading failed: {e}")
+            # Fallback to default
+            print("🔄 Falling back to default Tacotron2...")
+            tts = TTS("tts_models/en/ljspeech/tacotron2-DDC").to(DEVICE)
+            model_loaded = True
+            current_model = "tts_models/en/ljspeech/tacotron2-DDC"
+            voice_cloning_supported = False
+            return True
                 
         finally:
             sys.stdin = old_stdin