updated

app.py

@@ -7,14 +7,7 @@ from scipy.io import wavfile
 
 # Explicitly import Bark components
 from bark import generate_audio, SAMPLE_RATE
-from bark.generation import preload_models, load_model, generate_text_semantic
-
-# Add device detection
-device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-print("CUDA available:", torch.cuda.is_available())
-print("CUDA device count:", torch.cuda.device_count())
-if torch.cuda.is_available():
-    print("CUDA device name:", torch.cuda.get_device_name(0))
+from bark.generation import preload_models
 
 class VoiceCloningApp:
     def __init__(self):
@@ -26,11 +19,17 @@ class VoiceCloningApp:
         # Explicit model loading with error handling
         try:
             print("Attempting to load Bark models...")
-            preload_models(device=device)
-            print(f"Bark models loaded successfully on {device}")
+            # Remove device argument
+            preload_models()
+            print("Bark models loaded successfully.")
         except Exception as e:
             print(f"Error loading Bark models: {e}")
-            raise RuntimeError(f"Could not load Bark models: {e}")
+            # Log the full error for debugging
+            import traceback
+            traceback.print_exc()
+            
+            # Provide a more informative error message
+            raise RuntimeError(f"Could not load Bark models. Error: {e}")
 
     def process_reference_audio(self, audio_data):
         """Simple audio processing"""
@@ -63,21 +62,11 @@ class VoiceCloningApp:
             # Generate audio with explicit error handling
             print(f"Generating speech for text: {text}")
             
-            # Use alternative generation method
-            semantic_tokens = generate_text_semantic(
-                text,
-                history_prompt=None,
-                temp=0.7,
-                min_eos_p=0.05,
-                device=device
-            )
-            
-            # Generate audio from semantic tokens
+            # Simplified audio generation
             audio_array = generate_audio(
-                semantic_tokens,
+                text,
                 history_prompt=None,
-                temp=0.7,
-                device=device
+                temp=0.7
             )
             
             # Save generated audio
@@ -89,9 +78,16 @@ class VoiceCloningApp:
         
         except Exception as e:
             print(f"Speech generation error: {e}")
+            # Log the full error for debugging
+            import traceback
+            traceback.print_exc()
             return None, f"Error generating speech: {str(e)}"
 
 def create_interface():
+    # Create working directory if it doesn't exist
+    working_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "working_files")
+    os.makedirs(working_dir, exist_ok=True)
+    
     app = VoiceCloningApp()
     
     # Use the most basic Gradio theme to avoid font issues

requirements.txt

@@ -2,9 +2,6 @@ gradio>=3.50.2
 numpy
 scipy
 soundfile
-torch>=2.0.0+cu118
-torchvision>=0.15.0+cu118
-torchaudio>=2.0.0
+torch
 transformers
 git+https://github.com/suno-ai/bark.git
---extra-index-url https://download.pytorch.org/whl/cu118