abstractvoice 0.4.1__py3-none-any.whl → 0.5.0__py3-none-any.whl

abstractvoice/__init__.py

@@ -32,5 +32,5 @@ from .voice_manager import VoiceManager
 # Import simple APIs for third-party applications
 from .simple_model_manager import list_models, download_model, get_status, is_ready
 
-__version__ = "0.4.1"
+__version__ = "0.5.0"
 __all__ = ['VoiceManager', 'list_models', 'download_model', 'get_status', 'is_ready'] 

abstractvoice/examples/voice_cli.py

@@ -158,7 +158,7 @@ def main():
                     traceback.print_exc()
             return
         elif args.command == "download-models":
-            from abstractvoice.model_manager import download_models_cli
+            from abstractvoice.simple_model_manager import download_models_cli
             # Pass remaining arguments to download_models_cli
             import sys
             original_argv = sys.argv

abstractvoice/instant_setup.py

@@ -0,0 +1,83 @@
+"""
+Instant Setup Module for AbstractVoice
+Provides immediate TTS functionality with seamless model download.
+"""
+
+import os
+import sys
+from pathlib import Path
+
+# Essential model for instant functionality (lightweight, reliable)
+ESSENTIAL_MODEL = "tts_models/en/ljspeech/fast_pitch"
+ESSENTIAL_MODEL_SIZE = "~100MB"
+
+def ensure_instant_tts():
+    """
+    Ensure TTS is ready for immediate use.
+    Downloads essential model if needed with progress indicator.
+
+    Returns:
+        bool: True if TTS is ready, False if failed
+    """
+    try:
+        from TTS.api import TTS
+        from TTS.utils.manage import ModelManager
+
+        manager = ModelManager()
+
+        # Check if essential model is already cached
+        if is_model_cached(ESSENTIAL_MODEL):
+            return True
+
+        # Download essential model with user-friendly progress
+        print(f"🚀 AbstractVoice: Setting up TTS ({ESSENTIAL_MODEL_SIZE})...")
+        print(f"   This happens once and takes ~30 seconds")
+
+        try:
+            # Download with progress bar
+            tts = TTS(model_name=ESSENTIAL_MODEL, progress_bar=True)
+            print(f"✅ TTS ready! AbstractVoice is now fully functional.")
+            return True
+
+        except Exception as e:
+            print(f"❌ Setup failed: {e}")
+            print(f"💡 Try: pip install abstractvoice[all]")
+            return False
+
+    except ImportError as e:
+        print(f"❌ Missing dependencies: {e}")
+        print(f"💡 Install with: pip install abstractvoice[all]")
+        return False
+
+def is_model_cached(model_name):
+    """Check if a model is already cached."""
+    try:
+        from TTS.utils.manage import ModelManager
+        manager = ModelManager()
+
+        # Get cached models list
+        models_file = os.path.join(manager.output_prefix, ".models.json")
+        if os.path.exists(models_file):
+            import json
+            with open(models_file, 'r') as f:
+                cached_models = json.load(f)
+                return model_name in cached_models
+
+        # Fallback: check if model directory exists and has content
+        model_dir = model_name.replace("/", "--")
+        model_path = os.path.join(manager.output_prefix, model_dir)
+        return os.path.exists(model_path) and bool(os.listdir(model_path))
+
+    except:
+        # If anything fails, assume not cached
+        return False
+
+def get_instant_model():
+    """Get the essential model name for instant setup."""
+    return ESSENTIAL_MODEL
+
+if __name__ == "__main__":
+    # CLI test
+    print("🧪 Testing instant setup...")
+    success = ensure_instant_tts()
+    print(f"Result: {'✅ Ready' if success else '❌ Failed'}")

abstractvoice/simple_model_manager.py

@@ -31,37 +31,65 @@ class SimpleModelManager:
     """Simple, clean model manager for AbstractVoice."""
 
     # Essential model - guaranteed to work everywhere, reasonable size
-    ESSENTIAL_MODEL = "tts_models/en/ljspeech/fast_pitch"
+    # Changed from fast_pitch to tacotron2-DDC because fast_pitch downloads are failing
+    ESSENTIAL_MODEL = "tts_models/en/ljspeech/tacotron2-DDC"
 
     # Available models organized by language with metadata
     AVAILABLE_MODELS = {
         "en": {
+            "tacotron2": {
+                "model": "tts_models/en/ljspeech/tacotron2-DDC",
+                "name": "Linda (LJSpeech)",
+                "quality": "good",
+                "size_mb": 362,
+                "description": "Standard female voice (LJSpeech speaker)",
+                "requires_espeak": False,
+                "default": True
+            },
+            "jenny": {
+                "model": "tts_models/en/jenny/jenny",
+                "name": "Jenny",
+                "quality": "excellent",
+                "size_mb": 368,
+                "description": "Different female voice, clear and natural",
+                "requires_espeak": False,
+                "default": False
+            },
+            "ek1": {
+                "model": "tts_models/en/ek1/tacotron2",
+                "name": "Edward (EK1)",
+                "quality": "excellent",
+                "size_mb": 310,
+                "description": "Male voice with British accent",
+                "requires_espeak": False,
+                "default": False
+            },
+            "sam": {
+                "model": "tts_models/en/sam/tacotron-DDC",
+                "name": "Sam",
+                "quality": "good",
+                "size_mb": 370,
+                "description": "Different male voice, deeper tone",
+                "requires_espeak": False,
+                "default": False
+            },
             "fast_pitch": {
                 "model": "tts_models/en/ljspeech/fast_pitch",
-                "name": "Fast Pitch (English)",
+                "name": "Linda Fast (LJSpeech)",
                 "quality": "good",
                 "size_mb": 107,
-                "description": "Lightweight, reliable English voice",
+                "description": "Same speaker as Linda but faster engine",
                 "requires_espeak": False,
-                "default": True
+                "default": False
             },
             "vits": {
                 "model": "tts_models/en/ljspeech/vits",
-                "name": "VITS (English)",
+                "name": "Linda Premium (LJSpeech)",
                 "quality": "excellent",
                 "size_mb": 328,
-                "description": "High-quality English voice with natural prosody",
+                "description": "Same speaker as Linda but premium quality",
                 "requires_espeak": True,
                 "default": False
-            },
-            "tacotron2": {
-                "model": "tts_models/en/ljspeech/tacotron2-DDC",
-                "name": "Tacotron2 (English)",
-                "quality": "good",
-                "size_mb": 362,
-                "description": "Classic English voice, reliable",
-                "requires_espeak": False,
-                "default": False
             }
         },
         "fr": {
@@ -184,7 +212,7 @@ class SimpleModelManager:
             return False
 
     def download_model(self, model_name: str, progress_callback: Optional[Callable[[str, bool], None]] = None) -> bool:
-        """Download a specific model.
+        """Download a specific model with improved error handling.
 
         Args:
             model_name: TTS model name (e.g., 'tts_models/en/ljspeech/fast_pitch')
@@ -203,25 +231,56 @@ class SimpleModelManager:
         try:
             TTS, _ = _import_tts()
 
-            if self.debug_mode:
-                print(f"📥 Downloading {model_name}...")
+            print(f"📥 Downloading {model_name}...")
+            print(f"   This may take a few minutes depending on your connection...")
 
             start_time = time.time()
 
             # Initialize TTS to trigger download
-            tts = TTS(model_name=model_name, progress_bar=True)
+            # Set gpu=False to avoid CUDA errors on systems without GPU
+            try:
+                tts = TTS(model_name=model_name, progress_bar=True, gpu=False)
+
+                # Verify the model actually downloaded
+                if not self.is_model_cached(model_name):
+                    print(f"⚠️ Model download completed but not found in cache")
+                    return False
+
+            except Exception as init_error:
+                # Try alternative download method
+                error_msg = str(init_error).lower()
+                if "connection" in error_msg or "timeout" in error_msg:
+                    print(f"❌ Network error: Check your internet connection")
+                elif "not found" in error_msg:
+                    print(f"❌ Model '{model_name}' not found in registry")
+                else:
+                    print(f"❌ Download error: {init_error}")
+                raise
 
             download_time = time.time() - start_time
-            if self.debug_mode:
-                print(f"✅ Downloaded {model_name} in {download_time:.1f}s")
+            print(f"✅ Downloaded {model_name} in {download_time:.1f}s")
 
             if progress_callback:
                 progress_callback(model_name, True)
             return True
 
         except Exception as e:
-            if self.debug_mode:
-                print(f"❌ Failed to download {model_name}: {e}")
+            error_msg = str(e).lower()
+
+            # Provide helpful error messages
+            if "connection" in error_msg or "timeout" in error_msg:
+                print(f"❌ Failed to download {model_name}: Network issue")
+                print(f"   Check your internet connection and try again")
+            elif "permission" in error_msg:
+                print(f"❌ Failed to download {model_name}: Permission denied")
+                print(f"   Check write permissions for cache directory")
+            elif "space" in error_msg:
+                print(f"❌ Failed to download {model_name}: Insufficient disk space")
+            else:
+                print(f"❌ Failed to download {model_name}")
+                if self.debug_mode:
+                    print(f"   Error: {e}")
+
             if progress_callback:
                 progress_callback(model_name, False)
             return False
@@ -395,4 +454,86 @@ def get_status() -> str:
 def is_ready() -> bool:
     """Check if essential model is ready for immediate use."""
     manager = get_model_manager()
-    return manager.is_model_cached(manager.ESSENTIAL_MODEL)
+    return manager.is_model_cached(manager.ESSENTIAL_MODEL)
+
+
+def download_models_cli():
+    """Simple CLI entry point for downloading models."""
+    import argparse
+    import sys
+
+    parser = argparse.ArgumentParser(description="Download TTS models for offline use")
+    parser.add_argument("--essential", action="store_true",
+                       help="Download essential model (default)")
+    parser.add_argument("--all", action="store_true",
+                       help="Download all available models")
+    parser.add_argument("--model", type=str,
+                       help="Download specific model by name")
+    parser.add_argument("--language", type=str,
+                       help="Download models for specific language (en, fr, es, de, it)")
+    parser.add_argument("--status", action="store_true",
+                       help="Show current cache status")
+    parser.add_argument("--clear", action="store_true",
+                       help="Clear model cache")
+
+    args = parser.parse_args()
+
+    manager = get_model_manager(debug_mode=True)
+
+    if args.status:
+        print(get_status())
+        return
+
+    if args.clear:
+        # Ask for confirmation
+        response = input("⚠️ This will delete all downloaded TTS models. Continue? (y/N): ")
+        if response.lower() == 'y':
+            success = manager.clear_cache(confirm=True)
+            if success:
+                print("✅ Model cache cleared")
+            else:
+                print("❌ Failed to clear cache")
+        else:
+            print("Cancelled")
+        return
+
+    if args.model:
+        success = download_model(args.model)
+        if success:
+            print(f"✅ Downloaded {args.model}")
+        else:
+            print(f"❌ Failed to download {args.model}")
+        sys.exit(0 if success else 1)
+
+    if args.language:
+        # Language-specific downloads using our simple API
+        lang_models = {
+            'en': ['en.tacotron2', 'en.jenny', 'en.ek1'],
+            'fr': ['fr.css10_vits', 'fr.mai_tacotron2'],
+            'es': ['es.mai_tacotron2'],
+            'de': ['de.thorsten_vits'],
+            'it': ['it.mai_male_vits', 'it.mai_female_vits']
+        }
+
+        if args.language not in lang_models:
+            print(f"❌ Language '{args.language}' not supported")
+            print(f"   Available: {list(lang_models.keys())}")
+            sys.exit(1)
+
+        success = False
+        for model_id in lang_models[args.language]:
+            if download_model(model_id):
+                print(f"✅ Downloaded {model_id}")
+                success = True
+                break
+
+        sys.exit(0 if success else 1)
+
+    # Default: download essential model
+    print("📦 Downloading essential TTS model...")
+    success = download_model(manager.ESSENTIAL_MODEL)
+    if success:
+        print("✅ Essential model ready!")
+    else:
+        print("❌ Failed to download essential model")
+    sys.exit(0 if success else 1)

abstractvoice/tts/tts_engine.py

@@ -300,11 +300,24 @@ class NonBlockingAudioPlayer:
                     print(f"Error stopping audio stream: {e}")
             finally:
                 self.stream = None
-        
+
         self.is_playing = False
         with self.pause_lock:
             self.is_paused = False
         self.clear_queue()
+
+    def cleanup(self):
+        """Cleanup resources to prevent memory conflicts."""
+        try:
+            self.stop_stream()
+            # Clear any remaining references
+            self.current_audio = None
+            self.playback_complete_callback = None
+            if self.debug_mode:
+                print(" > Audio player cleaned up")
+        except Exception as e:
+            if self.debug_mode:
+                print(f"Audio cleanup warning: {e}")
     
     def play_audio(self, audio_array):
         """Add audio to the playback queue."""
@@ -509,58 +522,155 @@ class TTSEngine:
         self.is_paused_state = False  # Explicit paused state tracking
 
     def _load_with_simple_fallback(self, TTS, preferred_model: str, debug_mode: bool) -> tuple[bool, str]:
-        """Load TTS model with simple, effective strategy."""
+        """Load TTS model with bulletproof compatibility-first strategy."""
         from ..simple_model_manager import get_model_manager
 
         model_manager = get_model_manager(debug_mode=debug_mode)
 
-        # Strategy 1: Try preferred model if cached
-        if model_manager.is_model_cached(preferred_model):
+        # Step 1: Check espeak availability for smart model filtering
+        espeak_available = self._check_espeak_available()
+        if debug_mode and not espeak_available:
+            print(" > espeak-ng not found, will skip VITS models")
+
+        # Step 2: Try the REQUESTED model first if it's cached
+        cached_models = model_manager.get_cached_models()
+        if cached_models and debug_mode:
+            print(f" > Found {len(cached_models)} cached models")
+
+        # FORCE USER'S CHOICE: Try the specifically requested model first
+        if preferred_model in cached_models:
             try:
                 if debug_mode:
-                    print(f" > Using cached model: {preferred_model}")
+                    print(f" > LOADING REQUESTED MODEL: {preferred_model}")
+
+                # Safety check for Italian VITS models that might crash
+                if "it/" in preferred_model and "vits" in preferred_model:
+                    if debug_mode:
+                        print(f" > Italian VITS model detected - using safe loading...")
+
                 self.tts = TTS(model_name=preferred_model, progress_bar=self.debug_mode)
+
+                if debug_mode:
+                    print(f" > ✅ SUCCESS: Loaded requested model: {preferred_model}")
                 return True, preferred_model
+
             except Exception as e:
+                error_msg = str(e).lower()
                 if debug_mode:
-                    print(f" > Cached model failed: {e}")
+                    print(f" > ❌ Requested model failed: {e}")
+
+                # Special handling for Italian model crashes
+                if "it/" in preferred_model and ("segmentation" in error_msg or "crash" in error_msg):
+                    if debug_mode:
+                        print(f" > Italian model caused crash - marking as incompatible")
+                    # Force fallback for crashed Italian models
+                    pass
+
+                # Only fall back if the model actually failed to load, not due to dependencies
+
+        # Step 3: Only fall back to compatibility order if requested model failed
+        if debug_mode:
+            print(" > Requested model unavailable, trying fallback models...")
+
+        # Compatibility-first fallback order
+        fallback_models = [
+            "tts_models/en/ljspeech/tacotron2-DDC",  # Most reliable (Linda)
+            "tts_models/en/jenny/jenny",             # Different female speaker (Jenny)
+            "tts_models/en/ek1/tacotron2",           # Male British accent (Edward)
+            "tts_models/en/sam/tacotron-DDC",        # Different male voice (Sam)
+            "tts_models/en/ljspeech/fast_pitch",     # Lightweight alternative
+            "tts_models/en/ljspeech/glow-tts",       # Another alternative
+            "tts_models/en/vctk/vits",               # Multi-speaker (requires espeak)
+            "tts_models/en/ljspeech/vits",           # Premium (requires espeak)
+        ]
+
+        # Remove the preferred model from fallbacks to avoid duplicate attempts
+        fallback_models = [m for m in fallback_models if m != preferred_model]
 
-        # Strategy 2: Try essential model if cached
-        essential_model = model_manager.ESSENTIAL_MODEL
-        if essential_model != preferred_model and model_manager.is_model_cached(essential_model):
+        # Try fallback models
+        for model in fallback_models:
+            if model in cached_models:
+                # Skip VITS models if no espeak
+                if "vits" in model and not espeak_available:
+                    if debug_mode:
+                        print(f" > Skipping {model} (requires espeak-ng)")
+                    continue
+
+                try:
+                    if debug_mode:
+                        print(f" > Trying fallback model: {model}")
+                    self.tts = TTS(model_name=model, progress_bar=self.debug_mode)
+                    if debug_mode:
+                        print(f" > ✅ Successfully loaded fallback: {model}")
+                    return True, model
+                except Exception as e:
+                    if debug_mode:
+                        print(f" > ❌ Fallback {model} failed: {e}")
+
+        # Step 4: If no cached models work, try downloading requested model first
+        if debug_mode:
+            print(" > No cached models worked, attempting downloads...")
+
+        # Try downloading the requested model first
+        if "vits" not in preferred_model or espeak_available:
             try:
                 if debug_mode:
-                    print(f" > Using cached essential model: {essential_model}")
-                self.tts = TTS(model_name=essential_model, progress_bar=self.debug_mode)
-                return True, essential_model
+                    print(f" > Downloading requested model: {preferred_model}...")
+                success = model_manager.download_model(preferred_model)
+                if success:
+                    self.tts = TTS(model_name=preferred_model, progress_bar=self.debug_mode)
+                    if debug_mode:
+                        print(f" > ✅ Downloaded and loaded requested: {preferred_model}")
+                    return True, preferred_model
+                elif debug_mode:
+                    print(f" > ❌ Download failed for requested model: {preferred_model}")
             except Exception as e:
                 if debug_mode:
-                    print(f" > Essential model failed: {e}")
+                    print(f" > ❌ Failed to download/load requested model: {e}")
 
-        # Strategy 3: Download essential model (guaranteed to work)
-        try:
-            if debug_mode:
-                print(f" > Downloading essential model: {essential_model}")
-            success = model_manager.download_model(essential_model)
-            if success:
-                self.tts = TTS(model_name=essential_model, progress_bar=self.debug_mode)
-                return True, essential_model
-        except Exception as e:
-            if debug_mode:
-                print(f" > Essential model download failed: {e}")
+        # Step 5: If requested model download failed, try fallback downloads
+        for model in fallback_models:
+            # Skip VITS models if no espeak
+            if "vits" in model and not espeak_available:
+                continue
 
-        # Strategy 4: Try downloading preferred model
-        try:
-            if debug_mode:
-                print(f" > Attempting preferred model download: {preferred_model}")
-            self.tts = TTS(model_name=preferred_model, progress_bar=self.debug_mode)
-            return True, preferred_model
-        except Exception as e:
-            if debug_mode:
-                print(f" > Preferred model download failed: {e}")
+            try:
+                if debug_mode:
+                    print(f" > Downloading fallback: {model}...")
+
+                # First try to download
+                success = model_manager.download_model(model)
+                if success:
+                    # Then try to load
+                    self.tts = TTS(model_name=model, progress_bar=self.debug_mode)
+                    if debug_mode:
+                        print(f" > ✅ Downloaded and loaded fallback: {model}")
+                    return True, model
+                elif debug_mode:
+                    print(f" > ❌ Download failed for {model}")
+
+            except Exception as e:
+                if debug_mode:
+                    print(f" > ❌ Failed to load {model}: {e}")
 
         return False, None
 
+    def _check_espeak_available(self) -> bool:
+        """Check if espeak-ng is available on the system."""
+        import subprocess
+        try:
+            subprocess.run(['espeak-ng', '--version'],
+                         capture_output=True, check=True, timeout=5)
+            return True
+        except (subprocess.CalledProcessError, FileNotFoundError, subprocess.TimeoutExpired):
+            # Try alternative espeak command
+            try:
+                subprocess.run(['espeak', '--version'],
+                             capture_output=True, check=True, timeout=5)
+                return True
+            except:
+                return False
+
     def _handle_espeak_fallback(self, debug_mode: bool):
         """Handle espeak-related errors with fallback to non-phoneme models."""
         # Restore stdout to show user-friendly message
@@ -574,7 +684,7 @@ class TTSEngine:
         print("  • macOS:   brew install espeak-ng")
         print("  • Linux:   sudo apt-get install espeak-ng")
         print("  • Windows: conda install espeak-ng  (or see README)")
-        print("\nFalling back to fast_pitch (no espeak dependency)")
+        print("\nFalling back to compatible models (no espeak dependency)")
         print("="*70 + "\n")
 
         if not debug_mode:
@@ -582,12 +692,15 @@ class TTSEngine:
             null_out = open(os.devnull, 'w')
             sys.stdout = null_out
 
-        # Try non-phoneme models that don't require espeak
+        # Try non-phoneme models that don't require espeak (compatibility-first order)
         from TTS.api import TTS
         fallback_models = [
-            "tts_models/en/ljspeech/fast_pitch",
-            "tts_models/en/ljspeech/tacotron2-DDC",
-            "tts_models/en/ljspeech/glow-tts"
+            "tts_models/en/ljspeech/tacotron2-DDC",  # Most reliable (Linda)
+            "tts_models/en/jenny/jenny",             # Different female speaker (Jenny)
+            "tts_models/en/ek1/tacotron2",           # Male British accent (Edward)
+            "tts_models/en/sam/tacotron-DDC",        # Different male voice (Sam)
+            "tts_models/en/ljspeech/fast_pitch",     # Lightweight alternative
+            "tts_models/en/ljspeech/glow-tts"        # Another alternative
         ]
 
         tts_loaded = False