Update scripts/download_models.py

scripts/download_models.py

@@ -356,107 +356,10 @@ async def install_argos_packages():
 
 
 
-async def download_piper_models():
-    """Download Piper TTS voice models for Indian languages."""
-    print("\n" + "="*60)
-    print("Piper TTS Voice Models")
-    print("="*60)
-    
-    # Always use /tmp path to avoid permission issues in Docker
-    piper_dir = Path("/tmp") / ".local" / "share" / "piper" / "voices"
-    piper_dir.mkdir(parents=True, exist_ok=True)
-    print(f"Using Piper models path: {piper_dir}")
-    
-    # Piper voice models to download  
-    # Using direct GitHub release URLs for reliability
-    piper_models = {
-        "hi_IN": {
-            "display_name": "Hindi",
-            "model": "hi_IN-swarajya-medium",
-            "url": "https://github.com/rhasspy/piper/releases/download/2023.11.14-2/voice-hi-in-swarajya-medium.tar.gz",
-            "config_url": None  # Config included in tar.gz
-        },
-        "te_IN": {
-            "display_name": "Telugu",
-            "model": "te_IN-multi-medium",
-            "url": "https://github.com/rhasspy/piper/releases/download/2023.11.14-2/voice-te-in-multi-medium.tar.gz",
-            "config_url": None  # Config included in tar.gz
-        },
-        "ta_IN": {
-            "display_name": "Tamil",
-            "model": "ta_IN-multi-medium",
-            "url": "https://github.com/rhasspy/piper/releases/download/2023.11.14-2/voice-ta-in-multi-medium.tar.gz",
-            "config_url": None  # Config included in tar.gz
-        },
-    }
-    
-    # Prepare download list
-    downloads = []
-    skipped = []
-    
-    for lang_code, model_info in piper_models.items():
-        model_dir = piper_dir / model_info['model']
-        model_dir.mkdir(parents=True, exist_ok=True)
-        
-        model_file = model_dir / f"{model_info['model']}.onnx"
-        config_file = model_dir / f"{model_info['model']}.onnx.json"
-        
-        if model_file.exists() and config_file.exists():
-            skipped.append(model_info['display_name'])
-            continue
-        
-        # Add tar.gz download
-        if not model_file.exists():
-            tar_path = piper_dir / f"{model_info['model']}.tar.gz"
-            downloads.append((
-                model_info['url'],
-                tar_path,
-                f"Piper {model_info['display_name']}"
-            ))
-    
-    if skipped:
-        print(f"\nSkipping existing models: {', '.join(skipped)}")
-    
-    if not downloads:
-        print("\n✓ All Piper models already downloaded!")
-        return
-    
-    print(f"\nDownloading {len(downloads)} files...")
-    
-    # Download all models
-    async with ParallelDownloader(max_concurrent=3) as downloader:
-        results = await downloader.download_multiple(downloads)
-    
-    # Extract tar.gz files
-    print("\n\nExtracting models...")
-    import tarfile
-    
-    for lang_code, model_info in piper_models.items():
-        tar_path = piper_dir / f"{model_info['model']}.tar.gz"
-        model_dir = piper_dir / model_info['model']
-        
-        if tar_path.exists():
-            try:
-                print(f"  Extracting {model_info['display_name']}...")
-                with tarfile.open(tar_path, 'r:gz') as tar:
-                    tar.extractall(model_dir)
-                
-                # Clean up tar file
-                tar_path.unlink()
-                print(f"  ✓ {model_info['display_name']} extracted")
-            except Exception as e:
-                print(f"  ✗ {model_info['display_name']}: {str(e)[:50]}")
-    
-    print("\n✓ Piper model download complete!")
-    print("\n📋 Downloaded Piper voices for:")
-    for lang_code, model_info in piper_models.items():
-        print(f"  • {model_info['display_name']} ({lang_code})")
-
-
 def setup_tts_models():
-    """Setup TTS models (Coqui TTS for English)."""
+    """Setup TTS models (Coqui TTS downloads on first use)."""
     print("\n" + "="*60)
-    print("Coqui TTS Model Setup")
+    print("TTS Model Setup")
     print("="*60)
     
     # Create TTS directory
@@ -473,17 +376,23 @@ def setup_tts_models():
         print("  Or: pip install -r requirements.txt\n")
         return
     
-    print("\n📋 Coqui TTS Languages:")
-    print("  • English (en) - LJSpeech Tacotron2-DDC")
-    print("  • European languages (es, fr, de, it, pt, etc.)")
-    
-    print("\n📋 Piper TTS Languages (12 Indian languages):")
-    print("  • Hindi, Bengali, Tamil, Telugu")
-    print("  • Kannada, Malayalam, Marathi, Gujarati")
-    print("  • Punjabi, Urdu, Assamese, Odia")
-    
-    print("\n✓ Coqui TTS models auto-download on first use")
-    print("✓ Piper TTS models downloaded via download_piper_models()")
+    print("\n📋 Supported TTS Languages:")
+    print("  • English (en) - XTTS v2")
+    print("  • Hindi (hi), Bengali (bn), Telugu (te), Tamil (ta)")
+    print("  • Marathi (mr), Gujarati (gu), Kannada (kn), Malayalam (ml)")
+    print("  • Punjabi (pa), Urdu (ur), Assamese (as), Odia (or)")
+    print("  • Plus 15 more European/Asian languages via XTTS v2")
+    
+    print("\n✓ TTS models auto-download on first use (no pre-download needed)")
+    
+    # Optional: Pre-download commonly used models
+    print("\n💡 To pre-download models, use Python:")
+    print("  from TTS.api import TTS")
+    print("  # For English/European languages:")
+    print("  TTS('tts_models/multilingual/multi-dataset/xtts_v2')")
+    print("  # For Indian languages (example):")
+    print("  TTS('tts_models/hin/fairseq/vits')  # Hindi")
+    print("  TTS('tts_models/ben/fairseq/vits')  # Bengali")
     print()
 
 
@@ -508,7 +417,6 @@ async def main():
     try:
         await download_vosk_models()
         await install_argos_packages()
-        await download_piper_models()
         setup_tts_models()
         
         elapsed = time.time() - start_time
@@ -527,10 +435,10 @@ async def main():
         print("    ✓ English ↔ Hindi")
         print("    ✓ English ↔ Bengali")
         print("\n  TTS (Text-to-Speech):  13 languages")
-        print("    ✓ English (Coqui LJSpeech Tacotron2)")
-        print("    ✓ Hindi, Bengali, Telugu, Tamil (Piper TTS)")
-        print("    ✓ Marathi, Gujarati, Kannada, Malayalam (Piper TTS)")
-        print("    ✓ Punjabi, Urdu, Assamese, Odia (Piper TTS)")
+        print("    ✓ English")
+        print("    ✓ Hindi, Bengali, Telugu, Tamil")
+        print("    ✓ Marathi, Gujarati, Kannada, Malayalam")
+        print("    ✓ Punjabi, Urdu, Assamese, Odia")
         
         print("\n" + "="*70)
         print("\n🚀 Ready to Start!")