Add CosyVoice2-0.5B test scripts and download script

compare_models_english.py

@@ -0,0 +1,32 @@
+#\!/usr/bin/env python3
+import os
+import time
+
+print("=== Comparação: CosyVoice-300M vs CosyVoice2-0.5B ===")
+
+# Texto de teste em inglês
+test_text = "Hello\! This is a comparison test between CosyVoice models. We are testing English pronunciation quality."
+
+# Teste com modelo atual (300M)
+print("\n1. Testando modelo CosyVoice-300M-direct...")
+try:
+    from cosyvoice.cli.cosyvoice import CosyVoice
+    import torchaudio
+    
+    model1 = CosyVoice('pretrained_models/CosyVoice-300M-direct', load_jit=False, load_trt=False, fp16=False)
+    
+    start = time.time()
+    for i, j in enumerate(model1.inference_zero_shot(test_text, "Testing speech synthesis.", None, stream=False)):
+        torchaudio.save('english_300m.wav', j['tts_speech'], model1.sample_rate)
+        break
+    print(f"✅ Gerado em {time.time()-start:.1f}s - Arquivo: english_300m.wav")
+    
+except Exception as e:
+    print(f"❌ Erro no modelo 300M: {e}")
+
+# Teste com novo modelo (0.5B) - quando estiver disponível
+print("\n2. Modelo CosyVoice2-0.5B...")
+if os.path.exists('pretrained_models/CosyVoice2-0.5B/llm.pt'):
+    print("✅ Modelo baixado\! Pronto para testar.")
+else:
+    print("⏳ Ainda baixando... Execute novamente após o download.")

download_cosyvoice2.py

@@ -0,0 +1,8 @@
+from modelscope import snapshot_download
+import os
+
+print("Baixando CosyVoice2-0.5B (modelo mais recente)...")
+snapshot_download('iic/CosyVoice2-0.5B', 
+                  local_dir='pretrained_models/CosyVoice2-0.5B',
+                  cache_dir='./model_cache')
+print("Download concluído\!")

test_cosyvoice2_english.py

@@ -0,0 +1,52 @@
+#\!/usr/bin/env python3
+import os
+import sys
+import time
+
+print("=== CosyVoice2-0.5B English Test ===")
+
+# Configurar ambiente
+os.environ['HF_ENDPOINT'] = 'https://hf-mirror.com'
+
+try:
+    from cosyvoice.cli.cosyvoice import CosyVoice
+    import torchaudio
+    
+    # Usar o novo modelo CosyVoice2-0.5B
+    model_path = 'pretrained_models/CosyVoice2-0.5B'
+    
+    # Verificar se o modelo existe
+    if not os.path.exists(model_path):
+        print(f"❌ Modelo ainda não baixado em {model_path}")
+        print("Aguarde o download terminar...")
+        sys.exit(1)
+    
+    print("Carregando CosyVoice2-0.5B...")
+    start = time.time()
+    cosyvoice = CosyVoice(model_path, load_jit=False, load_trt=False, fp16=False)
+    print(f"✅ Modelo carregado em {time.time()-start:.1f}s")
+    
+    # Teste em inglês
+    text = "Hello\! This is CosyVoice version two point five B. The new model has better streaming performance and improved pronunciation for English text to speech synthesis."
+    prompt_text = "Welcome to the demonstration of our advanced speech synthesis system."
+    
+    print(f"\nTexto: {text}")
+    print("Gerando áudio...")
+    
+    start = time.time()
+    output_file = "cosyvoice2_english_test.wav"
+    for i, j in enumerate(cosyvoice.inference_zero_shot(text, prompt_text, None, stream=False)):
+        torchaudio.save(output_file, j['tts_speech'], cosyvoice.sample_rate)
+        break
+    
+    duration = time.time() - start
+    print(f"✅ Áudio gerado em {duration:.1f}s")
+    
+    if os.path.exists(output_file):
+        size = os.path.getsize(output_file) / 1024
+        print(f"📊 Arquivo: {output_file} ({size:.1f} KB)")
+        
+except Exception as e:
+    print(f"❌ Erro: {e}")
+    import traceback
+    traceback.print_exc()