Update app.py

app.py

@@ -1,5 +1,71 @@
+import sys
+import io, os, stat
+import subprocess
+import random
+from zipfile import ZipFile
+import uuid
 import time
+import torch
+import torchaudio
+import time
+# Mantenemos la descarga de MeCab
+os.system('python -m unidic download')
+
+# Mantenemos el acuerdo de CPML
+os.environ["COQUI_TOS_AGREED"] = "1"
+
+import langid
+import base64
+import csv
+from io import StringIO
+import datetime
+import re
+
+import gradio as gr
+from scipy.io.wavfile import write
+from pydub import AudioSegment
+
+from TTS.api import TTS
+from TTS.tts.configs.xtts_config import XttsConfig
+from TTS.tts.models.xtts import Xtts
+from TTS.utils.generic_utils import get_user_data_dir
+
+HF_TOKEN = os.environ.get("HF_TOKEN")
+
+from huggingface_hub import hf_hub_download
+import os
+from TTS.utils.manage import get_user_data_dir
+
+# Mantenemos la autenticación y descarga del modelo
+repo_id = "Blakus/Pedro_Lab_XTTS"
+local_dir = os.path.join(get_user_data_dir("tts"), "tts_models--multilingual--multi-dataset--xtts_v2")
+os.makedirs(local_dir, exist_ok=True)
+files_to_download = ["config.json", "model.pth", "vocab.json"]
+for file_name in files_to_download:
+    print(f"Downloading {file_name} from {repo_id}")
+    local_file_path = os.path.join(local_dir, file_name)
+    hf_hub_download(repo_id=repo_id, filename=file_name, local_dir=local_dir)
+
+# Cargamos configuración y modelo
+config_path = os.path.join(local_dir, "config.json")
+checkpoint_path = os.path.join(local_dir, "model.pth")
+vocab_path = os.path.join(local_dir, "vocab.json")
 
+config = XttsConfig()
+config.load_json(config_path)
+
+model = Xtts.init_from_config(config)
+model.load_checkpoint(config, checkpoint_path=checkpoint_path, vocab_path=vocab_path, eval=True, use_deepspeed=False)
+
+print("Modelo cargado en CPU")
+
+# Mantenemos variables globales y funciones auxiliares
+DEVICE_ASSERT_DETECTED = 0
+DEVICE_ASSERT_PROMPT = None
+DEVICE_ASSERT_LANG = None
+supported_languages = config.languages
+
+# Función de inferencia usando parámetros predeterminados del archivo de configuración
 def predict(prompt, language, audio_file_pth, mic_file_path, use_mic):
     try:
         if use_mic:
@@ -49,4 +115,34 @@ def predict(prompt, language, audio_file_pth, mic_file_path, use_mic):
 
     except Exception as e:
         print(f"Error detallado: {str(e)}")
-        return None, None, f"Error: {str(e)}"
+        return None, None, f"Error: {str(e)}"
+
+
+# Interfaz de Gradio actualizada sin sliders
+with gr.Blocks(theme=gr.themes.Base()) as demo:
+    gr.Markdown("# Sintetizador de Voz XTTS")
+    
+    with gr.Row():
+        with gr.Column():
+            input_text = gr.Textbox(label="Texto a sintetizar", placeholder="Escribe aquí el texto que quieres convertir a voz...")
+            language = gr.Dropdown(label="Idioma", choices=supported_languages, value="es")
+            audio_file = gr.Audio(label="Audio de referencia", type="filepath")
+            use_mic = gr.Checkbox(label="Usar micrófono")
+            mic_file = gr.Audio(label="Grabar con micrófono", source="microphone", type="filepath", visible=False)
+            
+            use_mic.change(fn=lambda x: gr.update(visible=x), inputs=[use_mic], outputs=[mic_file])
+            
+            generate_button = gr.Button("Generar voz")
+        
+        with gr.Column():
+            output_audio = gr.Audio(label="Audio generado")
+            waveform = gr.Image(label="Forma de onda")
+            metrics = gr.Textbox(label="Métricas")
+    
+    generate_button.click(
+        predict,
+        inputs=[input_text, language, audio_file, mic_file, use_mic],
+        outputs=[waveform, output_audio, metrics]
+    )
+
+demo.launch(debug=True)