Create app.py

app.py

@@ -0,0 +1,37 @@
+import torch
+import gradio as gr
+import soundfile as sf
+import tempfile
+
+from transformers import SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan
+from datasets import load_dataset
+
+# Carga de modelos y recursos
+processor = SpeechT5Processor.from_pretrained("microsoft/speecht5_tts")
+model = SpeechT5ForTextToSpeech.from_pretrained("microsoft/speecht5_tts")
+vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
+
+# Carga de embeddings de ejemplo (puedes permitir que el usuario elija un speaker también si quieres)
+embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
+speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0)
+
+# Función principal para generar voz
+def tts(text):
+    inputs = processor(text=text, return_tensors="pt")
+    with torch.no_grad():
+        speech = model.generate_speech(inputs["input_ids"], speaker_embeddings, vocoder=vocoder)
+    # Guardar a archivo temporal
+    with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as f:
+        sf.write(f.name, speech.numpy(), samplerate=16000)
+        return f.name
+
+# Interfaz con Gradio
+interface = gr.Interface(
+    fn=tts,
+    inputs=gr.Textbox(lines=2, placeholder="Introduce un texto..."),
+    outputs=gr.Audio(type="filepath"),
+    title="SpeechT5 TTS - Hugging Face Space",
+    description="Convierte texto a voz con el modelo SpeechT5 de Microsoft"
+)
+
+interface.launch()