Update app.py

app.py

@@ -1,23 +1,25 @@
 import gradio as gr
-from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
 import torch
 import scipy.io.wavfile
 import numpy as np
 
 def text_to_speech(text):
-    # Cargar el modelo y el procesador
-    model_name = "facebook/mms-tts-spa"
-    processor = AutoProcessor.from_pretrained(model_name)
-    model = AutoModelForSpeechSeq2Seq.from_pretrained(model_name)
+    # Cargar el modelo y el tokenizador
+    model_name = "facebook/tts_transformer-es-css10"
+    processor = AutoTokenizer.from_pretrained(model_name)
+    model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
 
     inputs = processor(text, return_tensors="pt")
     with torch.no_grad():
-        speech = model.generate(**inputs)
+        generated_ids = model.generate(inputs.input_ids)
     
-    waveform = processor.decode(speech.cpu().numpy(), skip_special_tokens=True)
+    generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
     
-    output_path = "/tmp/output.wav"
-    scipy.io.wavfile.write(output_path, model.config.sampling_rate, np.array(waveform))
+    # Aquí podrías generar un archivo de audio a partir del texto generado si tuvieras una función para ello
+    output_path = "/tmp/output.txt"
+    with open(output_path, "w") as f:
+        f.write(generated_text)
     
     return output_path, output_path
 
@@ -25,9 +27,9 @@ def text_to_speech(text):
 iface = gr.Interface(
     fn=text_to_speech,
     inputs="text",
-    outputs=[gr.Audio(type="filepath"), gr.File()],
+    outputs=[gr.Textbox(), gr.File()],
     title="Spanish Text-to-Speech",
-    description="Convert text to speech in Spanish using the facebook/mms-tts-spa model."
+    description="Convert text to speech in Spanish using the facebook/tts_transformer-es-css10 model."
 )
 
 iface.launch()