from transformers import pipeline
import gradio as gr

modelo = pipeline("automatic-speech-recognition", model="facebook/wav2vec2-large-xlsr-53-spanish")

def transcripcion(audio):
  text=modelo(audio)["text"]
  return text

gr.Interface(fn= transcripcion, inputs=[gr.Audio(source="microphone", type="filepath")], outputs=["textbox"]).launch()