app.py

from transformers import pipeline
import gradio as gr

modelo = pipeline("automatic-speech-recognition", model="facebook/wav2vec2-large-xlsr-53-spanish")
titulo = "Audio a texto"
desc = "Graba un audio (en español) y este será transcrito"

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

gr.Interface(
    transcribe,
    inputs=[gr.Audio(source="microphone", type="filepath")],
    outputs=["textbox"],
    title=titulo,
    description=desc,
).launch()