app.py

import os
import io
import gradio as gr
import subprocess
from google.cloud import speech
from google.api_core.client_options import ClientOptions

# Obtener la API Key desde las variables de entorno
try:
    API_KEY = os.environ["GOOGLE_API_KEY"]
except KeyError:
    raise ValueError("La API Key de Google no está disponible. Configúrala en los Secrets como 'GOOGLE_API_KEY'.")

# Configurar cliente de Google Speech-to-Text con API Key
client_options = ClientOptions(api_key=API_KEY)
client = speech.SpeechClient(client_options=client_options)

def convert_to_wav(input_file):
    """Convierte archivos de audio a formato WAV LINEAR16 si es necesario."""
    output_file = input_file + ".wav"
    command = [
        "ffmpeg", "-y", "-i", input_file,
        "-acodec", "pcm_s16le", "-ar", "44100", "-ac", "1", output_file
    ]
    subprocess.run(command, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
    return output_file

def transcribe(audio_file=None):
    """Transcribe audio a texto usando Google Cloud Speech-to-Text."""
    if audio_file is None:
        return "No se ha seleccionado ningún archivo.", ""

    # Convertir a WAV si es necesario
    if not audio_file.endswith(".wav"):
        audio_file = convert_to_wav(audio_file)

    # Configuración de la solicitud
    config = speech.RecognitionConfig(
        encoding=speech.RecognitionConfig.AudioEncoding.LINEAR16,
        sample_rate_hertz=44100,
        audio_channel_count=1,
        language_code="es-AR",
    )

    # Cargar el audio en binario
    with io.open(audio_file, "rb") as file:
        content = file.read()
        audio = speech.RecognitionAudio(content=content)

    # Realiza la transcripción
    response = client.recognize(config=config, audio=audio)

    transcript = []
    confidence = []

    # Leer la respuesta de la API
    for result in response.results:
        confidence.append(str(result.alternatives[0].confidence))
        transcript.append(result.alternatives[0].transcript)

    return ' '.join(transcript), '\n'.join(confidence)

# Configuración de la interfaz Gradio
output1 = gr.Textbox(label='Transcripción')
output2 = gr.Textbox(label='Confianza')

demo = gr.Interface(
    fn=transcribe,
    inputs=gr.Audio(sources=["microphone", "upload"], type="filepath", label="Subir o grabar audio"),
    outputs=[output1, output2],
    title='Demo Speech-to-Text con Google Cloud',
    description='<p>Grabar o subir un archivo de audio para convertir voz a texto usando Google Cloud Speech-to-Text.</p>'
)

demo.launch()