Spaces:
Running
Running
| import os | |
| import io | |
| import gradio as gr | |
| import subprocess | |
| import wave | |
| 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 get_audio_duration(file_path): | |
| """Obtiene la duración del archivo de audio en segundos.""" | |
| result = subprocess.run( | |
| ["ffprobe", "-i", file_path, "-show_entries", "format=duration", "-v", "quiet", "-of", "csv=p=0"], | |
| capture_output=True, text=True | |
| ) | |
| return float(result.stdout.strip()) | |
| def convert_to_wav(input_file): | |
| """Convierte archivos de audio a WAV LINEAR16 con la tasa de muestreo correcta.""" | |
| output_file = input_file + ".wav" | |
| command = [ | |
| "ffmpeg", "-y", "-i", input_file, | |
| "-acodec", "pcm_s16le", "-ar", "48000", "-ac", "1", output_file # 48000 Hz para evitar el error | |
| ] | |
| 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) | |
| # Obtener la duración del archivo | |
| duration = get_audio_duration(audio_file) | |
| # Configuración de la solicitud | |
| config = speech.RecognitionConfig( | |
| encoding=speech.RecognitionConfig.AudioEncoding.LINEAR16, | |
| sample_rate_hertz=48000, | |
| 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) | |
| if duration <= 60: | |
| # Si el audio dura menos de 1 minuto, usamos transcripción síncrona | |
| response = client.recognize(config=config, audio=audio) | |
| else: | |
| # Si el audio dura más de 1 minuto, mostramos un error (se necesita Google Cloud Storage) | |
| return "Error: El audio es muy largo para la transcripción síncrona. Se necesita Google Cloud Storage.", "" | |
| 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() | |