app.py

# interfaceV3.py
import gradio as gr
import shutil
import os
import subprocess
import sys

subprocess.run(["python", "src/setup_wav2lip.py"])

# AÑADIR RUTA DEL MÓDULO CUSTOM
sys.path.append(os.path.abspath("./src"))

from whisper_audio_transcriber import transcribe_audio, guardar_transcripcion
from call_openai_api import moni as rtff

# Rutas
AUDIO_RECORD_PATH = os.path.abspath("./assets/audio/grabacion_gradio.wav")
VIDEO_PATH = os.path.abspath("./assets/video/data_video_sun.mp4")
TRANSCRIPTION_TEXT_PATH = os.path.abspath("./results/transcripcion.txt")
RESULT_AUDIO_TEMP_PATH = os.path.abspath("./results/audiov2.wav")
RESULT_AUDIO_FINAL_PATH = os.path.abspath("./assets/audio/audio.wav")
RESULT_VIDEO_PATH = os.path.abspath("./results/result_voice.mp4")
TEXT_TO_SPEECH_PATH = os.path.abspath("./src/text_to_speech.py")
RUN_INFERENCE_PATH = os.path.abspath("./src/run_inference.py")


def transcribir_con_progreso(audio_path):
    progreso = gr.Progress()
    progreso(0, "Iniciando transcripción...")
    model_name = "openai/whisper-large"
    progreso(25, "Cargando modelo Whisper...")
    transcripcion = transcribe_audio(audio_path, model_name)
    progreso(75, "Guardando transcripción...")
    guardar_transcripcion(transcripcion, filename=TRANSCRIPTION_TEXT_PATH)
    progreso(100, "Transcripción completada.")
    return transcripcion


def generar_audio_desde_texto():
    print("Ejecutando text_to_speech...")
    result = subprocess.run(
        [sys.executable, TEXT_TO_SPEECH_PATH],
        capture_output=True,
        text=True
    )
    print("stdout:", result.stdout)
    print("stderr:", result.stderr)

    if result.returncode != 0:
        raise RuntimeError(f"Error ejecutando text_to_speech.py: {result.stderr}")

    if os.path.exists(RESULT_AUDIO_TEMP_PATH):
        os.makedirs(os.path.dirname(RESULT_AUDIO_FINAL_PATH), exist_ok=True)
        shutil.copy(RESULT_AUDIO_TEMP_PATH, RESULT_AUDIO_FINAL_PATH)
        print(f"Audio copiado a: {RESULT_AUDIO_FINAL_PATH}")
        return RESULT_AUDIO_FINAL_PATH
    else:
        print("Audio temporal no encontrado")
        return None


def procesar_video_audio():
    print("Iniciando procesamiento de video...")
    print("Audio de entrada:", RESULT_AUDIO_FINAL_PATH)
    print("Video de entrada:", VIDEO_PATH)

    result = subprocess.run(
        [sys.executable, RUN_INFERENCE_PATH, "--audio", RESULT_AUDIO_FINAL_PATH, "--video", VIDEO_PATH],
        capture_output=True,
        text=True
    )

    print("stdout:", result.stdout)
    print("stderr:", result.stderr)

    if os.path.exists(RESULT_VIDEO_PATH):
        print("Video generado:", RESULT_VIDEO_PATH)
        return RESULT_VIDEO_PATH
    else:
        print("No se generó el video")
        return None


def flujo_completo(audio_file_path):
    try:
        os.makedirs(os.path.dirname(AUDIO_RECORD_PATH), exist_ok=True)
        shutil.copy(audio_file_path, AUDIO_RECORD_PATH)
        print("Audio grabado copiado a:", AUDIO_RECORD_PATH)

        transcripcion = transcribir_con_progreso(AUDIO_RECORD_PATH)
        print("Texto transcrito:", transcripcion)

        respuesta_openai = rtff(TRANSCRIPTION_TEXT_PATH)
        print("Respuesta de OpenAI:", respuesta_openai)

        audio_generado = generar_audio_desde_texto()
        video_path = procesar_video_audio()

        return "Grabación recibida", AUDIO_RECORD_PATH, transcripcion, audio_generado, video_path

    except Exception as e:
        return (
            f"Error durante el flujo completo: {str(e)}",
            None,
            f"Error: {str(e)}",
            None,
            None
        )


def interfaz():
    with gr.Blocks() as demo:
        with gr.Row():
            with gr.Column():
                gr.Video(VIDEO_PATH, loop=True, autoplay=True, height=300, width=500)
                audio_input = gr.Audio(label="Graba tu voz", type="filepath", format="wav")
                estado_grabacion = gr.Textbox(label="Estado", interactive=False)

            with gr.Column():
                output_audio = gr.Audio(label="Audio grabado", interactive=False)
                output_audio_speech = gr.Audio(label="Audio TTS", interactive=False)
                video_resultado = gr.Video(label="Video procesado", interactive=False)
                texto_transcripcion = gr.Textbox(label="Texto transcrito")

        audio_input.change(
            flujo_completo,
            inputs=audio_input,
            outputs=[estado_grabacion, output_audio, texto_transcripcion, output_audio_speech, video_resultado]
        )

    return demo


if __name__ == "__main__":
    demo = interfaz()
    demo.launch()