app.py

import gradio as gr
import edge_tts
import asyncio
import os
import logging
import torch
from transformers import pipeline, set_seed
from moviepy.editor import *
from dotenv import load_dotenv

# Configurar logs visibles en Hugging Face
logging.basicConfig(level=logging.INFO, format="%(asctime)s [%(levelname)s] %(message)s")

# Cargar variables de entorno si hay .env
load_dotenv()

# Verificar si CUDA está disponible
device = 0 if torch.cuda.is_available() else -1

# Inicializar generador de texto
generator = pipeline("text-generation", model="gpt2", device=device)
set_seed(42)

# Asíncrono: convertir texto a voz con edge-tts
async def text_to_speech(text, output_path, voice="es-MX-DaliaNeural"):
    tts = edge_tts.Communicate(text=text, voice=voice)
    await tts.save(output_path)

def generate_video(prompt, background_music_path="musica.mp3"):
    logging.info("🚀 Generando guion con IA...")
    result = generator(prompt, max_length=500, do_sample=True, truncation=True)
    script = result[0]['generated_text']
    logging.info("🗣 Guion generado.")

    # Guardar guion a texto plano
    with open("guion.txt", "w") as f:
        f.write(script)

    # Convertir texto a voz (bloqueo controlado)
    output_audio = "voz.mp3"
    try:
        asyncio.run(text_to_speech(script, output_audio))
        logging.info("🎤 Voz generada.")
    except Exception as e:
        logging.error(f"❌ Error generando voz: {e}")
        return None, script

    # Cargar clip de voz
    voice_clip = AudioFileClip(output_audio)
    duration = voice_clip.duration

    # Video negro (fondo) + voz
    video = ColorClip(size=(1280, 720), color=(0, 0, 0), duration=duration)

    # Música en loop si es más corta que la voz
    if os.path.exists(background_music_path):
        music = AudioFileClip(background_music_path)
        if music.duration < duration:
            loops = int(duration // music.duration) + 1
            music = concatenate_audioclips([music] * loops)
        music = music.subclip(0, duration)
        final_audio = CompositeAudioClip([music.volumex(0.2), voice_clip])
    else:
        final_audio = voice_clip

    video = video.set_audio(final_audio)
    output_path = "video_generado.mp4"
    video.write_videofile(output_path, fps=24, codec="libx264", audio_codec="aac")

    return output_path, script

# Interfaz de Gradio
with gr.Blocks() as app:
    gr.Markdown("# 🎬 Generador de video IA + Voz + Música")
    prompt = gr.Textbox(label="Prompt del guion")
    boton = gr.Button("Generar video")
    salida_video = gr.Video()
    salida_texto = gr.Textbox(label="Guion generado")

    def ejecutar(prompt):
        video, script = generate_video(prompt)
        return video, script

    boton.click(ejecutar, inputs=prompt, outputs=[salida_video, salida_texto])

# Lanzar app
app.launch(debug=True)