app.py

import tempfile
import logging
import os
import asyncio
import gc
import psutil
from moviepy.editor import *
import edge_tts
import gradio as gr
from pydub import AudioSegment

# Configuración de Logs
logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")

# CONSTANTES DE ARCHIVOS
INTRO_VIDEO = "introvideo.mp4"
OUTRO_VIDEO = "outrovideo.mp4"
MUSIC_BG = "musicafondo.mp3"
EJEMPLO_VIDEO = "ejemplo.mp4"

# CONSTANTES DE LIMITACIONES
MAX_VIDEO_SIZE = 200 * 1024 * 1024  # Tamaño máximo en bytes (200MB)

# Configuración de chunks
SEGMENT_DURATION = 30  # Duración exacta entre transiciones (sin overlap)
TRANSITION_DURATION = 1.5  # Duración del efecto slide
PROCESSING_CHUNK = 120  # Procesar en bloques de 2 minutos para optimizar memoria

# Validar existencia de archivos
for file in [INTRO_VIDEO, OUTRO_VIDEO, MUSIC_BG, EJEMPLO_VIDEO]:
    if not os.path.exists(file):
        logging.error(f"Falta archivo necesario: {file}")
        raise FileNotFoundError(f"Falta: {file}")

def mostrar_uso_memoria():
    proceso = psutil.Process(os.getpid())
    memoria_uso = proceso.memory_info().rss / 1024 / 1024
    logging.info(f"Uso de memoria: {memoria_uso:.2f} MB")

def eliminar_archivo_tiempo(ruta, delay=3600):
    def eliminar():
        try:
            if os.path.exists(ruta):
                os.remove(ruta)
                logging.info(f"Archivo eliminado: {ruta}")
        except Exception as e:
            logging.error(f"Error al eliminar {ruta}: {e}")
    from threading import Timer
    Timer(delay, eliminar).start()

def validar_video(video_path):
    try:
        # Comprobar tamaño del archivo
        file_size = os.path.getsize(video_path)
        if file_size > MAX_VIDEO_SIZE:
            logging.warning(f"El video excede el tamaño máximo: {file_size/1024/1024:.2f}MB > {MAX_VIDEO_SIZE/1024/1024}MB")
            return False
        
        # Validar que es un video
        clip = VideoFileClip(video_path)
        duracion = clip.duration
        clip.close()
            
        return True
    except Exception as e:
        logging.error(f"El video no es válido: {e}")
        return False

def convertir_video(video_path):
    try:
        with tempfile.NamedTemporaryFile(delete=False, suffix=".mp4") as tmp_converted:
            output_path = tmp_converted.name
        
        # Convertir a un formato más eficiente pero manteniendo la resolución original
        os.system(f'ffmpeg -i "{video_path}" -c:v libx264 -crf 28 -preset ultrafast -c:a aac -b:a 96k "{output_path}" -y')
        
        # Comprobar si ahora cumple las limitaciones de tamaño
        if not validar_video(output_path):
            # Si sigue sin cumplir, aumentar la compresión pero sin cambiar la resolución
            os.system(f'ffmpeg -i "{output_path}" -c:v libx264 -crf 32 -preset ultrafast -c:a aac -b:a 64k "{output_path}.tmp" -y')
            os.remove(output_path)
            os.rename(f"{output_path}.tmp", output_path)
        
        return output_path
    except Exception as e:
        logging.error(f"Error al convertir el video: {e}")
        raise

async def generar_tts(texto, voz, duracion_total):
    try:
        if not texto.strip():
            raise ValueError("El texto para TTS no puede estar vacío.")
        # Limitar el texto a 1000 caracteres para procesar más rápido
        if len(texto) > 1000:
            texto = texto[:1000]
            logging.info("Texto para TTS truncado a 1000 caracteres para optimizar rendimiento")
        
        logging.info(f"Generando TTS con voz: {voz}")
        communicate = edge_tts.Communicate(texto, voz)
        with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmp_tts:
            await communicate.save(tmp_tts.name)
            tts_audio = AudioFileClip(tmp_tts.name)
            if tts_audio.duration > duracion_total:
                tts_audio = tts_audio.subclip(0, duracion_total)
            return tts_audio, tmp_tts.name
    except Exception as e:
        logging.error(f"Fallo en TTS: {str(e)}")
        raise

def crear_musica_fondo(duracion_total):
    bg_music = AudioSegment.from_mp3(MUSIC_BG)
    needed_ms = int(duracion_total * 1000)
    repeticiones = needed_ms // len(bg_music) + 1
    bg_music = bg_music * repeticiones
    bg_music = bg_music[:needed_ms].fade_out(1000)
    with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmp_bg:
        bg_music.export(tmp_bg.name, format="mp3")
        return AudioFileClip(tmp_bg.name).volumex(0.15), tmp_bg.name

def create_slide_transition(clip1, clip2, duration=TRANSITION_DURATION):
    # Obtener dimensiones de los clips
    width, height = clip1.size
    
    part1 = clip1.subclip(clip1.duration - duration)
    part2 = clip2.subclip(0, duration)
    transition = CompositeVideoClip([
        part1.fx(vfx.fadeout, duration),
        part2.fx(vfx.fadein, duration).set_position(
            lambda t: ('center', height - (height * (t/duration)))
        )
    ], size=(width, height)).set_duration(duration)
    return transition

def liberar_memoria(objetos_cerrar=None):
    """Forzar liberación de memoria cerrando objetos y llamando al recolector de basura"""
    if objetos_cerrar:
        for obj in objetos_cerrar:
            if obj is not None:
                try:
                    obj.close()
                except:
                    pass
    
    # Forzar recolección de basura
    gc.collect()
    mostrar_uso_memoria()

async def procesar_video(video_input, texto_tts, voz_seleccionada, progress=gr.Progress()):
    temp_files = []
    intro, outro, video_original = None, None, None
    segmentos_temp = []
    
    try:
        mostrar_uso_memoria()
        logging.info("Iniciando procesamiento")
        progress(0, desc="Validando video")
        
        if not validar_video(video_input):
            progress(0.05, desc="Optimizando formato de video")
            video_input = convertir_video(video_input)
            temp_files.append(video_input)
            
        progress(0.1, desc="Preparando video")
        # Cargamos el video original pero respetamos su resolución
        video_original = VideoFileClip(video_input)
        duracion_video = video_original.duration
        # Guardamos las dimensiones originales
        original_size = video_original.size
        video_original.close()  # Cerrar para liberar memoria
        
        # Información importante sobre el video original
        logging.info(f"Duración total del video: {duracion_video} segundos")
        logging.info(f"Resolución original: {original_size[0]}x{original_size[1]}")
        
        if duracion_video <= 0:
            raise ValueError("El video debe tener una duración mayor que cero.")
        
        progress(0.2, desc="Generando narración (TTS)")
        tts_audio, tts_path = await generar_tts(texto_tts, voz_seleccionada, duracion_video)
        temp_files.append(tts_path)
        
        progress(0.3, desc="Preparando música de fondo")
        bg_audio, bg_path = crear_musica_fondo(duracion_video)
        temp_files.append(bg_path)
        
        # Procesar por bloques para optimizar memoria
        num_chunks = int(duracion_video // PROCESSING_CHUNK) + (1 if duracion_video % PROCESSING_CHUNK > 0 else 0)
        logging.info(f"Procesando video en {num_chunks} bloques")
        
        for chunk_idx in range(num_chunks):
            chunk_start = chunk_idx * PROCESSING_CHUNK
            chunk_end = min((chunk_idx + 1) * PROCESSING_CHUNK, duracion_video)
            
            progress(0.35 + (0.45 * chunk_idx / num_chunks), 
                    desc=f"Procesando bloque {chunk_idx+1}/{num_chunks} ({chunk_start:.1f}s - {chunk_end:.1f}s)")
            
            # Cargar solo la porción del video que necesitamos
            chunk_video = VideoFileClip(video_input).subclip(chunk_start, chunk_end)
            
            # Extraer la porción de audio correspondiente a este bloque
            tts_chunk_end = min(chunk_end, tts_audio.duration)
            chunk_tts = None
            if chunk_start < tts_audio.duration:
                chunk_tts = tts_audio.subclip(chunk_start, tts_chunk_end)
            
            chunk_bg = bg_audio.subclip(chunk_start, chunk_end)
            
            # Crear la mezcla de audio para este bloque
            audio_chunks = [chunk_bg]
            if chunk_video.audio:
                audio_chunks.append(chunk_video.audio.volumex(0.5))
            if chunk_tts:
                audio_chunks.append(chunk_tts.volumex(0.85))
            
            chunk_audio_final = CompositeAudioClip(audio_chunks)
            chunk_video = chunk_video.set_audio(chunk_audio_final)
            
            # Procesar las transiciones dentro de este chunk si es necesario
            if chunk_end - chunk_start > SEGMENT_DURATION:
                segments_in_chunk = []
                segments_count = int((chunk_end - chunk_start) // SEGMENT_DURATION) + \
                                (1 if (chunk_end - chunk_start) % SEGMENT_DURATION > 0 else 0)
                
                for i in range(segments_count):
                    seg_start = i * SEGMENT_DURATION
                    seg_end = min(seg_start + SEGMENT_DURATION, chunk_end - chunk_start)
                    segment = chunk_video.subclip(seg_start, seg_end)
                    
                    if i == 0:
                        segments_in_chunk.append(segment)
                    else:
                        prev_segment = segments_in_chunk[-1]
                        transition = create_slide_transition(prev_segment, segment)
                        
                        prev_end = prev_segment.duration - TRANSITION_DURATION
                        if prev_end > 0:
                            segments_in_chunk[-1] = prev_segment.subclip(0, prev_end)
                        
                        segments_in_chunk.append(transition)
                        segments_in_chunk.append(segment)
                
                chunk_processed = concatenate_videoclips(segments_in_chunk, method="compose")
            else:
                chunk_processed = chunk_video
            
            # Guardar este chunk procesado como archivo temporal
            with tempfile.NamedTemporaryFile(delete=False, suffix=f"_chunk{chunk_idx}.mp4") as chunk_file:
                chunk_path = chunk_file.name
                chunk_processed.write_videofile(
                    chunk_path,
                    codec="libx264", 
                    audio_codec="aac",
                    preset="ultrafast",
                    bitrate="1M",
                    ffmpeg_params=["-crf", "28"],
                    verbose=False
                )
                segmentos_temp.append(chunk_path)
            
            # Liberar memoria
            chunk_video.close()
            chunk_processed.close()
            liberar_memoria()
        
        # Liberar memoria antes de procesar intro/outro
        liberar_memoria([tts_audio, bg_audio])
        tts_audio = bg_audio = None
        
        # Añadir intro y outro - conservar resolución original para consistencia
        progress(0.85, desc="Preparando intro y outro")
        intro = VideoFileClip(INTRO_VIDEO)
        with tempfile.NamedTemporaryFile(delete=False, suffix="_intro.mp4") as tmp_intro:
            intro.write_videofile(
                tmp_intro.name, 
                codec="libx264", 
                audio_codec="aac", 
                preset="ultrafast", 
                bitrate="1M",
                ffmpeg_params=["-crf", "28"],
                verbose=False
            )
            segmentos_temp.insert(0, tmp_intro.name)  # Intro al principio
        intro.close()
        
        outro = VideoFileClip(OUTRO_VIDEO)
        with tempfile.NamedTemporaryFile(delete=False, suffix="_outro.mp4") as tmp_outro:
            outro.write_videofile(
                tmp_outro.name, 
                codec="libx264", 
                audio_codec="aac", 
                preset="ultrafast", 
                bitrate="1M",
                ffmpeg_params=["-crf", "28"],
                verbose=False
            )
            segmentos_temp.append(tmp_outro.name)  # Outro al final
        outro.close()
        
        # Unir todos los segmentos con ffmpeg
        progress(0.9, desc="Generando video final")
        with tempfile.NamedTemporaryFile(suffix=".txt", delete=False) as concat_file:
            # Escribir archivo de lista para concatenación
            for segment in segmentos_temp:
                concat_file.write(f"file '{segment}'\n".encode())
            concat_path = concat_file.name
            
        with tempfile.NamedTemporaryFile(suffix=".mp4", delete=False) as tmp_final:
            output_path = tmp_final.name
            os.system(f'ffmpeg -f concat -safe 0 -i "{concat_path}" -c copy "{output_path}" -y')
            
            # Limpiar archivos temporales
            os.remove(concat_path)
            for segment in segmentos_temp:
                if os.path.exists(segment):
                    os.remove(segment)
            
            eliminar_archivo_tiempo(output_path, 3600)  # Eliminación después de 1 hora
            progress(1.0, desc="¡Video listo!")
            logging.info(f"Video final guardado: {output_path}")
            mostrar_uso_memoria()
            return output_path
    except Exception as e:
        logging.error(f"Fallo general: {str(e)}")
        raise
    finally:
        try:
            liberar_memoria([video_original, intro, outro])
            for file in temp_files:
                try:
                    if os.path.exists(file):
                        os.remove(file)
                except Exception as e:
                    logging.warning(f"Error limpiando {file}: {e}")
            for segment in segmentos_temp:
                try:
                    if os.path.exists(segment):
                        os.remove(segment)
                except Exception as e:
                    logging.warning(f"Error limpiando segmento {segment}: {e}")
        except Exception as e:
            logging.warning(f"Error al cerrar recursos: {str(e)}")

# Interfaz Gradio
with gr.Blocks() as demo:
    gr.Markdown("# Editor de Video con IA")
    with gr.Tab("Principal"):
        video_input = gr.Video(label="Subir video")
        texto_tts = gr.Textbox(
            label="Texto para TTS (máx. 1000 caracteres)",
            lines=3,
            placeholder="Escribe aquí tu texto..."
        )
        voz_seleccionada = gr.Dropdown(
            label="Voz",
            choices=[
                "es-ES-AlvaroNeural", "es-MX-BeatrizNeural", 
                "es-ES-ElviraNeural", "es-MX-JavierNeural",
                "es-AR-ElenaNeural", "es-AR-TomasNeural", 
                "es-CL-CatalinaNeural", "es-CL-LorenzoNeural", 
                "es-CO-SofiaNeural", "es-CO-GonzaloNeural", 
                "es-PE-CamilaNeural", "es-PE-AlexNeural", 
                "es-VE-MariaNeural", "es-VE-ManuelNeural", 
                "es-US-AlonsoNeural", "es-US-PalomaNeural", 
                "es-ES-AbrilNeural", "es-ES-DarioNeural", 
                "es-ES-HelenaRUS", "es-ES-LauraNeural", 
                "es-ES-PabloNeural", "es-ES-TriniNeural", 
                "en-US-AriaNeural", "en-US-GuyNeural", 
                "en-US-JennyNeural", "en-US-AmberNeural", 
                "en-US-AnaNeural", "en-US-AshleyNeural", 
                "en-US-BrandonNeural", "en-US-ChristopherNeural", 
                "en-US-CoraNeural", "en-US-DavisNeural", 
                "en-US-ElizabethNeural", "en-US-EricNeural", 
                "en-US-GinaNeural", "en-US-JacobNeural", 
                "en-US-JaneNeural", "en-US-JasonNeural", 
                "en-US-MichelleNeural", "en-US-MonicaNeural", 
                "en-US-SaraNeural", "en-US-SteffanNeural", 
                "en-US-TonyNeural", "en-US-YaraNeural", 
                "fr-FR-AlainNeural", "fr-FR-BrigitteNeural", 
                "fr-FR-CelesteNeural", "fr-FR-ClaudeNeural", 
                "fr-FR-CoralieNeural", "fr-FR-DeniseNeural", 
                "fr-FR-EloiseNeural", "fr-FR-HenriNeural", 
                "fr-FR-JacquelineNeural", "fr-FR-JeromeNeural", 
                "fr-FR-JosephineNeural", "fr-FR-MauriceNeural", 
                "fr-FR-YvesNeural", "fr-FR-YvetteNeural", 
                "de-DE-AmalaNeural", "de-DE-BerndNeural", 
                "de-DE-ChristophNeural", "de-DE-ConradNeural", 
                "de-DE-ElkeNeural", "de-DE-GiselaNeural", 
                "de-DE-KasperNeural", "de-DE-KatjaNeural", 
                "de-DE-KillianNeural", "de-DE-KlarissaNeural", 
                "de-DE-KlausNeural", "de-DE-LouisaNeural", 
                "de-DE-MajaNeural", "de-DE-RalfNeural", 
                "de-DE-TanjaNeural", "de-DE-ViktoriaNeural", 
                "it-IT-BenignoNeural", "it-IT-CalimeroNeural", 
                "it-IT-CataldoNeural", "it-IT-DiegoNeural", 
                "it-IT-ElsaNeural", "it-IT-FabiolaNeural", 
                "it-IT-GianniNeural", "it-IT-ImeldaNeural", 
                "it-IT-IrmaNeural", "it-IT-IsabellaNeural", 
                "it-IT-LisandroNeural", "it-IT-PalmiraNeural", 
                "it-IT-PierinaNeural", "it-IT-RinaldoNeural", 
                "ja-JP-AoiNeural", "ja-JP-DaichiNeural", 
                "ja-JP-HarukaNeural", "ja-JP-KeitaNeural", 
                "ja-JP-MayuNeural", "ja-JP-NanamiNeural", 
                "ja-JP-NaokiNeural", "ja-JP-ShioriNeural"
            ],
            value="es-ES-AlvaroNeural"
        )
        procesar_btn = gr.Button("Generar Video (Modo Optimizado)")
        video_output = gr.Video(label="Video Procesado")
    with gr.Accordion("Ejemplos de Uso", open=False):
        gr.Examples(
            examples=[[EJEMPLO_VIDEO, "¡Hola! Esto es una prueba. Suscríbete al canal."]],
            inputs=[video_input, texto_tts],
            label="Ejemplos"
        )
    procesar_btn.click(
        procesar_video,
        inputs=[video_input, texto_tts, voz_seleccionada],
        outputs=video_output
    )

    gr.Markdown("""
    ### ℹ️ Notas importantes:
    - **Optimizaciones para Hugging Face Spaces:**
      - Procesamiento por bloques para videos largos
      - Máximo tamaño de archivo: 200MB
      - Mantiene la resolución original del video
      - Texto TTS limitado a 1000 caracteres
    - Las transiciones ocurren cada 30 segundos
    - El video contiene intro y outro predefinidos
    - El archivo generado se elimina después de 1 hora
    - Para videos de alta calidad, considera usar este código localmente
    """)

if __name__ == "__main__":
    # Instalar psutil si no está disponible
    try:
        import psutil
    except ImportError:
        os.system("pip install psutil")
        import psutil
        
    demo.queue().launch()