import tempfile import logging import os import asyncio 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" GLITCH_SOUND = "fxsound.mp3" EJEMPLO_VIDEO = "ejemplo.mp4" # Validar existencia de archivos for file in [INTRO_VIDEO, OUTRO_VIDEO, MUSIC_BG, GLITCH_SOUND, EJEMPLO_VIDEO]: if not os.path.exists(file): logging.error(f"Falta archivo necesario: {file}") raise FileNotFoundError(f"Falta: {file}") def eliminar_archivo_tiempo(ruta, delay=1800): 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() async def procesar_audio(texto, voz, duracion_video, audio_original): temp_files = [] try: if not texto.strip(): raise ValueError("El texto para TTS no puede estar vacío.") def dividir_texto(texto, max_length=3000): return [texto[i:i + max_length] for i in range(0, len(texto), max_length)] fragmentos = dividir_texto(texto) audios_tts = [] for fragmento in fragmentos: communicate = edge_tts.Communicate(fragmento, voz) with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmp_tts: await communicate.save(tmp_tts.name) tts_audio = AudioFileClip(tmp_tts.name) temp_files.append(tmp_tts.name) audios_tts.append(tts_audio) tts_audio_final = concatenate_audioclips(audios_tts) if tts_audio_final.duration > duracion_video: tts_audio_final = tts_audio_final.subclip(0, duracion_video) needed_ms = int(duracion_video * 1000) bg_music = AudioSegment.from_mp3(MUSIC_BG) 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") bg_audio = AudioFileClip(tmp_bg.name).volumex(0.15) temp_files.append(tmp_bg.name) audios = [bg_audio.set_duration(duracion_video)] if audio_original: audios.append(audio_original.volumex(0.7)) audios.append(tts_audio_final.volumex(0.85).set_start(0)) audio_final = CompositeAudioClip(audios).set_duration(duracion_video) return audio_final except Exception as e: logging.error(f" fallo en audio: {str(e)}") raise finally: for file in temp_files: try: os.remove(file) except Exception as e: logging.warning(f"Error limpiando {file}: {e}") def aplicar_glitch(video_clip): def glitch_effect(frame): import numpy as np frame = frame.copy() # Corrección clave para evitar errores height, width, _ = frame.shape offset = np.random.randint(5, 15) if height > 0: frame[offset:, :] = np.roll(frame[:-offset, :], -offset, axis=0) return frame return video_clip.fl_image(glitch_effect) async def procesar_video(video_input, texto_tts, voz_seleccionada): try: # Carga optimizada intro = VideoFileClip(INTRO_VIDEO, target_resolution=(1080, 1920)) outro = VideoFileClip(OUTRO_VIDEO, target_resolution=(1080, 1920)) video_original = VideoFileClip(video_input, target_resolution=(1080, 1920)) audio_original = video_original.audio # Liberar recursos intro.reader.close() outro.reader.close() video_original.reader.close() duracion_video = video_original.duration audio_final = await procesar_audio( texto_tts, voz_seleccionada, duracion_video, audio_original ) # Configuración de cortes dinámicos segment_duration = 18 # Duración visible del segmento overlap = 2 # Segundos eliminados en cada corte total_segments = int((duracion_video) // (segment_duration)) + 1 segments = [] glitch_clips = [] glitch_sound = AudioFileClip(GLITCH_SOUND).volumex(0.5) start_time = 0 for i in range(total_segments): end_time = start_time + segment_duration + overlap end_time = min(end_time, duracion_video) # Extraer segmento completo full_segment = video_original.subclip(start_time, end_time) # Aplicar transición/glitch if i > 0: # Tomar 0.5 segundos para glitch glitch_part = full_segment.subclip(0, 0.5) glitch_part = aplicar_glitch(glitch_part) # Combinar con el resto del segmento processed_segment = concatenate_videoclips([ glitch_part, full_segment.subclip(0.5) ], method="compose") # Agregar sonido de glitch glitch_sound_clip = glitch_sound.set_start(start_time) glitch_clips.append(glitch_sound_clip) else: processed_segment = full_segment segments.append(processed_segment) start_time += segment_duration # Avanzar sin los 2 segundos de overlap # Combinar todos los segmentos video_final = concatenate_videoclips(segments, method="compose") video_con_audio = video_final.set_audio(audio_final) # Agregar intro y outro intro = VideoFileClip(INTRO_VIDEO, target_resolution=(1080, 1920)) outro = VideoFileClip(OUTRO_VIDEO, target_resolution=(1080, 1920)) video_final = concatenate_videoclips([intro, video_con_audio, outro], method="compose") # Renderizado final with tempfile.NamedTemporaryFile(suffix=".mp4", delete=False) as tmp: video_final.write_videofile( tmp.name, codec="libx264", audio_codec="aac", fps=video_original.fps, threads=4, bitrate="5M", ffmpeg_params=[ "-preset", "ultrafast", "-crf", "23", "-movflags", "+faststart", "-vf", "scale=1920:1080" ], verbose=False ) eliminar_archivo_tiempo(tmp.name, 1800) return tmp.name except Exception as e: logging.error(f" fallo general: {str(e)}") raise finally: try: intro.close() outro.close() video_original.close() except: pass # 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", lines=3, placeholder="Escribe aquí tu texto..." ) voz_seleccionada = gr.Dropdown( label="Voz", choices=["es-ES-AlvaroNeural", "es-MX-BeatrizNeural"], value="es-ES-AlvaroNeural" ) procesar_btn = gr.Button("Generar Video") 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 ) if __name__ == "__main__": demo.queue().launch()