Spaces:
Sleeping
Sleeping
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" | |
EJEMPLO_VIDEO = "ejemplo.mp4" | |
# 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 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: | |
# Validar texto | |
if not texto.strip(): | |
raise ValueError("El texto para TTS no puede estar vacío.") | |
# Dividir el texto en fragmentos si es demasiado largo | |
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: | |
# Generar TTS | |
communicate = edge_tts.Communicate(fragmento, voz) | |
with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmp_tts: | |
try: | |
await communicate.save(tmp_tts.name) | |
except edge_tts.exceptions.NoAudioReceived as e: | |
logging.error(f"Error en TTS: {str(e)}") | |
raise ValueError("No se pudo generar el audio. Verifica tu conexión o los parámetros del TTS.") | |
tts_audio = AudioFileClip(tmp_tts.name) | |
temp_files.append(tmp_tts.name) | |
audios_tts.append(tts_audio) | |
# Combinar todos los fragmentos de TTS | |
tts_audio_final = concatenate_audioclips(audios_tts) | |
# Limitar TTS al video | |
if tts_audio_final.duration > duracion_video: | |
tts_audio_final = tts_audio_final.subclip(0, duracion_video) | |
# Preparar música de fondo en loop | |
bg_music = AudioSegment.from_mp3(MUSIC_BG) | |
needed_ms = int(duracion_video * 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") | |
bg_audio = AudioFileClip(tmp_bg.name).volumex(0.15) | |
temp_files.append(tmp_bg.name) | |
# Combinar audios | |
audios = [bg_audio.set_duration(duracion_video)] | |
if audio_original: | |
audios.append(audio_original.volumex(0.7)) # Audio original al 70% | |
audios.append(tts_audio_final.volumex(0.85).set_start(0)) # TTS al 85% | |
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}") | |
async def procesar_video(video_input, texto_tts, voz_seleccionada): | |
try: | |
# Cargar componentes | |
intro = VideoFileClip(INTRO_VIDEO) | |
outro = VideoFileClip(OUTRO_VIDEO) | |
video_original = VideoFileClip(video_input) | |
audio_original = video_original.audio | |
# Duración del video editado (sin intro/outro) | |
duracion_video = video_original.duration | |
# Procesar audio | |
audio_final = await procesar_audio( | |
texto_tts, | |
voz_seleccionada, | |
duracion_video, | |
audio_original | |
) | |
# Redimensionar temporalmente el video a un tamaño muy grande (4000px de ancho) | |
video_temporal = video_original.resize(width=4000) | |
# Redimensionar nuevamente al tamaño final deseado (1920x1080) | |
target_width = 1920 | |
target_height = 1080 | |
video_resized = video_temporal.resize(height=target_height) | |
# Manejar márgenes si es necesario | |
current_width = video_resized.w | |
current_height = video_resized.h | |
if current_width < target_width: | |
# Agregar márgenes laterales | |
margin_left = (target_width - current_width) // 2 | |
margin_right = target_width - current_width - margin_left | |
video_final = video_resized.margin( | |
left=margin_left, | |
right=margin_right, | |
color=(0, 0, 0) # Negro | |
) | |
else: | |
# Recortar horizontalmente si es más ancho que 1920 | |
video_final = video_resized.crop( | |
x1=(current_width - target_width) // 2, | |
x2=current_width - (current_width - target_width) // 2 | |
) | |
# Combinar video con audio | |
video_con_audio = video_final.set_audio(audio_final) | |
# Concatenar intro + video + outro SIN alteraciones | |
video_final = concatenate_videoclips( | |
[intro, video_con_audio, outro], | |
method="compose", # Evitar problemas de grid | |
padding=0 # Sin espacio entre clips | |
) | |
# Renderizar video final con metadatos correctos | |
with tempfile.NamedTemporaryFile(suffix=".mp4", delete=False) as tmp: | |
video_final.write_videofile( | |
tmp.name, | |
codec="libx264", | |
audio_codec="aac", | |
fps=video_original.fps, # Mantener FPS original | |
threads=4, | |
ffmpeg_params=[ | |
"-aspect", "16:9", # Forzar relación de aspecto | |
"-vf", "scale=1920:1080" # Forzar escalado explícito | |
], | |
verbose=False | |
) | |
eliminar_archivo_tiempo(tmp.name) | |
return tmp.name | |
except Exception as e: | |
logging.error(f" fallo general: {str(e)}") | |
raise | |
# 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() |