Spaces:
Sleeping
Sleeping
import os | |
import logging | |
import tempfile | |
import requests | |
from datetime import datetime | |
import gradio as gr | |
import torch | |
from transformers import GPT2Tokenizer, GPT2LMHeadModel | |
from keybert import KeyBERT | |
from TTS.api import TTS | |
from moviepy.editor import VideoFileClip, concatenate_videoclips, AudioFileClip, CompositeAudioClip, concatenate_audioclips, AudioClip | |
import re | |
import math | |
import shutil | |
import json | |
from collections import Counter | |
import threading | |
import time | |
# Variable global para TTS | |
tts_model = None | |
# Configuración de logging | |
logging.basicConfig( | |
level=logging.DEBUG, | |
format='%(asctime)s - %(levelname)s - %(message)s', | |
handlers=[ | |
logging.StreamHandler(), | |
logging.FileHandler('video_generator_full.log', encoding='utf-8') | |
] | |
) | |
logger = logging.getLogger(__name__) | |
logger.info("="*80) | |
logger.info("INICIO DE EJECUCIÓN - GENERADOR DE VIDEOS") | |
logger.info("="*80) | |
# Clave API de Pexels | |
PEXELS_API_KEY = os.environ.get("PEXELS_API_KEY") | |
if not PEXELS_API_KEY: | |
logger.critical("NO SE ENCONTRÓ PEXELS_API_KEY EN VARIABLES DE ENTORNO") | |
# Inicialización de modelos | |
MODEL_NAME = "datificate/gpt2-small-spanish" | |
logger.info(f"Inicializando modelo GPT-2: {MODEL_NAME}") | |
tokenizer = None | |
model = None | |
try: | |
tokenizer = GPT2Tokenizer.from_pretrained(MODEL_NAME) | |
model = GPT2LMHeadModel.from_pretrained(MODEL_NAME).eval() | |
if tokenizer.pad_token is None: | |
tokenizer.pad_token = tokenizer.eos_token | |
logger.info(f"Modelo GPT-2 cargado | Vocabulario: {len(tokenizer)} tokens") | |
except Exception as e: | |
logger.error(f"FALLA CRÍTICA al cargar GPT-2: {str(e)}", exc_info=True) | |
tokenizer = model = None | |
logger.info("Cargando modelo KeyBERT...") | |
kw_model = None | |
try: | |
kw_model = KeyBERT('distilbert-base-multilingual-cased') | |
logger.info("KeyBERT inicializado correctamente") | |
except Exception as e: | |
logger.error(f"FALLA al cargar KeyBERT: {str(e)}", exc_info=True) | |
kw_model = None | |
def schedule_deletion(directory_path, delay_seconds): | |
"""Programa la eliminación de un directorio después de un cierto tiempo.""" | |
logger.info(f"PROGRAMADA eliminación del directorio '{directory_path}' en {delay_seconds / 3600:.1f} horas.") | |
time.sleep(delay_seconds) | |
try: | |
if os.path.isdir(directory_path): | |
shutil.rmtree(directory_path) | |
logger.info(f"Directorio temporal '{directory_path}' eliminado exitosamente.") | |
else: | |
logger.warning(f"No se pudo eliminar: '{directory_path}' no es un directorio válido o ya fue eliminado.") | |
except Exception as e: | |
logger.error(f"Error durante la eliminación programada de '{directory_path}': {str(e)}") | |
def buscar_videos_pexels(query, api_key, per_page=5): | |
if not api_key: | |
logger.warning("No se puede buscar en Pexels: API Key no configurada.") | |
return [] | |
logger.debug(f"Buscando en Pexels: '{query}' | Resultados: {per_page}") | |
headers = {"Authorization": api_key} | |
try: | |
params = { | |
"query": query, | |
"per_page": per_page, | |
"orientation": "landscape", | |
"size": "medium" | |
} | |
response = requests.get( | |
"https://api.pexels.com/videos/search", | |
headers=headers, | |
params=params, | |
timeout=20 | |
) | |
response.raise_for_status() | |
data = response.json() | |
videos = data.get('videos', []) | |
logger.info(f"Pexels: {len(videos)} videos encontrados para '{query}'") | |
return videos | |
except requests.exceptions.RequestException as e: | |
logger.error(f"Error de conexión Pexels para '{query}': {str(e)}") | |
except json.JSONDecodeError: | |
logger.error(f"Pexels: JSON inválido recibido | Status: {response.status_code} | Respuesta: {response.text[:200]}...") | |
except Exception as e: | |
logger.error(f"Error inesperado Pexels para '{query}': {str(e)}", exc_info=True) | |
return [] | |
def generate_script(prompt, max_length=150): | |
logger.info(f"Generando guión | Prompt: '{prompt[:50]}...' | Longitud máxima: {max_length}") | |
if not tokenizer or not model: | |
logger.warning("Modelos GPT-2 no disponibles - Usando prompt original como guion.") | |
return prompt.strip() | |
instruction_phrase_start = "Escribe un guion corto, interesante y coherente sobre:" | |
ai_prompt = f"{instruction_phrase_start} {prompt}" | |
try: | |
inputs = tokenizer(ai_prompt, return_tensors="pt", truncation=True, max_length=512) | |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu") | |
model.to(device) | |
inputs = {k: v.to(device) for k, v in inputs.items()} | |
outputs = model.generate( | |
**inputs, | |
max_length=max_length + inputs[list(inputs.keys())[0]].size(1), | |
do_sample=True, | |
top_p=0.9, | |
top_k=40, | |
temperature=0.7, | |
repetition_penalty=1.2, | |
pad_token_id=tokenizer.pad_token_id, | |
eos_token_id=tokenizer.eos_token_id, | |
no_repeat_ngram_size=3 | |
) | |
text = tokenizer.decode(outputs[0], skip_special_tokens=True) | |
cleaned_text = text.strip() | |
try: | |
instruction_end_idx = text.find(instruction_phrase_start) | |
if instruction_end_idx != -1: | |
cleaned_text = text[instruction_end_idx + len(instruction_phrase_start):].strip() | |
logger.debug("Instrucción inicial encontrada y eliminada del guión generado.") | |
else: | |
instruction_start_idx = text.find(instruction_phrase_start) | |
if instruction_start_idx != -1: | |
prompt_in_output_idx = text.find(prompt, instruction_start_idx) | |
if prompt_in_output_idx != -1: | |
cleaned_text = text[prompt_in_output_idx + len(prompt):].strip() | |
logger.debug("Instrucción base y prompt encontrados y eliminados del guión generado.") | |
else: | |
cleaned_text = text[instruction_start_idx + len(instruction_phrase_start):].strip() | |
logger.debug("Instrucción base encontrada, eliminada del guión generado (sin prompt detectado).") | |
except Exception as e: | |
logger.warning(f"Error durante la limpieza heurística del guión de IA: {e}. Usando texto generado sin limpieza adicional.") | |
cleaned_text = re.sub(r'<[^>]+>', '', text).strip() | |
if not cleaned_text or len(cleaned_text) < 10: | |
logger.warning("El guión generado parece muy corto o vacío después de la limpieza. Usando el texto generado original (sin limpieza heurística).") | |
cleaned_text = re.sub(r'<[^>]+>', '', text).strip() | |
cleaned_text = re.sub(r'<[^>]+>', '', cleaned_text).strip() | |
cleaned_text = cleaned_text.lstrip(':').strip() | |
cleaned_text = cleaned_text.lstrip('.').strip() | |
sentences = cleaned_text.split('.') | |
if sentences and sentences[0].strip(): | |
final_text = sentences[0].strip() + '.' | |
if len(sentences) > 1 and sentences[1].strip() and len(final_text.split()) < max_length * 0.7: | |
final_text += " " + sentences[1].strip() + "." | |
final_text = final_text.replace("..", ".") | |
logger.info(f"Guion generado final (Truncado a 100 chars): '{final_text[:100]}...'") | |
return final_text.strip() | |
logger.info(f"Guion generado final (sin oraciones completas detectadas - Truncado): '{cleaned_text[:100]}...'") | |
return cleaned_text.strip() | |
except Exception as e: | |
logger.error(f"Error generando guion con GPT-2 (fuera del bloque de limpieza): {str(e)}", exc_info=True) | |
logger.warning("Usando prompt original como guion debido al error de generación.") | |
return prompt.strip() | |
def text_to_speech(text, output_path, voice=None): | |
logger.info(f"Convirtiendo texto a voz con Coqui TTS | Caracteres: {len(text)} | Salida: {output_path}") | |
if not text or not text.strip(): | |
logger.warning("Texto vacío para TTS") | |
return False | |
try: | |
tts = TTS(model_name="tts_models/es/css10/vits", progress_bar=False, gpu=False) | |
text = text.replace("na hora", "A la hora") | |
text = re.sub(r'[^\w\s.,!?áéíóúñÁÉÍÓÚÑ]', '', text) | |
if len(text) > 500: | |
logger.warning("Texto demasiado largo, truncando a 500 caracteres") | |
text = text[:500] | |
tts.tts_to_file(text=text, file_path=output_path) | |
if os.path.exists(output_path) and os.path.getsize(output_path) > 1000: | |
logger.info(f"Audio creado: {output_path} | Tamaño: {os.path.getsize(output_path)} bytes") | |
return True | |
else: | |
logger.error("Archivo de audio vacío o no creado") | |
return False | |
except Exception as e: | |
logger.error(f"Error TTS: {str(e)}", exc_info=True) | |
return False | |
def download_video_file(url, temp_dir): | |
if not url: | |
logger.warning("URL de video no proporcionada para descargar") | |
return None | |
try: | |
logger.info(f"Descargando video desde: {url[:80]}...") | |
os.makedirs(temp_dir, exist_ok=True) | |
file_name = f"video_dl_{datetime.now().strftime('%Y%m%d_%H%M%S_%f')}.mp4" | |
output_path = os.path.join(temp_dir, file_name) | |
with requests.get(url, stream=True, timeout=60) as r: | |
r.raise_for_status() | |
with open(output_path, 'wb') as f: | |
for chunk in r.iter_content(chunk_size=8192): | |
f.write(chunk) | |
if os.path.exists(output_path) and os.path.getsize(output_path) > 1000: | |
logger.info(f"Video descargado exitosamente: {output_path} | Tamaño: {os.path.getsize(output_path)} bytes") | |
return output_path | |
else: | |
logger.warning(f"Descarga parece incompleta o vacía para {url[:80]}... Archivo: {output_path} Tamaño: {os.path.getsize(output_path) if os.path.exists(output_path) else 'N/A'} bytes") | |
if os.path.exists(output_path): | |
os.remove(output_path) | |
return None | |
except requests.exceptions.RequestException as e: | |
logger.error(f"Error de descarga para {url[:80]}... : {str(e)}") | |
except Exception as e: | |
logger.error(f"Error inesperado descargando {url[:80]}... : {str(e)}", exc_info=True) | |
return None | |
def loop_audio_to_length(audio_clip, target_duration): | |
logger.debug(f"Ajustando audio | Duración actual: {audio_clip.duration:.2f}s | Objetivo: {target_duration:.2f}s") | |
if audio_clip is None or audio_clip.duration is None or audio_clip.duration <= 0: | |
logger.warning("Input audio clip is invalid (None or zero duration), cannot loop.") | |
try: | |
sr = getattr(audio_clip, 'fps', 44100) if audio_clip else 44100 | |
return AudioClip(lambda t: 0, duration=target_duration, sr=sr) | |
except Exception as e: | |
logger.error(f"Could not create silence clip: {e}", exc_info=True) | |
return AudioFileClip(filename="") | |
if audio_clip.duration >= target_duration: | |
logger.debug("Audio clip already longer or equal to target. Trimming.") | |
trimmed_clip = audio_clip.subclip(0, target_duration) | |
if trimmed_clip.duration is None or trimmed_clip.duration <= 0: | |
logger.error("Trimmed audio clip is invalid.") | |
try: trimmed_clip.close() | |
except: pass | |
return AudioFileClip(filename="") | |
return trimmed_clip | |
loops = math.ceil(target_duration / audio_clip.duration) | |
logger.debug(f"Creando {loops} loops de audio") | |
audio_segments = [audio_clip] * loops | |
looped_audio = None | |
final_looped_audio = None | |
try: | |
looped_audio = concatenate_audioclips(audio_segments) | |
if looped_audio.duration is None or looped_audio.duration <= 0: | |
logger.error("Concatenated audio clip is invalid (None or zero duration).") | |
raise ValueError("Invalid concatenated audio.") | |
final_looped_audio = looped_audio.subclip(0, target_duration) | |
if final_looped_audio.duration is None or final_looped_audio.duration <= 0: | |
logger.error("Final subclipped audio clip is invalid (None or zero duration).") | |
raise ValueError("Invalid final subclipped audio.") | |
return final_looped_audio | |
except Exception as e: | |
logger.error(f"Error concatenating/subclipping audio clips for looping: {str(e)}", exc_info=True) | |
try: | |
if audio_clip.duration is not None and audio_clip.duration > 0: | |
logger.warning("Returning original audio clip (may be too short).") | |
return audio_clip.subclip(0, min(audio_clip.duration, target_duration)) | |
except: | |
pass | |
logger.error("Fallback to original audio clip failed.") | |
return AudioFileClip(filename="") | |
finally: | |
if looped_audio is not None and looped_audio is not final_looped_audio: | |
try: looped_audio.close() | |
except: pass | |
def extract_visual_keywords_from_script(script_text): | |
logger.info("Extrayendo palabras clave del guion") | |
if not script_text or not script_text.strip(): | |
logger.warning("Guion vacío, no se pueden extraer palabras clave.") | |
return ["naturaleza", "ciudad", "paisaje"] | |
clean_text = re.sub(r'[^\w\sáéíóúñÁÉÍÓÚÑ]', '', script_text) | |
keywords_list = [] | |
if kw_model: | |
try: | |
logger.debug("Intentando extracción con KeyBERT...") | |
keywords1 = kw_model.extract_keywords(clean_text, keyphrase_ngram_range=(1, 1), stop_words='spanish', top_n=5) | |
keywords2 = kw_model.extract_keywords(clean_text, keyphrase_ngram_range=(2, 2), stop_words='spanish', top_n=3) | |
all_keywords = keywords1 + keywords2 | |
all_keywords.sort(key=lambda item: item[1], reverse=True) | |
seen_keywords = set() | |
for keyword, score in all_keywords: | |
formatted_keyword = keyword.lower().replace(" ", "+") | |
if formatted_keyword and formatted_keyword not in seen_keywords: | |
keywords_list.append(formatted_keyword) | |
seen_keywords.add(formatted_keyword) | |
if len(keywords_list) >= 5: | |
break | |
if keywords_list: | |
logger.debug(f"Palabras clave extraídas por KeyBERT: {keywords_list}") | |
return keywords_list | |
except Exception as e: | |
logger.warning(f"KeyBERT falló: {str(e)}. Intentando método simple.") | |
logger.debug("Extrayendo palabras clave con método simple...") | |
words = clean_text.lower().split() | |
stop_words = {"el", "la", "los", "las", "de", "en", "y", "a", "que", "es", "un", "una", "con", "para", "del", "al", "por", "su", "sus", "se", "lo", "le", "me", "te", "nos", "os", "les", "mi", "tu", | |
"nuestro", "vuestro", "este", "ese", "aquel", "esta", "esa", "aquella", "esto", "eso", "aquello", "mis", "tus", | |
"nuestros", "vuestros", "estas", "esas", "aquellas", "si", "no", "más", "menos", "sin", "sobre", "bajo", "entre", "hasta", "desde", "durante", "mediante", "según", "versus", "via", "cada", "todo", "todos", "toda", "todas", "poco", "pocos", "poca", "pocas", "mucho", "muchos", "mucha", "muchas", "varios", "varias", "otro", "otros", "otra", "otras", "mismo", "misma", "mismos", "mismas", "tan", "tanto", "tanta", "tantos", "tantas", "tal", "tales", "cual", "cuales", "cuyo", "cuya", "cuyos", "cuyas", "quien", "quienes", "cuan", "cuanto", "cuanta", "cuantos", "cuantas", "como", "donde", "cuando", "porque", "aunque", "mientras", "siempre", "nunca", "jamás", "muy", "casi", "solo", "solamente", "incluso", "apenas", "quizás", "tal vez", "acaso", "claro", "cierto", "obvio", "evidentemente", "realmente", "simplemente", "generalmente", "especialmente", "principalmente", "posiblemente", "probablemente", "difícilmente", "fácilmente", "rápidamente", "lentamente", "bien", "mal", "mejor", "peor", "arriba", "abajo", "adelante", "atrás", "cerca", "lejos", "dentro", "fuera", "encima", "debajo", "frente", "detrás", "antes", "después", "luego", "pronto", "tarde", "todavía", "ya", "aun", "aún", "quizá"} | |
valid_words = [word for word in words if len(word) > 3 and word not in stop_words] | |
if not valid_words: | |
logger.warning("No se encontraron palabras clave válidas con método simple. Usando palabras clave predeterminadas.") | |
return ["naturaleza", "ciudad", "paisaje"] | |
word_counts = Counter(valid_words) | |
top_keywords = [word.replace(" ", "+") for word, _ in word_counts.most_common(5)] | |
if not top_keywords: | |
logger.warning("El método simple no produjo keywords. Usando palabras clave predeterminadas.") | |
return ["naturaleza", "ciudad", "paisaje"] | |
logger.info(f"Palabras clave finales: {top_keywords}") | |
return top_keywords | |
def crear_video(prompt_type, input_text, musica_file=None): | |
logger.info("="*80) | |
logger.info(f"INICIANDO CREACIÓN DE VIDEO | Tipo: {prompt_type}") | |
logger.debug(f"Input: '{input_text[:100]}...'") | |
start_time = datetime.now() | |
temp_dir_intermediate = None | |
TARGET_RESOLUTION = (1280, 720) # *** MODIFICACIÓN: Resolución 720p *** | |
audio_tts_original = None | |
musica_audio_original = None | |
audio_tts = None | |
musica_audio = None | |
video_base = None | |
video_final = None | |
source_clips = [] | |
clips_to_concatenate = [] | |
try: | |
# 1. Generar o usar guion | |
if prompt_type == "Generar Guion con IA": | |
guion = generate_script(input_text) | |
else: | |
guion = input_text.strip() | |
logger.info(f"Guion final ({len(guion)} chars): '{guion[:100]}...'") | |
if not guion.strip(): | |
logger.error("El guion resultante está vacío o solo contiene espacios.") | |
raise ValueError("El guion está vacío.") | |
guion = guion.replace("na hora", "A la hora") | |
temp_dir_intermediate = tempfile.mkdtemp(prefix="video_gen_intermediate_") | |
logger.info(f"Directorio temporal intermedio creado: {temp_dir_intermediate}") | |
temp_intermediate_files = [] | |
# 2. Generar audio de voz | |
logger.info("Generando audio de voz...") | |
voz_path = os.path.join(temp_dir_intermediate, "voz.mp3") | |
tts_success = text_to_speech(guion, voz_path) | |
if not tts_success or not os.path.exists(voz_path) or os.path.getsize(voz_path) <= 1000: | |
logger.error(f"Fallo en la generación de voz. Archivo de audio no creado o es muy pequeño: {voz_path}") | |
raise ValueError("Error generando voz a partir del guion (fallo de TTS).") | |
temp_intermediate_files.append(voz_path) | |
audio_tts_original = AudioFileClip(voz_path) | |
if audio_tts_original.reader is None or audio_tts_original.duration is None or audio_tts_original.duration <= 0: | |
logger.critical("Clip de audio TTS inicial es inválido (reader is None o duración <= 0).") | |
try: audio_tts_original.close() | |
except: pass | |
audio_tts_original = None | |
raise ValueError("Audio de voz generado es inválido después de procesamiento inicial.") | |
audio_tts = audio_tts_original | |
audio_duration = audio_tts_original.duration | |
logger.info(f"Duración audio voz: {audio_duration:.2f} segundos") | |
if audio_duration < 1.0: | |
logger.error(f"Duración audio voz ({audio_duration:.2f}s) es muy corta.") | |
raise ValueError("Generated voice audio is too short (min 1 second required).") | |
# 3. Extraer palabras clave | |
logger.info("Extrayendo palabras clave...") | |
try: | |
keywords = extract_visual_keywords_from_script(guion) | |
logger.info(f"Palabras clave identificadas: {keywords}") | |
except Exception as e: | |
logger.error(f"Error extrayendo keywords: {str(e)}", exc_info=True) | |
keywords = ["naturaleza", "paisaje"] | |
if not keywords: | |
keywords = ["video", "background"] | |
# 4. Buscar y descargar videos | |
logger.info("Buscando videos en Pexels...") | |
videos_data = [] | |
total_desired_videos = 10 | |
per_page_per_keyword = max(1, total_desired_videos // len(keywords)) | |
for keyword in keywords: | |
if len(videos_data) >= total_desired_videos: break | |
try: | |
videos = buscar_videos_pexels(keyword, PEXELS_API_KEY, per_page=per_page_per_keyword) | |
if videos: | |
videos_data.extend(videos) | |
except Exception as e: | |
logger.warning(f"Error buscando videos para '{keyword}': {str(e)}") | |
if len(videos_data) < total_desired_videos / 2: | |
logger.warning(f"Pocos videos encontrados ({len(videos_data)}). Intentando con palabras clave genéricas.") | |
for keyword in ["nature", "city", "background", "abstract"]: | |
if len(videos_data) >= total_desired_videos: break | |
try: | |
videos = buscar_videos_pexels(keyword, PEXELS_API_KEY, per_page=2) | |
if videos: | |
videos_data.extend(videos) | |
except Exception as e: | |
logger.warning(f"Error buscando videos genéricos para '{keyword}': {str(e)}") | |
if not videos_data: | |
logger.error("No se encontraron videos en Pexels para ninguna palabra clave.") | |
raise ValueError("No se encontraron videos adecuados en Pexels.") | |
video_paths = [] | |
logger.info(f"Intentando descargar {len(videos_data)} videos encontrados...") | |
for video in videos_data: | |
if 'video_files' not in video or not video['video_files']: continue | |
try: | |
best_quality = next((vf for vf in sorted(video['video_files'], key=lambda x: x.get('width', 0), reverse=True) if 'link' in vf), None) | |
if best_quality: | |
path = download_video_file(best_quality['link'], temp_dir_intermediate) | |
if path: | |
video_paths.append(path) | |
temp_intermediate_files.append(path) | |
except Exception as e: | |
logger.warning(f"Error procesando/descargando video {video.get('id')}: {str(e)}") | |
logger.info(f"Descargados {len(video_paths)} archivos de video utilizables.") | |
if not video_paths: | |
logger.error("No se pudo descargar ningún archivo de video utilizable.") | |
raise ValueError("No se pudo descargar ningún video utilizable de Pexels.") | |
# 5. Procesar y concatenar clips de video | |
logger.info("Procesando y concatenando videos descargados...") | |
current_duration = 0 | |
min_clip_duration = 0.5 | |
max_clip_segment = 10.0 | |
for i, path in enumerate(video_paths): | |
if current_duration >= audio_duration + max_clip_segment: break | |
clip = None | |
try: | |
clip = VideoFileClip(path) | |
source_clips.append(clip) | |
if clip.reader is None or clip.duration is None or clip.duration <= 0: continue | |
remaining_needed = audio_duration - current_duration | |
if remaining_needed > 0: | |
segment_duration = min(min(clip.duration, max_clip_segment), remaining_needed + min_clip_duration) | |
segment_duration = max(min_clip_duration, segment_duration) | |
if segment_duration >= min_clip_duration: | |
try: | |
sub_raw = clip.subclip(0, segment_duration) | |
# *** MODIFICACIÓN: Redimensionar y recortar cada subclip a 720p *** | |
sub = sub_raw.resize(height=TARGET_RESOLUTION[1]).crop(x_center='center', width=TARGET_RESOLUTION[0], height=TARGET_RESOLUTION[1]) | |
sub_raw.close() # Liberar memoria del clip intermedio | |
if sub.reader is None or sub.duration is None or sub.duration <= 0: | |
try: sub.close() | |
except: pass | |
continue | |
clips_to_concatenate.append(sub) | |
current_duration += sub.duration | |
logger.debug(f"Segmento añadido: {sub.duration:.1f}s (total video: {current_duration:.1f}/{audio_duration:.1f}s)") | |
except Exception as sub_e: | |
logger.warning(f"Error creando subclip de {path}: {str(sub_e)}") | |
except Exception as e: | |
logger.warning(f"Error procesando video {path}: {str(e)}", exc_info=True) | |
if not clips_to_concatenate: | |
logger.error("No hay segmentos de video válidos disponibles para crear la secuencia.") | |
raise ValueError("No hay segmentos de video válidos disponibles para crear el video.") | |
logger.info(f"Concatenando {len(clips_to_concatenate)} segmentos de video.") | |
video_base = concatenate_videoclips(clips_to_concatenate, method="chain") | |
for seg in clips_to_concatenate: | |
try: seg.close() | |
except: pass | |
clips_to_concatenate = [] | |
if video_base.duration < audio_duration: | |
logger.info(f"Video base ({video_base.duration:.2f}s) más corto que audio ({audio_duration:.2f}s). Repitiendo...") | |
video_base = video_base.loop(duration=audio_duration) | |
if video_base.duration > audio_duration: | |
logger.info(f"Recortando video base ({video_base.duration:.2f}s) para coincidir con audio ({audio_duration:.2f}s).") | |
video_base = video_base.subclip(0, audio_duration) | |
if video_base is None or video_base.duration is None or video_base.duration <= 0: | |
raise ValueError("Video base final es inválido.") | |
# 6. Manejar música de fondo | |
logger.info("Procesando audio...") | |
final_audio = audio_tts_original | |
if musica_file: | |
try: | |
music_path = os.path.join(temp_dir_intermediate, "musica_bg.mp3") | |
shutil.copyfile(musica_file, music_path) | |
temp_intermediate_files.append(music_path) | |
musica_audio_original = AudioFileClip(music_path) | |
if musica_audio_original.duration > 0: | |
musica_audio = loop_audio_to_length(musica_audio_original, video_base.duration) | |
if musica_audio and musica_audio.duration > 0: | |
final_audio = CompositeAudioClip([musica_audio.volumex(0.2), audio_tts_original.volumex(1.0)]) | |
except Exception as e: | |
logger.warning(f"Error procesando música: {str(e)}", exc_info=True) | |
if final_audio.duration is not None and abs(final_audio.duration - video_base.duration) > 0.2: | |
logger.warning(f"Ajustando duración de audio final ({final_audio.duration:.2f}s) a la del video ({video_base.duration:.2f}s).") | |
final_audio = final_audio.subclip(0, video_base.duration) | |
# 7. Crear video final | |
logger.info("Renderizando video final...") | |
video_final = video_base.set_audio(final_audio) | |
output_filename = "final_video.mp4" | |
output_path = os.path.join(temp_dir_intermediate, output_filename) | |
logger.info(f"Escribiendo video final a: {output_path}") | |
video_final.write_videofile( | |
filename=output_path, fps=24, threads=4, codec="libx264", | |
audio_codec="aac", preset="medium", logger='bar' | |
) | |
total_time = (datetime.now() - start_time).total_seconds() | |
logger.info(f"PROCESO DE VIDEO FINALIZADO | Output: {output_path} | Tiempo total: {total_time:.2f}s") | |
return output_path | |
except Exception as e: | |
logger.critical(f"ERROR CRÍTICO en crear_video: {str(e)}", exc_info=True) | |
raise | |
finally: | |
logger.info("Iniciando limpieza de clips y archivos temporales intermedios...") | |
clips_to_close = [video_base, video_final, audio_tts_original, musica_audio_original, musica_audio] + source_clips + clips_to_concatenate | |
for clip in clips_to_close: | |
if clip: | |
try: clip.close() | |
except Exception: pass | |
if temp_dir_intermediate and os.path.exists(temp_dir_intermediate): | |
final_output_in_temp = os.path.join(temp_dir_intermediate, "final_video.mp4") | |
for path in temp_intermediate_files: | |
if os.path.isfile(path) and path != final_output_in_temp: | |
try: | |
os.remove(path) | |
logger.debug(f"Eliminando archivo temporal intermedio: {path}") | |
except Exception: pass | |
logger.info(f"Directorio temporal intermedio {temp_dir_intermediate} persistirá para que Gradio lea el video final.") | |
def run_app(prompt_type, prompt_ia, prompt_manual, musica_file): | |
logger.info("="*80) | |
logger.info("SOLICITUD RECIBIDA EN INTERFAZ") | |
input_text = prompt_ia if prompt_type == "Generar Guion con IA" else prompt_manual | |
output_video = None | |
output_file = gr.update(value=None, visible=False) | |
status_msg = gr.update(value="⏳ Procesando...", interactive=False) | |
if not input_text or not input_text.strip(): | |
logger.warning("Texto de entrada vacío.") | |
status_msg = gr.update(value="⚠️ Por favor, ingresa texto para el guion o el tema.", interactive=False) | |
return output_video, output_file, status_msg | |
try: | |
logger.info("Llamando a crear_video...") | |
video_path = crear_video(prompt_type, input_text, musica_file) | |
if video_path and os.path.exists(video_path): | |
logger.info(f"crear_video retornó path: {video_path}") | |
output_video = video_path | |
output_file = gr.update(value=video_path, visible=True) | |
status_msg = gr.update(value="✅ Video generado exitosamente.", interactive=False) | |
# *** MODIFICACIÓN: Programar la eliminación automática del directorio del video *** | |
temp_dir_to_delete = os.path.dirname(video_path) | |
delay_hours = 3 | |
deletion_thread = threading.Thread( | |
target=schedule_deletion, | |
args=(temp_dir_to_delete, delay_hours * 3600) | |
) | |
deletion_thread.daemon = True | |
deletion_thread.start() | |
else: | |
logger.error(f"crear_video no retornó un path válido o el archivo no existe: {video_path}") | |
status_msg = gr.update(value="❌ Error: La generación del video falló.", interactive=False) | |
except Exception as e: | |
logger.critical(f"Error crítico durante la creación del video: {str(e)}", exc_info=True) | |
status_msg = gr.update(value=f"❌ Error inesperado: {str(e)}", interactive=False) | |
finally: | |
logger.info("Fin del handler run_app.") | |
return output_video, output_file, status_msg | |
with gr.Blocks(title="Generador de Videos con IA", theme=gr.themes.Soft(), css=""" | |
.gradio-container {max-width: 800px; margin: auto;} | |
h1 {text-align: center;} | |
""") as app: | |
gr.Markdown("# 🎬 Generador Automático de Videos con IA") | |
gr.Markdown("Genera videos cortos a partir de un tema o guion, usando imágenes de archivo de Pexels y voz generada.") | |
with gr.Row(): | |
with gr.Column(): | |
prompt_type = gr.Radio( | |
["Generar Guion con IA", "Usar Mi Guion"], | |
label="Método de Entrada", | |
value="Generar Guion con IA" | |
) | |
with gr.Column(visible=True) as ia_guion_column: | |
prompt_ia = gr.Textbox( | |
label="Tema para IA", | |
lines=2, | |
placeholder="Ej: Un paisaje natural con montañas y ríos...", | |
max_lines=4 | |
) | |
with gr.Column(visible=False) as manual_guion_column: | |
prompt_manual = gr.Textbox( | |
label="Tu Guion Completo", | |
lines=5, | |
placeholder="Ej: En este video exploraremos los misterios del océano...", | |
max_lines=10 | |
) | |
musica_input = gr.Audio( | |
label="Música de fondo (opcional)", | |
type="filepath" | |
) | |
generate_btn = gr.Button("✨ Generar Video", variant="primary") | |
with gr.Column(): | |
video_output = gr.Video( | |
label="Previsualización del Video Generado", | |
interactive=False, | |
height=400 | |
) | |
file_output = gr.File( | |
label="Descargar Archivo de Video", | |
interactive=False, | |
visible=False | |
) | |
status_output = gr.Textbox( | |
label="Estado", | |
interactive=False, | |
show_label=False, | |
placeholder="Esperando acción..." | |
) | |
prompt_type.change( | |
lambda x: (gr.update(visible=x == "Generar Guion con IA"), | |
gr.update(visible=x == "Usar Mi Guion")), | |
inputs=prompt_type, | |
outputs=[ia_guion_column, manual_guion_column] | |
) | |
generate_btn.click( | |
lambda: (None, None, gr.update(value="⏳ Procesando... Esto puede tomar varios minutos.")), | |
outputs=[video_output, file_output, status_output], | |
queue=True, | |
).then( | |
run_app, | |
inputs=[prompt_type, prompt_ia, prompt_manual, musica_input], | |
outputs=[video_output, file_output, status_output] | |
) | |
gr.Markdown("### Instrucciones:") | |
gr.Markdown("1. **Clave API de Pexels:** Asegúrate de tener la variable de entorno `PEXELS_API_KEY`.\n" | |
"2. **Selecciona el método** y escribe tu tema o guion.\n" | |
"3. **Sube música** (opcional).\n" | |
"4. Haz clic en **Generar Video** y espera.\n" | |
"5. El video generado se eliminará automáticamente del servidor después de 3 horas.") | |
if __name__ == "__main__": | |
logger.info("Iniciando aplicación Gradio...") | |
try: | |
app.launch(server_name="0.0.0.0", server_port=7860, share=False) | |
except Exception as e: | |
logger.critical(f"No se pudo iniciar la app: {str(e)}", exc_info=True) | |
raise |