Spaces:
Running
Running
| import os | |
| import asyncio | |
| import logging | |
| import tempfile | |
| import requests | |
| from datetime import datetime | |
| import edge_tts | |
| import gradio as gr | |
| import torch | |
| from transformers import GPT2Tokenizer, GPT2LMHeadModel | |
| from keybert import KeyBERT | |
| 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 time | |
| # 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) | |
| # CAMBIO: Lista estática de voces como en script 2 | |
| VOICES = { | |
| "Español": [ | |
| "es-ES-JuanNeural", "es-ES-ElviraNeural", "es-MX-JorgeNeural", "es-MX-DaliaNeural", | |
| "es-AR-TomasNeural", "es-AR-ElenaNeural", "es-CO-GonzaloNeural", "es-CO-SalomeNeural", | |
| "es-CL-LorenzoNeural", "es-CL-CatalinaNeural", "es-PE-AlexNeural", "es-PE-CamilaNeural", | |
| "es-VE-PaolaNeural", "es-VE-SebastianNeural", "es-US-AlonsoNeural", "es-US-PalomaNeural" | |
| ] | |
| } | |
| # CAMBIO: Función para obtener las voces en formato de dropdown como en script 2 | |
| def get_voice_choices(): | |
| choices = [] | |
| for region, voice_list in VOICES.items(): | |
| for voice_id in voice_list: | |
| choices.append((f"{voice_id} ({region})", voice_id)) | |
| return choices | |
| # CAMBIO: Usamos lista estática en lugar de edge_tts.list_voices() | |
| AVAILABLE_VOICES = get_voice_choices() | |
| DEFAULT_VOICE_ID = "es-ES-JuanNeural" | |
| DEFAULT_VOICE_NAME = next((name for name, vid in AVAILABLE_VOICES if vid == DEFAULT_VOICE_ID), DEFAULT_VOICE_ID) | |
| logger.info(f"Voz por defecto seleccionada: {DEFAULT_VOICE_NAME} ({DEFAULT_VOICE_ID})") | |
| # 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 con reintentos (como en script 1b) | |
| MODEL_NAME = "datificate/gpt2-small-spanish" | |
| logger.info(f"Inicializando modelo GPT-2: {MODEL_NAME}") | |
| tokenizer = None | |
| model = None | |
| MAX_RETRIES = 3 | |
| for attempt in range(MAX_RETRIES): | |
| try: | |
| tokenizer = GPT2Tokenizer.from_pretrained(MODEL_NAME, timeout=60) | |
| model = GPT2LMHeadModel.from_pretrained(MODEL_NAME, timeout=60).eval() | |
| if tokenizer.pad_token is None: | |
| tokenizer.pad_token = tokenizer.eos_token | |
| logger.info(f"Modelo GPT-2 cargado | Vocabulario: {len(tokenizer)} tokens") | |
| break | |
| except Exception as e: | |
| logger.warning(f"Intento {attempt + 1} falló al cargar GPT-2: {str(e)}") | |
| if attempt < MAX_RETRIES - 1: | |
| time.sleep(5) | |
| else: | |
| logger.error(f"FALLA CRÍTICA al cargar GPT-2 después de {MAX_RETRIES} intentos: {str(e)}") | |
| 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 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} | |
| for attempt in range(MAX_RETRIES): | |
| 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=60 | |
| ) | |
| 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.warning(f"Intento {attempt + 1} falló en Pexels: {str(e)}") | |
| if attempt < MAX_RETRIES - 1: | |
| time.sleep(5) | |
| else: | |
| logger.error(f"Error de conexión Pexels para '{query}' después de {MAX_RETRIES} intentos: {str(e)}") | |
| 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: | |
| prompt_in_output_idx = text.lower().find(prompt.lower()) | |
| if prompt_in_output_idx != -1: | |
| cleaned_text = text[prompt_in_output_idx + len(prompt):].strip() | |
| logger.debug("Texto limpiado tomando parte después del prompt original.") | |
| else: | |
| instruction_start_idx = text.find(instruction_phrase_start) | |
| if instruction_start_idx != -1: | |
| cleaned_text = text[instruction_start_idx + len(instruction_phrase_start):].strip() | |
| logger.debug("Texto limpiado tomando parte después de la frase de instrucción base.") | |
| else: | |
| logger.warning("No se pudo identificar el inicio del guión generado. Usando texto generado completo.") | |
| cleaned_text = text.strip() | |
| except Exception as e: | |
| logger.warning(f"Error durante la limpieza heurística del guion 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 heurística. Usando el texto generado original (sin limpieza adicional).") | |
| 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: {str(e)}", exc_info=True) | |
| logger.warning("Usando prompt original como guion debido al error de generación.") | |
| return prompt.strip() | |
| async def text_to_speech(text, output_path, voice): | |
| logger.info(f"Convirtiendo texto a voz | Caracteres: {len(text)} | Voz: {voice} | Salida: {output_path}") | |
| if not text or not text.strip(): | |
| logger.warning("Texto vacío para TTS") | |
| return False | |
| try: | |
| communicate = edge_tts.Communicate(text, voice) | |
| await communicate.save(output_path) | |
| if os.path.exists(output_path) and os.path.getsize(output_path) > 100: | |
| logger.info(f"Audio guardado exitosamente en: {output_path} | Tamaño: {os.path.getsize(output_path)} bytes") | |
| return True | |
| else: | |
| logger.error(f"TTS guardó un archivo pequeño o vacío en: {output_path}") | |
| return False | |
| except Exception as e: | |
| logger.error(f"Error en TTS con voz '{voice}': {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) | |
| for attempt in range(MAX_RETRIES): | |
| try: | |
| 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]}...") | |
| if os.path.exists(output_path): | |
| os.remove(output_path) | |
| return None | |
| except requests.exceptions.RequestException as e: | |
| logger.warning(f"Intento {attempt + 1} falló al descargar video: {str(e)}") | |
| if attempt < MAX_RETRIES - 1: | |
| time.sleep(5) | |
| else: | |
| logger.error(f"Error descargando video después de {MAX_RETRIES} intentos: {str(e)}") | |
| return None | |
| 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, selected_voice, musica_file=None): | |
| logger.info("="*80) | |
| logger.info(f"INICIANDO CREACIÓN DE VIDEO | Tipo: {prompt_type}") | |
| logger.debug(f"Input: '{input_text[:100]}...'") | |
| logger.info(f"Voz seleccionada: {selected_voice}") | |
| # CAMBIO: Validación simple de la voz como en script 2 | |
| voice_ids = [vid for _, vid in AVAILABLE_VOICES] | |
| if selected_voice not in voice_ids: | |
| logger.warning(f"Voz seleccionada '{selected_voice}' no es válida. Usando voz por defecto: {DEFAULT_VOICE_ID}") | |
| selected_voice = DEFAULT_VOICE_ID | |
| start_time = datetime.now() | |
| temp_dir_intermediate = None | |
| 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 = [] | |
| temp_intermediate_files = [] | |
| try: | |
| 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.") | |
| temp_dir_intermediate = tempfile.mkdtemp(prefix="video_gen_intermediate_") | |
| logger.info(f"Directorio temporal intermedio creado: {temp_dir_intermediate}") | |
| logger.info("Generando audio de voz...") | |
| voz_path = os.path.join(temp_dir_intermediate, "voz.mp3") | |
| # CAMBIO: Usamos la voz seleccionada directamente, sin fallbacks complejos | |
| logger.info(f"Intentando TTS con voz: {selected_voice}...") | |
| tts_success = asyncio.run(text_to_speech(guion, voz_path, voice=selected_voice)) | |
| if not tts_success: | |
| 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.") | |
| try: audio_tts_original.close() | |
| except: pass | |
| raise ValueError("Audio de voz generado es inválido.") | |
| 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.") | |
| 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"] | |
| 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.") | |
| generic_keywords = ["nature", "city", "background", "abstract"] | |
| for keyword in generic_keywords: | |
| 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.") | |
| 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']: | |
| logger.debug(f"Saltando video sin archivos de video: {video.get('id')}") | |
| continue | |
| try: | |
| best_quality = None | |
| for vf in sorted(video['video_files'], key=lambda x: x.get('width', 0) * x.get('height', 0), reverse=True): | |
| if 'link' in vf: | |
| best_quality = vf | |
| break | |
| if best_quality and 'link' in 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.") | |
| 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: | |
| logger.debug(f"Video base suficiente.") | |
| break | |
| clip = None | |
| try: | |
| logger.debug(f"[{i+1}/{len(video_paths)}] Abriendo clip: {path}") | |
| clip = VideoFileClip(path) | |
| source_clips.append(clip) | |
| if clip.reader is None or clip.duration is None or clip.duration <= 0: | |
| logger.warning(f"[{i+1}/{len(video_paths)}] Clip fuente {path} parece inválido.") | |
| continue | |
| remaining_needed = audio_duration - current_duration | |
| potential_use_duration = min(clip.duration, max_clip_segment) | |
| if remaining_needed > 0: | |
| segment_duration = min(potential_use_duration, remaining_needed + min_clip_duration) | |
| segment_duration = max(min_clip_duration, segment_duration) | |
| segment_duration = min(segment_duration, clip.duration) | |
| if segment_duration >= min_clip_duration: | |
| try: | |
| sub = clip.subclip(0, segment_duration) | |
| if sub.reader is None or sub.duration is None or sub.duration <= 0: | |
| logger.warning(f"[{i+1}/{len(video_paths)}] Subclip inválido.") | |
| try: sub.close() | |
| except: pass | |
| continue | |
| clips_to_concatenate.append(sub) | |
| current_duration += sub.duration | |
| except Exception as sub_e: | |
| logger.warning(f"[{i+1}/{len(video_paths)}] Error creando subclip: {str(sub_e)}") | |
| continue | |
| except Exception as e: | |
| logger.warning(f"[{i+1}/{len(video_paths)}] Error procesando video {path}: {str(e)}", exc_info=True) | |
| continue | |
| logger.info(f"Procesamiento de clips fuente finalizado. Se obtuvieron {len(clips_to_concatenate)} segmentos válidos.") | |
| if not clips_to_concatenate: | |
| logger.error("No hay segmentos de video válidos disponibles.") | |
| raise ValueError("No hay segmentos de video válidos.") | |
| logger.info(f"Concatenando {len(clips_to_concatenate)} segmentos de video.") | |
| concatenated_base = None | |
| try: | |
| concatenated_base = concatenate_videoclips(clips_to_concatenate, method="chain") | |
| logger.info(f"Duración video base después de concatenación inicial: {concatenated_base.duration:.2f}s") | |
| if concatenated_base is None or concatenated_base.duration is None or concatenated_base.duration <= 0: | |
| logger.critical("Video base concatenado es inválido.") | |
| raise ValueError("Fallo al crear video base válido.") | |
| except Exception as e: | |
| logger.critical(f"Error durante la concatenación inicial: {str(e)}", exc_info=True) | |
| raise ValueError("Fallo durante la concatenación de video inicial.") | |
| finally: | |
| for clip_segment in clips_to_concatenate: | |
| try: clip_segment.close() | |
| except: pass | |
| clips_to_concatenate = [] | |
| video_base = concatenated_base | |
| final_video_base = video_base | |
| if final_video_base.duration < audio_duration: | |
| logger.info(f"Video base ({final_video_base.duration:.2f}s) es más corto que el audio ({audio_duration:.2f}s). Repitiendo...") | |
| num_full_repeats = int(audio_duration // final_video_base.duration) | |
| remaining_duration = audio_duration % final_video_base.duration | |
| repeated_clips_list = [final_video_base] * num_full_repeats | |
| if remaining_duration > 0: | |
| try: | |
| remaining_clip = final_video_base.subclip(0, remaining_duration) | |
| if remaining_clip is None or remaining_clip.duration is None or remaining_clip.duration <= 0: | |
| logger.warning(f"Subclip generado para duración restante {remaining_duration:.2f}s es inválido.") | |
| try: remaining_clip.close() | |
| except: pass | |
| else: | |
| repeated_clips_list.append(remaining_clip) | |
| except Exception as e: | |
| logger.warning(f"Error creando subclip para duración restante {remaining_duration:.2f}s: {str(e)}") | |
| if repeated_clips_list: | |
| logger.info(f"Concatenando {len(repeated_clips_list)} partes para repetición.") | |
| video_base_repeated = None | |
| try: | |
| video_base_repeated = concatenate_videoclips(repeated_clips_list, method="chain") | |
| logger.info(f"Duración del video base repetido: {video_base_repeated.duration:.2f}s") | |
| if video_base_repeated is None or video_base_repeated.duration is None or video_base_repeated.duration <= 0: | |
| logger.critical("Video base repetido concatenado es inválido.") | |
| raise ValueError("Fallo al crear video base repetido válido.") | |
| if final_video_base is not video_base_repeated: | |
| try: final_video_base.close() | |
| except: pass | |
| final_video_base = video_base_repeated | |
| except Exception as e: | |
| logger.critical(f"Error durante la concatenación de repetición: {str(e)}", exc_info=True) | |
| raise ValueError("Fallo durante la repetición de video.") | |
| finally: | |
| for clip in repeated_clips_list: | |
| if clip is not final_video_base: | |
| try: clip.close() | |
| except: pass | |
| if final_video_base.duration > audio_duration: | |
| logger.info(f"Recortando video base ({final_video_base.duration:.2f}s) para coincidir con audio ({audio_duration:.2f}s).") | |
| trimmed_video_base = None | |
| try: | |
| trimmed_video_base = final_video_base.subclip(0, audio_duration) | |
| if trimmed_video_base is None or trimmed_video_base.duration is None or trimmed_video_base.duration <= 0: | |
| logger.critical("Video base recortado es inválido.") | |
| raise ValueError("Fallo al crear video base recortado válido.") | |
| if final_video_base is not trimmed_video_base: | |
| try: final_video_base.close() | |
| except: pass | |
| final_video_base = trimmed_video_base | |
| except Exception as e: | |
| logger.critical(f"Error durante el recorte: {str(e)}", exc_info=True) | |
| raise ValueError("Fallo durante el recorte de video.") | |
| if final_video_base is None or final_video_base.duration is None or final_video_base.duration <= 0: | |
| logger.critical("Video base final es inválido.") | |
| raise ValueError("Video base final es inválido.") | |
| if final_video_base.size is None or final_video_base.size[0] <= 0 or final_video_base.size[1] <= 0: | |
| logger.critical(f"Video base final tiene tamaño inválido: {final_video_base.size}.") | |
| raise ValueError("Video base final tiene tamaño inválido.") | |
| video_base = final_video_base | |
| logger.info("Procesando audio...") | |
| final_audio = audio_tts_original | |
| musica_audio_looped = None | |
| if musica_file: | |
| musica_audio_original = None | |
| try: | |
| music_path = os.path.join(temp_dir_intermediate, "musica_bg.mp3") | |
| shutil.copyfile(musica_file, music_path) | |
| temp_intermediate_files.append(music_path) | |
| logger.info(f"Música de fondo copiada a: {music_path}") | |
| musica_audio_original = AudioFileClip(music_path) | |
| if musica_audio_original.reader is None or musica_audio_original.duration is None or musica_audio_original.duration <= 0: | |
| logger.warning("Clip de música de fondo parece inválido.") | |
| try: musica_audio_original.close() | |
| except: pass | |
| musica_audio_original = None | |
| else: | |
| musica_audio_looped = loop_audio_to_length(musica_audio_original, video_base.duration) | |
| logger.debug(f"Música ajustada a duración del video: {musica_audio_looped.duration:.2f}s") | |
| if musica_audio_looped is None or musica_audio_looped.duration is None or musica_audio_looped.duration <= 0: | |
| logger.warning("Clip de música de fondo loopeado es inválido.") | |
| try: musica_audio_looped.close() | |
| except: pass | |
| musica_audio_looped = None | |
| if musica_audio_looped: | |
| composite_audio = CompositeAudioClip([ | |
| musica_audio_looped.volumex(0.2), | |
| audio_tts_original.volumex(1.0) | |
| ]) | |
| if composite_audio.duration is None or composite_audio.duration <= 0: | |
| logger.warning("Clip de audio compuesto es inválido.") | |
| try: composite_audio.close() | |
| except: pass | |
| final_audio = audio_tts_original | |
| else: | |
| logger.info("Mezcla de audio completada (voz + música).") | |
| final_audio = composite_audio | |
| musica_audio = musica_audio_looped | |
| except Exception as e: | |
| logger.warning(f"Error procesando música de fondo: {str(e)}", exc_info=True) | |
| final_audio = audio_tts_original | |
| musica_audio = None | |
| logger.warning("Usando solo audio de voz debido a un error con la música.") | |
| if final_audio.duration is not None and abs(final_audio.duration - video_base.duration) > 0.2: | |
| logger.warning(f"Duración del audio final ({final_audio.duration:.2f}s) difiere del video base ({video_base.duration:.2f}s).") | |
| try: | |
| if final_audio.duration > video_base.duration: | |
| trimmed_final_audio = final_audio.subclip(0, video_base.duration) | |
| if trimmed_final_audio is None or trimmed_final_audio.duration <= 0: | |
| logger.warning("Audio final recortado es inválido.") | |
| try: trimmed_final_audio.close() | |
| except: pass | |
| else: | |
| if final_audio is not trimmed_final_audio: | |
| try: final_audio.close() | |
| except: pass | |
| final_audio = trimmed_final_audio | |
| except Exception as e: | |
| logger.warning(f"Error ajustando duración del audio final: {str(e)}") | |
| logger.info("Renderizando video final...") | |
| video_final = video_base.set_audio(final_audio) | |
| if video_final is None or video_final.duration is None or video_final.duration <= 0: | |
| logger.critical("Clip de video final (con audio) es inválido.") | |
| raise ValueError("Clip de video final es inválido.") | |
| 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( | |
| 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 ValueError as ve: | |
| logger.error(f"ERROR CONTROLADO en crear_video: {str(ve)}") | |
| raise ve | |
| except Exception as e: | |
| logger.critical(f"ERROR CRÍTICO NO CONTROLADO en crear_video: {str(e)}", exc_info=True) | |
| raise e | |
| finally: | |
| logger.info("Iniciando limpieza de clips y archivos temporales intermedios...") | |
| for clip in source_clips: | |
| try: | |
| clip.close() | |
| except Exception as e: | |
| logger.warning(f"Error cerrando clip de video fuente: {str(e)}") | |
| for clip_segment in clips_to_concatenate: | |
| try: | |
| clip_segment.close() | |
| except Exception as e: | |
| logger.warning(f"Error cerrando segmento de video: {str(e)}") | |
| if musica_audio is not None: | |
| try: | |
| musica_audio.close() | |
| except Exception as e: | |
| logger.warning(f"Error cerrando musica_audio: {str(e)}") | |
| if musica_audio_original is not None and musica_audio_original is not musica_audio: | |
| try: | |
| musica_audio_original.close() | |
| except Exception as e: | |
| logger.warning(f"Error cerrando musica_audio_original: {str(e)}") | |
| if audio_tts is not None and audio_tts is not audio_tts_original: | |
| try: | |
| audio_tts.close() | |
| except Exception as e: | |
| logger.warning(f"Error cerrando audio_tts: {str(e)}") | |
| if audio_tts_original is not None: | |
| try: | |
| audio_tts_original.close() | |
| except Exception as e: | |
| logger.warning(f"Error cerrando audio_tts_original: {str(e)}") | |
| if video_final is not None: | |
| try: | |
| video_final.close() | |
| except Exception as e: | |
| logger.warning(f"Error cerrando video_final: {str(e)}") | |
| elif video_base is not None and video_base is not video_final: | |
| try: | |
| video_base.close() | |
| except Exception as e: | |
| logger.warning(f"Error cerrando video_base: {str(e)}") | |
| 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: | |
| try: | |
| if os.path.isfile(path) and path != final_output_in_temp: | |
| logger.debug(f"Eliminando archivo temporal intermedio: {path}") | |
| os.remove(path) | |
| except Exception as e: | |
| logger.warning(f"No se pudo eliminar archivo temporal intermedio {path}: {str(e)}") | |
| def run_app(prompt_type, prompt_ia, prompt_manual, musica_file, selected_voice): | |
| 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 = None | |
| status_msg = gr.update(value="⏳ Procesando...", interactive=False) | |
| if not input_text or not input_text.strip(): | |
| logger.warning("Texto de entrada vacío.") | |
| return None, None, gr.update(value="⚠️ Por favor, ingresa texto para el guion o el tema.", interactive=False) | |
| # CAMBIO: Validación simple de la voz en run_app | |
| voice_ids = [vid for _, vid in AVAILABLE_VOICES] | |
| if selected_voice not in voice_ids: | |
| logger.warning(f"Voz seleccionada inválida: '{selected_voice}'. Usando voz por defecto: {DEFAULT_VOICE_ID}") | |
| selected_voice = DEFAULT_VOICE_ID | |
| logger.info(f"Voz final a usar (ID): {selected_voice}") | |
| try: | |
| logger.info("Llamando a crear_video...") | |
| video_path = crear_video(prompt_type, input_text, selected_voice, 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 = video_path | |
| status_msg = gr.update(value="✅ Video generado exitosamente.", interactive=False) | |
| else: | |
| logger.error(f"crear_video no retornó un path válido: {video_path}") | |
| status_msg = gr.update(value="❌ Error: La generación del video falló o el archivo no se creó correctamente.", interactive=False) | |
| except ValueError as ve: | |
| logger.warning(f"Error de validación durante la creación del video: {str(ve)}") | |
| status_msg = gr.update(value=f"⚠️ Error de validación: {str(ve)}", 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) | |
| return output_video, output_file, status_msg | |
| # Interfaz de Gradio | |
| 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 al amanecer, mostrando la belleza de la naturaleza...", | |
| max_lines=4, | |
| value="" | |
| ) | |
| 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. Veremos la vida marina fascinante y los arrecifes de coral vibrantes. ¡Acompáñanos en esta aventura subacuática!", | |
| max_lines=10, | |
| value="" | |
| ) | |
| musica_input = gr.Audio( | |
| label="Música de fondo (opcional)", | |
| type="filepath", | |
| interactive=True, | |
| value=None | |
| ) | |
| # CAMBIO: Dropdown de voces simplificado | |
| voice_dropdown = gr.Dropdown( | |
| label="Seleccionar Voz para Guion", | |
| choices=AVAILABLE_VOICES, | |
| value=DEFAULT_VOICE_ID, | |
| interactive=True | |
| ) | |
| 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...", | |
| value="Esperando entrada..." | |
| ) | |
| 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.", interactive=False)), | |
| outputs=[video_output, file_output, status_output], | |
| queue=True, | |
| ).then( | |
| run_app, | |
| inputs=[prompt_type, prompt_ia, prompt_manual, musica_input, voice_dropdown], | |
| outputs=[video_output, file_output, status_output] | |
| ).then( | |
| lambda video_path, file_path, status_msg: gr.update(visible=file_path is not None), | |
| inputs=[video_output, file_output, status_output], | |
| outputs=[file_output] | |
| ) | |
| gr.Markdown("### Instrucciones:") | |
| gr.Markdown(""" | |
| 1. **Clave API de Pexels:** Asegúrate de haber configurado la variable de entorno `PEXELS_API_KEY` con tu clave. | |
| 2. **Selecciona el tipo de entrada**: "Generar Guion con IA" o "Usar Mi Guion". | |
| 3. **Sube música** (opcional): Selecciona un archivo de audio (MP3, WAV, etc.). | |
| 4. **Selecciona la voz** deseada del desplegable. | |
| 5. **Haz clic en "✨ Generar Video"**. | |
| 6. Espera a que se procese el video. Verás el estado. | |
| 7. La previsualización aparecerá si es posible, y siempre un enlace **Descargar Archivo de Video** se mostrará si la generación fue exitosa. | |
| 8. Revisa `video_generator_full.log` para detalles si hay errores. | |
| """) | |
| gr.Markdown("---") | |
| gr.Markdown("Desarrollado por [Tu Nombre/Empresa/Alias - Opcional]") | |
| if __name__ == "__main__": | |
| logger.info("Verificando dependencias críticas...") | |
| try: | |
| from moviepy.editor import ColorClip | |
| try: | |
| temp_clip = ColorClip((100,100), color=(255,0,0), duration=0.1) | |
| temp_clip.close() | |
| logger.info("Clips base de MoviePy creados y cerrados exitosamente. FFmpeg parece accesible.") | |
| except Exception as e: | |
| logger.critical(f"Fallo al crear clip base de MoviePy. A menudo indica problemas con FFmpeg/ImageMagick. Error: {e}", exc_info=True) | |
| except Exception as e: | |
| logger.critical(f"Fallo al importar MoviePy. Asegúrate de que está instalado. Error: {e}", exc_info=True) | |
| 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 |