Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -9,7 +9,6 @@ import gradio as gr
|
|
9 |
import torch
|
10 |
from transformers import GPT2Tokenizer, GPT2LMHeadModel
|
11 |
from keybert import KeyBERT
|
12 |
-
# Importación correcta
|
13 |
from moviepy.editor import VideoFileClip, concatenate_videoclips, AudioFileClip, CompositeAudioClip, concatenate_audioclips, AudioClip
|
14 |
import re
|
15 |
import math
|
@@ -32,7 +31,6 @@ logger.info("INICIO DE EJECUCIÓN - GENERADOR DE VIDEOS")
|
|
32 |
logger.info("="*80)
|
33 |
|
34 |
# Diccionario de voces TTS disponibles organizadas por idioma
|
35 |
-
# Puedes expandir esta lista si conoces otros IDs de voz de Edge TTS
|
36 |
VOCES_DISPONIBLES = {
|
37 |
"Español (España)": {
|
38 |
"es-ES-JuanNeural": "Juan (España) - Masculino",
|
@@ -100,37 +98,26 @@ def get_voice_choices():
|
|
100 |
choices = []
|
101 |
for region, voices in VOCES_DISPONIBLES.items():
|
102 |
for voice_id, voice_name in voices.items():
|
103 |
-
# Formato: (Texto a mostrar en el dropdown, Valor que se pasa)
|
104 |
choices.append((f"{voice_name} ({region})", voice_id))
|
105 |
return choices
|
106 |
|
107 |
# Obtener las voces al inicio del script
|
108 |
-
|
109 |
-
|
110 |
-
# AVAILABLE_VOICES = asyncio.run(get_available_voices())
|
111 |
-
AVAILABLE_VOICES = get_voice_choices() # <-- Usamos la lista predefinida y aplanada
|
112 |
-
# Establecer una voz por defecto inicial
|
113 |
-
DEFAULT_VOICE_ID = "es-ES-JuanNeural" # ID de Juan
|
114 |
-
|
115 |
-
# Buscar el nombre amigable para la voz por defecto si existe
|
116 |
DEFAULT_VOICE_NAME = DEFAULT_VOICE_ID
|
117 |
for text, voice_id in AVAILABLE_VOICES:
|
118 |
if voice_id == DEFAULT_VOICE_ID:
|
119 |
DEFAULT_VOICE_NAME = text
|
120 |
break
|
121 |
-
# Si Juan no está en la lista (ej. lista de fallback), usar la primera voz disponible
|
122 |
if DEFAULT_VOICE_ID not in [v[1] for v in AVAILABLE_VOICES]:
|
123 |
DEFAULT_VOICE_ID = AVAILABLE_VOICES[0][1] if AVAILABLE_VOICES else "en-US-AriaNeural"
|
124 |
-
DEFAULT_VOICE_NAME = AVAILABLE_VOICES[0][0] if AVAILABLE_VOICES else "Aria (United States) - Female"
|
125 |
-
|
126 |
logger.info(f"Voz por defecto seleccionada (ID): {DEFAULT_VOICE_ID}")
|
127 |
|
128 |
-
|
129 |
# Clave API de Pexels
|
130 |
PEXELS_API_KEY = os.environ.get("PEXELS_API_KEY")
|
131 |
if not PEXELS_API_KEY:
|
132 |
logger.critical("NO SE ENCONTRÓ PEXELS_API_KEY EN VARIABLES DE ENTORNO")
|
133 |
-
# raise ValueError("API key de Pexels no configurada")
|
134 |
|
135 |
# Inicialización de modelos
|
136 |
MODEL_NAME = "datificate/gpt2-small-spanish"
|
@@ -170,7 +157,6 @@ def buscar_videos_pexels(query, api_key, per_page=5):
|
|
170 |
"orientation": "landscape",
|
171 |
"size": "medium"
|
172 |
}
|
173 |
-
|
174 |
response = requests.get(
|
175 |
"https://api.pexels.com/videos/search",
|
176 |
headers=headers,
|
@@ -178,20 +164,19 @@ def buscar_videos_pexels(query, api_key, per_page=5):
|
|
178 |
timeout=20
|
179 |
)
|
180 |
response.raise_for_status()
|
181 |
-
|
182 |
data = response.json()
|
183 |
videos = data.get('videos', [])
|
184 |
logger.info(f"Pexels: {len(videos)} videos encontrados para '{query}'")
|
185 |
return videos
|
186 |
-
|
187 |
except requests.exceptions.RequestException as e:
|
188 |
logger.error(f"Error de conexión Pexels para '{query}': {str(e)}")
|
|
|
189 |
except json.JSONDecodeError:
|
190 |
-
logger.error(f"Pexels: JSON inválido recibido | Status: {response.status_code}
|
|
|
191 |
except Exception as e:
|
192 |
-
logger.error(f"Error inesperado Pexels para '{query}': {str(e)}"
|
193 |
-
|
194 |
-
return []
|
195 |
|
196 |
def generate_script(prompt, max_length=150):
|
197 |
logger.info(f"Generando guión | Prompt: '{prompt[:50]}...' | Longitud máxima: {max_length}")
|
@@ -207,7 +192,6 @@ def generate_script(prompt, max_length=150):
|
|
207 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
208 |
model.to(device)
|
209 |
inputs = {k: v.to(device) for k, v in inputs.items()}
|
210 |
-
|
211 |
outputs = model.generate(
|
212 |
**inputs,
|
213 |
max_length=max_length + inputs[list(inputs.keys())[0]].size(1),
|
@@ -220,247 +204,147 @@ def generate_script(prompt, max_length=150):
|
|
220 |
eos_token_id=tokenizer.eos_token_id,
|
221 |
no_repeat_ngram_size=3
|
222 |
)
|
223 |
-
|
224 |
text = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
225 |
-
|
226 |
-
|
227 |
-
|
228 |
-
|
229 |
-
|
230 |
-
|
231 |
-
if
|
232 |
-
|
233 |
-
|
234 |
-
logger.debug("Texto limpiado tomando parte después del prompt original.")
|
235 |
else:
|
236 |
-
|
237 |
-
|
238 |
-
if instruction_start_idx != -1:
|
239 |
-
# Tomar texto después de la frase base (puede incluir el prompt)
|
240 |
-
cleaned_text = text[instruction_start_idx + len(instruction_phrase_start):].strip()
|
241 |
-
logger.debug("Texto limpiado tomando parte después de la frase de instrucción base.")
|
242 |
-
else:
|
243 |
-
# Si ni la frase de instrucción ni el prompt se encuentran, usar el texto original
|
244 |
-
logger.warning("No se pudo identificar el inicio del guión generado. Usando texto generado completo.")
|
245 |
-
cleaned_text = text.strip() # Limpieza básica
|
246 |
-
|
247 |
-
|
248 |
-
except Exception as e:
|
249 |
-
logger.warning(f"Error durante la limpieza heurística del guión de IA: {e}. Usando texto generado sin limpieza adicional.")
|
250 |
-
cleaned_text = re.sub(r'<[^>]+>', '', text).strip() # Limpieza básica como fallback
|
251 |
-
|
252 |
-
# Asegurarse de que el texto resultante no sea solo la instrucción o vacío
|
253 |
-
if not cleaned_text or len(cleaned_text) < 10: # Umbral de longitud mínima
|
254 |
-
logger.warning("El guión generado parece muy corto o vacío después de la limpieza heurística. Usando el texto generado original (sin limpieza adicional).")
|
255 |
-
cleaned_text = re.sub(r'<[^>]+>', '', text).strip() # Fallback al texto original limpio
|
256 |
-
|
257 |
-
# Limpieza final de caracteres especiales y espacios sobrantes
|
258 |
cleaned_text = re.sub(r'<[^>]+>', '', cleaned_text).strip()
|
259 |
-
cleaned_text = cleaned_text.lstrip(':').strip()
|
260 |
-
cleaned_text = cleaned_text.lstrip('.').strip() # Quitar posibles '.' al inicio
|
261 |
-
|
262 |
-
|
263 |
-
# Intentar obtener al menos una oración completa si es posible para un inicio más limpio
|
264 |
sentences = cleaned_text.split('.')
|
265 |
if sentences and sentences[0].strip():
|
266 |
final_text = sentences[0].strip() + '.'
|
267 |
-
|
268 |
-
|
269 |
-
|
270 |
-
final_text = final_text.replace("..", ".") # Limpiar doble punto
|
271 |
-
|
272 |
logger.info(f"Guion generado final (Truncado a 100 chars): '{final_text[:100]}...'")
|
273 |
return final_text.strip()
|
274 |
-
|
275 |
-
|
276 |
-
return cleaned_text.strip() # Si no se puede formar una oración, devolver el texto limpio tal cual
|
277 |
-
|
278 |
except Exception as e:
|
279 |
-
logger.error(f"Error generando guion con GPT-2
|
280 |
-
logger.warning("Usando prompt original como guion debido al error de generación.")
|
281 |
return prompt.strip()
|
282 |
|
283 |
-
# Función TTS ahora recibe la voz a usar
|
284 |
async def text_to_speech(text, output_path, voice):
|
285 |
-
logger.info(f"Convirtiendo texto a voz | Caracteres: {len(text)} | Voz: {voice}
|
286 |
if not text or not text.strip():
|
287 |
logger.warning("Texto vacío para TTS")
|
288 |
return False
|
289 |
-
|
290 |
try:
|
291 |
communicate = edge_tts.Communicate(text, voice)
|
292 |
await communicate.save(output_path)
|
293 |
-
|
294 |
if os.path.exists(output_path) and os.path.getsize(output_path) > 100:
|
295 |
-
logger.info(f"Audio guardado exitosamente en: {output_path}
|
296 |
return True
|
297 |
-
|
298 |
-
|
299 |
-
return False
|
300 |
-
|
301 |
except Exception as e:
|
302 |
-
logger.error(f"Error en TTS con voz '{voice}': {str(e)}"
|
303 |
return False
|
304 |
|
305 |
def download_video_file(url, temp_dir):
|
306 |
if not url:
|
307 |
-
logger.warning("URL de video no proporcionada
|
308 |
return None
|
309 |
-
|
310 |
try:
|
311 |
logger.info(f"Descargando video desde: {url[:80]}...")
|
312 |
os.makedirs(temp_dir, exist_ok=True)
|
313 |
file_name = f"video_dl_{datetime.now().strftime('%Y%m%d_%H%M%S_%f')}.mp4"
|
314 |
output_path = os.path.join(temp_dir, file_name)
|
315 |
-
|
316 |
with requests.get(url, stream=True, timeout=60) as r:
|
317 |
r.raise_for_status()
|
318 |
with open(output_path, 'wb') as f:
|
319 |
for chunk in r.iter_content(chunk_size=8192):
|
320 |
f.write(chunk)
|
321 |
-
|
322 |
if os.path.exists(output_path) and os.path.getsize(output_path) > 1000:
|
323 |
-
|
324 |
-
|
325 |
-
|
326 |
-
|
327 |
-
|
328 |
-
|
329 |
-
return None
|
330 |
-
|
331 |
except requests.exceptions.RequestException as e:
|
332 |
-
logger.error(f"Error de descarga para {url[:80]}
|
|
|
333 |
except Exception as e:
|
334 |
-
logger.error(f"Error inesperado descargando {url[:80]}
|
335 |
-
|
336 |
-
return None
|
337 |
|
338 |
def loop_audio_to_length(audio_clip, target_duration):
|
339 |
logger.debug(f"Ajustando audio | Duración actual: {audio_clip.duration:.2f}s | Objetivo: {target_duration:.2f}s")
|
340 |
-
|
341 |
if audio_clip is None or audio_clip.duration is None or audio_clip.duration <= 0:
|
342 |
-
logger.warning("Input audio clip is invalid
|
343 |
-
|
344 |
-
|
345 |
-
return AudioClip(lambda t: 0, duration=target_duration, sr=sr)
|
346 |
-
except Exception as e:
|
347 |
-
logger.error(f"Could not create silence clip: {e}", exc_info=True)
|
348 |
-
return AudioFileClip(filename="")
|
349 |
-
|
350 |
if audio_clip.duration >= target_duration:
|
351 |
-
logger.debug("Audio clip
|
352 |
-
|
353 |
-
if trimmed_clip.duration is None or trimmed_clip.duration <= 0:
|
354 |
-
logger.error("Trimmed audio clip is invalid.")
|
355 |
-
try: trimmed_clip.close()
|
356 |
-
except: pass
|
357 |
-
return AudioFileClip(filename="")
|
358 |
-
return trimmed_clip
|
359 |
-
|
360 |
loops = math.ceil(target_duration / audio_clip.duration)
|
361 |
logger.debug(f"Creando {loops} loops de audio")
|
362 |
-
|
363 |
-
audio_segments = [audio_clip] * loops
|
364 |
-
looped_audio = None
|
365 |
-
final_looped_audio = None
|
366 |
try:
|
367 |
-
|
368 |
-
|
369 |
-
|
370 |
-
|
371 |
-
raise ValueError("Invalid concatenated audio.")
|
372 |
-
|
373 |
-
final_looped_audio = looped_audio.subclip(0, target_duration)
|
374 |
-
|
375 |
-
if final_looped_audio.duration is None or final_looped_audio.duration <= 0:
|
376 |
-
logger.error("Final subclipped audio clip is invalid (None or zero duration).")
|
377 |
-
raise ValueError("Invalid final subclipped audio.")
|
378 |
-
|
379 |
-
return final_looped_audio
|
380 |
-
|
381 |
except Exception as e:
|
382 |
-
logger.error(f"Error
|
383 |
-
|
384 |
-
if audio_clip.duration is not None and audio_clip.duration > 0:
|
385 |
-
logger.warning("Returning original audio clip (may be too short).")
|
386 |
-
return audio_clip.subclip(0, min(audio_clip.duration, target_duration))
|
387 |
-
except:
|
388 |
-
pass
|
389 |
-
logger.error("Fallback to original audio clip failed.")
|
390 |
-
return AudioFileClip(filename="")
|
391 |
-
|
392 |
-
finally:
|
393 |
-
if looped_audio is not None and looped_audio is not final_looped_audio:
|
394 |
-
try: looped_audio.close()
|
395 |
-
except: pass
|
396 |
-
|
397 |
|
398 |
def extract_visual_keywords_from_script(script_text):
|
399 |
logger.info("Extrayendo palabras clave del guion")
|
400 |
if not script_text or not script_text.strip():
|
401 |
-
logger.warning("Guion vacío
|
402 |
return ["naturaleza", "ciudad", "paisaje"]
|
403 |
-
|
404 |
clean_text = re.sub(r'[^\w\sáéíóúñÁÉÍÓÚÑ]', '', script_text)
|
405 |
-
keywords_list = []
|
406 |
-
|
407 |
if kw_model:
|
408 |
try:
|
409 |
-
logger.debug("Intentando extracción con KeyBERT...")
|
410 |
keywords1 = kw_model.extract_keywords(clean_text, keyphrase_ngram_range=(1, 1), stop_words='spanish', top_n=5)
|
411 |
keywords2 = kw_model.extract_keywords(clean_text, keyphrase_ngram_range=(2, 2), stop_words='spanish', top_n=3)
|
412 |
-
|
413 |
all_keywords = keywords1 + keywords2
|
414 |
all_keywords.sort(key=lambda item: item[1], reverse=True)
|
415 |
-
|
416 |
seen_keywords = set()
|
417 |
-
for keyword,
|
418 |
formatted_keyword = keyword.lower().replace(" ", "+")
|
419 |
if formatted_keyword and formatted_keyword not in seen_keywords:
|
420 |
keywords_list.append(formatted_keyword)
|
421 |
seen_keywords.add(formatted_keyword)
|
422 |
if len(keywords_list) >= 5:
|
423 |
break
|
424 |
-
|
425 |
if keywords_list:
|
426 |
logger.debug(f"Palabras clave extraídas por KeyBERT: {keywords_list}")
|
427 |
return keywords_list
|
428 |
-
|
429 |
except Exception as e:
|
430 |
-
logger.warning(f"KeyBERT falló: {str(e)}.
|
431 |
-
|
432 |
logger.debug("Extrayendo palabras clave con método simple...")
|
433 |
words = clean_text.lower().split()
|
434 |
-
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"
|
435 |
-
"nuestro", "vuestro", "este", "ese", "aquel", "esta", "esa", "aquella", "esto", "eso", "aquello", "mis", "tus",
|
436 |
-
"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"}
|
437 |
-
|
438 |
valid_words = [word for word in words if len(word) > 3 and word not in stop_words]
|
439 |
-
|
440 |
if not valid_words:
|
441 |
-
logger.warning("No se encontraron palabras clave válidas
|
442 |
-
return ["espiritual", "terror", "matrix", "arcontes", "galaxia"
|
443 |
-
|
444 |
word_counts = Counter(valid_words)
|
445 |
top_keywords = [word.replace(" ", "+") for word, _ in word_counts.most_common(5)]
|
446 |
-
|
447 |
-
if not top_keywords:
|
448 |
-
logger.warning("El método simple no produjo keywords. Usando palabras clave predeterminadas.")
|
449 |
-
return ["espiritual", "terror", "matrix", "arcontes", "galaxia", "creepy", "magia", "gangstalking","conspiracy",]
|
450 |
-
|
451 |
logger.info(f"Palabras clave finales: {top_keywords}")
|
452 |
return top_keywords
|
453 |
|
454 |
-
# crear_video ahora recibe la voz seleccionada
|
455 |
def crear_video(prompt_type, input_text, selected_voice, musica_file=None):
|
456 |
logger.info("="*80)
|
457 |
logger.info(f"INICIANDO CREACIÓN DE VIDEO | Tipo: {prompt_type}")
|
458 |
logger.debug(f"Input: '{input_text[:100]}...'")
|
459 |
logger.info(f"Voz seleccionada: {selected_voice}")
|
460 |
-
|
461 |
start_time = datetime.now()
|
462 |
-
temp_dir_intermediate =
|
463 |
-
|
|
|
464 |
audio_tts_original = None
|
465 |
musica_audio_original = None
|
466 |
audio_tts = None
|
@@ -472,403 +356,147 @@ def crear_video(prompt_type, input_text, selected_voice, musica_file=None):
|
|
472 |
|
473 |
try:
|
474 |
# 1. Generar o usar guion
|
475 |
-
if prompt_type == "Generar Guion con IA"
|
476 |
-
guion = generate_script(input_text)
|
477 |
-
else:
|
478 |
-
guion = input_text.strip()
|
479 |
-
|
480 |
logger.info(f"Guion final ({len(guion)} chars): '{guion[:100]}...'")
|
481 |
-
|
482 |
if not guion.strip():
|
483 |
-
logger.error("El guion resultante está vacío o solo contiene espacios.")
|
484 |
raise ValueError("El guion está vacío.")
|
485 |
|
486 |
-
|
487 |
-
logger.info(f"Directorio temporal intermedio creado: {temp_dir_intermediate}")
|
488 |
-
temp_intermediate_files = []
|
489 |
-
|
490 |
-
# 2. Generar audio de voz usando la voz seleccionada, con reintentos si falla
|
491 |
-
logger.info("Generando audio de voz...")
|
492 |
voz_path = os.path.join(temp_dir_intermediate, "voz.mp3")
|
493 |
-
|
494 |
-
tts_voices_to_try = [selected_voice]
|
495 |
-
fallback_juan = "es-ES-JuanNeural"
|
496 |
-
fallback_elvira = "es-ES-ElviraNeural"
|
497 |
-
|
498 |
-
if fallback_juan and fallback_juan != selected_voice and fallback_juan not in tts_voices_to_try:
|
499 |
-
tts_voices_to_try.append(fallback_juan)
|
500 |
-
if fallback_elvira and fallback_elvira != selected_voice and fallback_elvira not in tts_voices_to_try:
|
501 |
-
tts_voices_to_try.append(fallback_elvira)
|
502 |
-
|
503 |
tts_success = False
|
504 |
tried_voices = set()
|
505 |
|
506 |
for current_voice in tts_voices_to_try:
|
507 |
-
if
|
|
|
508 |
tried_voices.add(current_voice)
|
509 |
-
|
510 |
logger.info(f"Intentando TTS con voz: {current_voice}...")
|
511 |
-
|
512 |
-
|
513 |
-
|
514 |
-
logger.info(f"TTS exitoso con voz '{current_voice}'.")
|
515 |
-
break
|
516 |
-
except Exception as e:
|
517 |
-
logger.warning(f"Fallo al generar TTS con voz '{current_voice}': {str(e)}", exc_info=True)
|
518 |
-
pass
|
519 |
|
520 |
if not tts_success or not os.path.exists(voz_path) or os.path.getsize(voz_path) <= 100:
|
521 |
-
|
522 |
-
raise ValueError("Error generando voz a partir del guion (fallo de TTS).")
|
523 |
|
524 |
temp_intermediate_files.append(voz_path)
|
525 |
-
|
526 |
audio_tts_original = AudioFileClip(voz_path)
|
527 |
-
|
528 |
-
|
529 |
-
logger.critical("Clip de audio TTS inicial es inválido (reader is None o duración <= 0) *después* de crear AudioFileClip.")
|
530 |
-
try: audio_tts_original.close()
|
531 |
-
except: pass
|
532 |
-
audio_tts_original = None
|
533 |
-
if os.path.exists(voz_path):
|
534 |
-
try: os.remove(voz_path)
|
535 |
-
except: pass
|
536 |
-
if voz_path in temp_intermediate_files:
|
537 |
-
temp_intermediate_files.remove(voz_path)
|
538 |
-
|
539 |
-
raise ValueError("Audio de voz generado es inválido después de procesamiento inicial.")
|
540 |
-
|
541 |
audio_tts = audio_tts_original
|
542 |
audio_duration = audio_tts_original.duration
|
543 |
logger.info(f"Duración audio voz: {audio_duration:.2f} segundos")
|
544 |
-
|
545 |
if audio_duration < 1.0:
|
546 |
-
|
547 |
-
raise ValueError("Generated voice audio is too short (min 1 second required).")
|
548 |
-
# 3. Extraer palabras clave
|
549 |
-
logger.info("Extrayendo palabras clave...")
|
550 |
-
try:
|
551 |
-
keywords = extract_visual_keywords_from_script(guion)
|
552 |
-
logger.info(f"Palabras clave identificadas: {keywords}")
|
553 |
-
except Exception as e:
|
554 |
-
logger.error(f"Error extrayendo keywords: {str(e)}", exc_info=True)
|
555 |
-
keywords = ["naturaleza", "paisaje"]
|
556 |
|
|
|
|
|
557 |
if not keywords:
|
558 |
-
|
|
|
559 |
|
560 |
# 4. Buscar y descargar videos
|
561 |
-
logger.info("Buscando videos en Pexels...")
|
562 |
videos_data = []
|
563 |
total_desired_videos = 10
|
564 |
per_page_per_keyword = max(1, total_desired_videos // len(keywords))
|
565 |
-
|
566 |
for keyword in keywords:
|
567 |
-
if len(videos_data) >= total_desired_videos:
|
568 |
-
|
569 |
-
|
570 |
-
|
571 |
-
videos_data.extend(videos)
|
572 |
-
logger.info(f"Encontrados {len(videos)} videos para '{keyword}'. Total data: {len(videos_data)}")
|
573 |
-
except Exception as e:
|
574 |
-
logger.warning(f"Error buscando videos para '{keyword}': {str(e)}")
|
575 |
|
576 |
if len(videos_data) < total_desired_videos / 2:
|
577 |
-
|
578 |
-
generic_keywords = ["mystery", "alien", "ufo", "conspiracy", "paranormal", "supernatural", "horror", "fear", "suspense", "secret", "government", "cover_up", "simulation", "matrix", "apocalypse", "dystopian", "shadow", "occult", "unexplained", "creepy", "extraterrestrial", "abduction", "experiment", "secret_society", "illuminati", "new_world_order", "ancient_aliens", "ufo_sighting", "cryptid", "bigfoot", "loch_ness", "ghost", "haunting", "spirit", "demon", "possession", "exorcism", "witchcraft", "ritual", "cursed", "urban_legend", "myth", "legend", "folklore", "scary", "terror", "panic", "anxiety", "dread", "nightmare", "dark", "gloomy", "fog", "haunted", "cemetery", "asylum", "abandoned", "ruins", "underground", "tunnel", "bunker", "lab", "experiment", "government_secret", "mind_control", "brainwash", "propaganda", "surveillance", "spy", "whistleblower", "leak", "anonymous", "hack", "cyber", "virtual_reality", "ai", "artificial_intelligence", "robot", "cyborg", "apocalyptic", "post_apocalyptic", "zombie", "outbreak", "pandemic", "contagion", "biohazard", "radiation", "nuclear", "doomsday", "armageddon", "revelation", "prophecy", "symbolism", "hidden_meaning", "enigma", "puzzle", "code", "cipher", "mysterious", "unidentified", "anomaly", "glitch", "time_travel", "parallel_universe", "dimension", "portal"]
|
579 |
for keyword in generic_keywords:
|
580 |
-
|
581 |
-
|
582 |
-
|
583 |
-
|
584 |
-
videos_data.extend(videos)
|
585 |
-
logger.info(f"Encontrados {len(videos)} videos para '{keyword}' (genérico). Total data: {len(videos_data)}")
|
586 |
-
except Exception as e:
|
587 |
-
logger.warning(f"Error buscando videos genéricos para '{keyword}': {str(e)}")
|
588 |
|
589 |
if not videos_data:
|
590 |
-
|
591 |
-
raise ValueError("No se encontraron videos adecuados en Pexels.")
|
592 |
|
593 |
video_paths = []
|
594 |
-
logger.info(f"Intentando descargar {len(videos_data)} videos encontrados...")
|
595 |
for video in videos_data:
|
596 |
if 'video_files' not in video or not video['video_files']:
|
597 |
-
logger.debug(f"Saltando video sin archivos de video: {video.get('id')}")
|
598 |
continue
|
|
|
|
|
|
|
|
|
|
|
|
|
599 |
|
600 |
-
try:
|
601 |
-
best_quality = None
|
602 |
-
for vf in sorted(video['video_files'], key=lambda x: x.get('width', 0) * x.get('height', 0), reverse=True):
|
603 |
-
if 'link' in vf:
|
604 |
-
best_quality = vf
|
605 |
-
break
|
606 |
-
|
607 |
-
if best_quality and 'link' in best_quality:
|
608 |
-
path = download_video_file(best_quality['link'], temp_dir_intermediate)
|
609 |
-
if path:
|
610 |
-
video_paths.append(path)
|
611 |
-
temp_intermediate_files.append(path)
|
612 |
-
logger.info(f"Video descargado OK desde {best_quality['link'][:50]}...")
|
613 |
-
else:
|
614 |
-
logger.warning(f"No se pudo descargar video desde {best_quality['link'][:50]}...")
|
615 |
-
else:
|
616 |
-
logger.warning(f"No se encontró enlace de descarga válido para video {video.get('id')}.")
|
617 |
-
|
618 |
-
except Exception as e:
|
619 |
-
logger.warning(f"Error procesando/descargando video {video.get('id')}: {str(e)}")
|
620 |
-
|
621 |
-
logger.info(f"Descargados {len(video_paths)} archivos de video utilizables.")
|
622 |
if not video_paths:
|
623 |
-
|
624 |
-
raise ValueError("No se pudo descargar ningún video utilizable de Pexels.")
|
625 |
|
626 |
# 5. Procesar y concatenar clips de video
|
627 |
-
logger.info("Procesando y concatenando videos descargados...")
|
628 |
current_duration = 0
|
629 |
min_clip_duration = 0.5
|
630 |
max_clip_segment = 10.0
|
631 |
-
|
632 |
for i, path in enumerate(video_paths):
|
633 |
if current_duration >= audio_duration + max_clip_segment:
|
634 |
-
logger.debug(f"Video base suficiente ({current_duration:.1f}s >= {audio_duration:.1f}s + {max_clip_segment:.1f}s buffer). Dejando de procesar clips fuente restantes.")
|
635 |
break
|
636 |
-
|
637 |
-
clip = None
|
638 |
try:
|
639 |
-
logger.debug(f"[{i+1}/{len(video_paths)}] Abriendo clip: {path}")
|
640 |
clip = VideoFileClip(path)
|
641 |
source_clips.append(clip)
|
642 |
-
|
643 |
-
if clip.reader is None or clip.duration is None or clip.duration <= 0:
|
644 |
-
logger.warning(f"[{i+1}/{len(video_paths)}] Clip fuente {path} parece inválido (reader is None o duración <= 0). Saltando.")
|
645 |
continue
|
646 |
-
|
647 |
remaining_needed = audio_duration - current_duration
|
648 |
-
|
649 |
-
|
650 |
-
|
651 |
-
|
652 |
-
|
653 |
-
segment_duration = min(segment_duration, clip.duration)
|
654 |
-
|
655 |
-
if segment_duration >= min_clip_duration:
|
656 |
-
try:
|
657 |
-
sub = clip.subclip(0, segment_duration)
|
658 |
-
if sub.reader is None or sub.duration is None or sub.duration <= 0:
|
659 |
-
logger.warning(f"[{i+1}/{len(video_paths)}] Subclip generado de {path} es inválido. Saltando.")
|
660 |
-
try: sub.close()
|
661 |
-
except: pass
|
662 |
-
continue
|
663 |
-
|
664 |
-
clips_to_concatenate.append(sub)
|
665 |
-
current_duration += sub.duration
|
666 |
-
logger.debug(f"[{i+1}/{len(video_paths)}] Segmento añadido: {sub.duration:.1f}s (total video: {current_duration:.1f}/{audio_duration:.1f}s)")
|
667 |
-
|
668 |
-
except Exception as sub_e:
|
669 |
-
logger.warning(f"[{i+1}/{len(video_paths)}] Error creando subclip de {path} ({segment_duration:.1f}s): {str(sub_e)}")
|
670 |
-
continue
|
671 |
-
else:
|
672 |
-
logger.debug(f"[{i+1}/{len(video_paths)}] Clip {path} ({clip.duration:.1f}s) no contribuye un segmento suficiente ({segment_duration:.1f}s necesario). Saltando.")
|
673 |
-
else:
|
674 |
-
logger.debug(f"[{i+1}/{len(video_paths)}] Duración de video base ya alcanzada. Saltando clip.")
|
675 |
-
|
676 |
except Exception as e:
|
677 |
-
logger.warning(f"
|
678 |
-
continue
|
679 |
-
|
680 |
-
logger.info(f"Procesamiento de clips fuente finalizado. Se obtuvieron {len(clips_to_concatenate)} segmentos válidos.")
|
681 |
|
682 |
if not clips_to_concatenate:
|
683 |
-
|
684 |
-
raise ValueError("No hay segmentos de video válidos disponibles para crear el video.")
|
685 |
|
686 |
-
|
687 |
-
|
688 |
-
|
689 |
-
concatenated_base = concatenate_videoclips(clips_to_concatenate, method="chain")
|
690 |
-
logger.info(f"Duración video base después de concatenación inicial: {concatenated_base.duration:.2f}s")
|
691 |
-
|
692 |
-
if concatenated_base is None or concatenated_base.duration is None or concatenated_base.duration <= 0:
|
693 |
-
logger.critical("Video base concatenado es inválido después de la primera concatenación (None o duración cero).")
|
694 |
-
raise ValueError("Fallo al crear video base válido a partir de segmentos.")
|
695 |
-
|
696 |
-
except Exception as e:
|
697 |
-
logger.critical(f"Error durante la concatenación inicial: {str(e)}", exc_info=True)
|
698 |
-
raise ValueError("Fallo durante la concatenación de video inicial.")
|
699 |
-
finally:
|
700 |
-
for clip_segment in clips_to_concatenate:
|
701 |
-
try: clip_segment.close()
|
702 |
-
except: pass
|
703 |
-
clips_to_concatenate = []
|
704 |
-
|
705 |
-
video_base = concatenated_base
|
706 |
-
|
707 |
-
final_video_base = video_base
|
708 |
|
709 |
-
|
710 |
-
|
711 |
-
|
712 |
-
|
713 |
-
|
714 |
-
|
715 |
-
repeated_clips_list = [final_video_base] * num_full_repeats
|
716 |
if remaining_duration > 0:
|
717 |
-
|
718 |
-
|
719 |
-
|
720 |
-
|
721 |
-
|
722 |
-
except: pass
|
723 |
-
else:
|
724 |
-
repeated_clips_list.append(remaining_clip)
|
725 |
-
logger.debug(f"Añadiendo segmento restante: {remaining_duration:.2f}s")
|
726 |
-
|
727 |
-
except Exception as e:
|
728 |
-
logger.warning(f"Error creando subclip para duración restante {remaining_duration:.2f}s: {str(e)}")
|
729 |
-
|
730 |
-
if repeated_clips_list:
|
731 |
-
logger.info(f"Concatenando {len(repeated_clips_list)} partes para repetición.")
|
732 |
-
video_base_repeated = None
|
733 |
-
try:
|
734 |
-
video_base_repeated = concatenate_videoclips(repeated_clips_list, method="chain")
|
735 |
-
logger.info(f"Duración del video base repetido: {video_base_repeated.duration:.2f}s")
|
736 |
-
|
737 |
-
if video_base_repeated is None or video_base_repeated.duration is None or video_base_repeated.duration <= 0:
|
738 |
-
logger.critical("Video base repetido concatenado es inválido.")
|
739 |
-
raise ValueError("Fallo al crear video base repetido válido.")
|
740 |
-
|
741 |
-
if final_video_base is not video_base_repeated:
|
742 |
-
try: final_video_base.close()
|
743 |
-
except: pass
|
744 |
-
|
745 |
-
final_video_base = video_base_repeated
|
746 |
-
|
747 |
-
except Exception as e:
|
748 |
-
logger.critical(f"Error durante la concatenación de repetición: {str(e)}", exc_info=True)
|
749 |
-
raise ValueError("Fallo durante la repetición de video.")
|
750 |
-
finally:
|
751 |
-
if 'repeated_clips_list' in locals():
|
752 |
-
for clip in repeated_clips_list:
|
753 |
-
if clip is not final_video_base:
|
754 |
-
try: clip.close()
|
755 |
-
except: pass
|
756 |
-
|
757 |
-
|
758 |
-
if final_video_base.duration > audio_duration:
|
759 |
-
logger.info(f"Recortando video base ({final_video_base.duration:.2f}s) para que coincida con la duración del audio ({audio_duration:.2f}s).")
|
760 |
-
trimmed_video_base = None
|
761 |
-
try:
|
762 |
-
trimmed_video_base = final_video_base.subclip(0, audio_duration)
|
763 |
-
if trimmed_video_base is None or trimmed_video_base.duration is None or trimmed_video_base.duration <= 0:
|
764 |
-
logger.critical("Video base recortado es inválido.")
|
765 |
-
raise ValueError("Fallo al crear video base recortado válido.")
|
766 |
-
|
767 |
-
if final_video_base is not trimmed_video_base:
|
768 |
-
try: final_video_base.close()
|
769 |
-
except: pass
|
770 |
-
|
771 |
-
final_video_base = trimmed_video_base
|
772 |
-
|
773 |
-
except Exception as e:
|
774 |
-
logger.critical(f"Error durante el recorte: {str(e)}", exc_info=True)
|
775 |
-
raise ValueError("Fallo durante el recorte de video.")
|
776 |
-
|
777 |
-
|
778 |
-
if final_video_base is None or final_video_base.duration is None or final_video_base.duration <= 0:
|
779 |
-
logger.critical("Video base final es inválido antes de audio/escritura (None o duración cero).")
|
780 |
-
raise ValueError("Video base final es inválido.")
|
781 |
-
|
782 |
-
if final_video_base.size is None or final_video_base.size[0] <= 0 or final_video_base.size[1] <= 0:
|
783 |
-
logger.critical(f"Video base final tiene tamaño inválido: {final_video_base.size}. No se puede escribir video.")
|
784 |
-
raise ValueError("Video base final tiene tamaño inválido antes de escribir.")
|
785 |
-
|
786 |
-
video_base = final_video_base
|
787 |
|
788 |
# 6. Manejar música de fondo
|
789 |
-
|
790 |
-
|
791 |
-
final_audio = audio_tts_original
|
792 |
-
|
793 |
-
musica_audio_looped = None
|
794 |
-
|
795 |
if musica_file:
|
796 |
-
musica_audio_original = None
|
797 |
try:
|
798 |
music_path = os.path.join(temp_dir_intermediate, "musica_bg.mp3")
|
799 |
shutil.copyfile(musica_file, music_path)
|
800 |
temp_intermediate_files.append(music_path)
|
801 |
-
logger.info(f"Música de fondo copiada a: {music_path}")
|
802 |
-
|
803 |
musica_audio_original = AudioFileClip(music_path)
|
804 |
-
|
805 |
-
|
806 |
-
|
807 |
-
|
808 |
-
|
809 |
-
musica_audio_original = None
|
810 |
-
else:
|
811 |
-
musica_audio_looped = loop_audio_to_length(musica_audio_original, video_base.duration)
|
812 |
-
logger.debug(f"Música ajustada a duración del video: {musica_audio_looped.duration:.2f}s")
|
813 |
-
|
814 |
-
if musica_audio_looped is None or musica_audio_looped.duration is None or musica_audio_looped.duration <= 0:
|
815 |
-
logger.warning("Clip de música de fondo loopeado es inválido. Saltando música.")
|
816 |
-
try: musica_audio_looped.close()
|
817 |
-
except: pass
|
818 |
-
musica_audio_looped = None
|
819 |
-
|
820 |
-
|
821 |
-
if musica_audio_looped:
|
822 |
-
composite_audio = CompositeAudioClip([
|
823 |
-
musica_audio_looped.volumex(0.2), # Volumen 20% para música
|
824 |
-
audio_tts_original.volumex(1.0) # Volumen 100% para voz
|
825 |
])
|
826 |
-
|
827 |
-
if composite_audio.duration is None or composite_audio.duration <= 0:
|
828 |
-
logger.warning("Clip de audio compuesto es inválido (None o duración cero). Usando solo audio de voz.")
|
829 |
-
try: composite_audio.close()
|
830 |
-
except: pass
|
831 |
-
final_audio = audio_tts_original
|
832 |
-
else:
|
833 |
-
logger.info("Mezcla de audio completada (voz + música).")
|
834 |
-
final_audio = composite_audio
|
835 |
-
musica_audio = musica_audio_looped # Asignar para limpieza
|
836 |
-
|
837 |
-
except Exception as e:
|
838 |
-
logger.warning(f"Error procesando música de fondo: {str(e)}", exc_info=True)
|
839 |
-
final_audio = audio_tts_original
|
840 |
-
musica_audio = None
|
841 |
-
logger.warning("Usando solo audio de voz debido a un error con la música.")
|
842 |
-
|
843 |
-
|
844 |
-
if final_audio.duration is not None and abs(final_audio.duration - video_base.duration) > 0.2:
|
845 |
-
logger.warning(f"Duración del audio final ({final_audio.duration:.2f}s) difiere significativamente del video base ({video_base.duration:.2f}s). Intentando recorte.")
|
846 |
-
try:
|
847 |
-
if final_audio.duration > video_base.duration:
|
848 |
-
trimmed_final_audio = final_audio.subclip(0, video_base.duration)
|
849 |
-
if trimmed_final_audio is None or trimmed_final_audio.duration <= 0:
|
850 |
-
logger.warning("Audio final recortado es inválido. Usando audio final original.")
|
851 |
-
try: trimmed_final_audio.close()
|
852 |
-
except: pass
|
853 |
-
else:
|
854 |
-
if final_audio is not trimmed_final_audio:
|
855 |
-
try: final_audio.close()
|
856 |
-
except: pass
|
857 |
-
final_audio = trimmed_final_audio
|
858 |
-
logger.warning("Audio final recortado para que coincida con la duración del video.")
|
859 |
except Exception as e:
|
860 |
-
|
|
|
861 |
|
862 |
-
|
863 |
-
|
864 |
|
865 |
-
# 7.
|
866 |
-
|
867 |
-
output_filename = f"video_{int(
|
868 |
output_path = os.path.join(temp_dir_intermediate, output_filename)
|
869 |
permanent_path = f"/tmp/{output_filename}"
|
870 |
-
|
871 |
-
# Escribir el video
|
872 |
video_final.write_videofile(
|
873 |
output_path,
|
874 |
fps=24,
|
@@ -876,172 +504,95 @@ def crear_video(prompt_type, input_text, selected_voice, musica_file=None):
|
|
876 |
codec="libx264",
|
877 |
audio_codec="aac",
|
878 |
preset="medium",
|
879 |
-
ffmpeg_params=[
|
880 |
-
'-vf', 'scale=1920:1080:force_original_aspect_ratio=decrease,pad=1920:1080:-1:-1:color=black',
|
881 |
-
'-crf', '23'
|
882 |
-
],
|
883 |
logger='bar'
|
884 |
)
|
885 |
|
886 |
-
|
887 |
-
|
888 |
-
shutil.copy(output_path, permanent_path) # Usamos copy() en lugar de move()
|
889 |
-
logger.info(f"Video guardado permanentemente en: {permanent_path}")
|
890 |
-
except Exception as move_error:
|
891 |
-
logger.error(f"Error moviendo archivo: {str(move_error)}. Usando path original.")
|
892 |
-
permanent_path = output_path
|
893 |
-
|
894 |
-
# Cierra los clips para liberar memoria
|
895 |
-
try:
|
896 |
-
video_final.close()
|
897 |
-
if 'video_base' in locals() and video_base is not None and video_base is not video_final:
|
898 |
-
video_base.close()
|
899 |
-
except Exception as close_error:
|
900 |
-
logger.error(f"Error cerrando clips: {str(close_error)}")
|
901 |
-
|
902 |
total_time = (datetime.now() - start_time).total_seconds()
|
903 |
-
logger.info(f"
|
904 |
-
|
905 |
return permanent_path
|
906 |
|
907 |
except ValueError as ve:
|
908 |
-
|
909 |
-
|
910 |
except Exception as e:
|
911 |
-
logger.critical(f"
|
912 |
-
raise
|
913 |
finally:
|
914 |
-
|
915 |
-
|
916 |
-
for clip in source_clips:
|
917 |
try:
|
918 |
clip.close()
|
919 |
-
except
|
920 |
-
|
921 |
-
|
922 |
-
|
923 |
-
|
924 |
-
|
925 |
-
|
926 |
-
|
927 |
-
|
928 |
-
if musica_audio is not None:
|
929 |
try:
|
930 |
musica_audio.close()
|
931 |
-
except
|
932 |
-
|
933 |
-
|
934 |
-
|
935 |
-
|
936 |
-
|
937 |
-
|
938 |
-
|
939 |
-
|
940 |
-
|
941 |
-
|
942 |
-
|
943 |
-
|
944 |
-
logger.warning(f"Error cerrando audio_tts (procesada) en finally: {str(e)}")
|
945 |
-
|
946 |
-
if audio_tts_original is not None:
|
947 |
-
try:
|
948 |
-
audio_tts_original.close()
|
949 |
-
except Exception as e:
|
950 |
-
logger.warning(f"Error cerrando audio_tts_original en finally: {str(e)}")
|
951 |
-
|
952 |
-
if video_final is not None:
|
953 |
try:
|
954 |
video_final.close()
|
955 |
-
except
|
956 |
-
|
957 |
-
|
958 |
-
|
959 |
-
|
960 |
-
|
961 |
-
|
962 |
-
|
963 |
-
if temp_dir_intermediate and os.path.exists(temp_dir_intermediate):
|
964 |
-
final_output_in_temp = os.path.join(temp_dir_intermediate, output_filename)
|
965 |
-
|
966 |
-
for path in temp_intermediate_files:
|
967 |
-
try:
|
968 |
-
if os.path.isfile(path) and path != final_output_in_temp and path != permanent_path:
|
969 |
-
logger.debug(f"Eliminando archivo temporal intermedio: {path}")
|
970 |
-
os.remove(path)
|
971 |
-
elif os.path.isfile(path) and (path == final_output_in_temp or path == permanent_path):
|
972 |
-
logger.debug(f"Saltando eliminación del archivo de video final: {path}")
|
973 |
-
except Exception as e:
|
974 |
-
logger.warning(f"No se pudo eliminar archivo temporal intermedio {path}: {str(e)}")
|
975 |
-
|
976 |
-
logger.info(f"Directorio temporal intermedio {temp_dir_intermediate} persistirá para que Gradio lea el video final.")
|
977 |
|
978 |
-
|
979 |
-
def run_app(prompt_type, prompt_ia, prompt_manual, musica_file, selected_voice): # <-- Recibe el valor del Dropdown
|
980 |
logger.info("="*80)
|
981 |
logger.info("SOLICITUD RECIBIDA EN INTERFAZ")
|
982 |
-
|
983 |
-
# Elegir el texto de entrada basado en el prompt_type
|
984 |
input_text = prompt_ia if prompt_type == "Generar Guion con IA" else prompt_manual
|
985 |
-
|
986 |
output_video = None
|
987 |
output_file = None
|
988 |
-
status_msg = gr.update(value="⏳ Procesando..."
|
989 |
|
990 |
if not input_text or not input_text.strip():
|
991 |
logger.warning("Texto de entrada vacío.")
|
992 |
-
|
993 |
-
return None, None, gr.update(value="⚠️ Por favor, ingresa texto para el guion o el tema.", interactive=False)
|
994 |
|
995 |
-
# Validar la voz seleccionada. Si no es válida, usar la por defecto.
|
996 |
-
# AVAILABLE_VOICES se obtiene al inicio. Hay que buscar si el voice_id existe en la lista de pares (nombre, id)
|
997 |
voice_ids_disponibles = [v[1] for v in AVAILABLE_VOICES]
|
998 |
if selected_voice not in voice_ids_disponibles:
|
999 |
-
logger.warning(f"Voz
|
1000 |
-
selected_voice = DEFAULT_VOICE_ID
|
1001 |
-
else:
|
1002 |
-
logger.info(f"Voz seleccionada validada: {selected_voice}")
|
1003 |
-
|
1004 |
-
|
1005 |
-
logger.info(f"Tipo de entrada: {prompt_type}")
|
1006 |
-
logger.debug(f"Texto de entrada: '{input_text[:100]}...'")
|
1007 |
-
if musica_file:
|
1008 |
-
logger.info(f"Archivo de música recibido: {musica_file}")
|
1009 |
-
else:
|
1010 |
-
logger.info("No se proporcionó archivo de música.")
|
1011 |
-
logger.info(f"Voz final a usar (ID): {selected_voice}") # Loguear el ID de la voz final
|
1012 |
|
1013 |
try:
|
1014 |
-
|
1015 |
-
# Pasar el input_text elegido, la voz seleccionada (el ID) y el archivo de música a crear_video
|
1016 |
-
video_path = crear_video(prompt_type, input_text, selected_voice, musica_file) # <-- PASAR selected_voice (ID) a crear_video
|
1017 |
-
|
1018 |
if video_path and os.path.exists(video_path):
|
1019 |
-
|
1020 |
-
|
1021 |
-
|
1022 |
-
output_file = video_path # Establecer valor del componente de archivo para descarga
|
1023 |
-
status_msg = gr.update(value="✅ Video generado exitosamente.", interactive=False)
|
1024 |
else:
|
1025 |
-
|
1026 |
-
status_msg = gr.update(value="❌ Error: La generación del video falló o el archivo no se creó correctamente.", interactive=False)
|
1027 |
-
|
1028 |
except ValueError as ve:
|
1029 |
-
logger.warning(f"Error de validación
|
1030 |
-
status_msg = gr.update(value=f"⚠️ Error
|
1031 |
except Exception as e:
|
1032 |
-
logger.critical(f"Error crítico
|
1033 |
-
status_msg = gr.update(value=f"❌ Error inesperado: {str(e)}"
|
1034 |
-
|
1035 |
-
logger.info("Fin del handler run_app.")
|
1036 |
-
return output_video, output_file, status_msg
|
1037 |
-
|
1038 |
|
1039 |
# Interfaz de Gradio
|
1040 |
-
with gr.Blocks(title="Generador de Videos con IA", theme=gr.themes.Soft()
|
1041 |
-
.gradio-container {max-width: 800px; margin: auto;}
|
1042 |
-
h1 {text-align: center;}
|
1043 |
-
""") as app:
|
1044 |
-
|
1045 |
gr.Markdown("# 🎬 Generador Automático de Videos con IA")
|
1046 |
gr.Markdown("Genera videos cortos a partir de un tema o guion, usando imágenes de archivo de Pexels y voz generada.")
|
1047 |
|
@@ -1052,44 +603,32 @@ with gr.Blocks(title="Generador de Videos con IA", theme=gr.themes.Soft(), css="
|
|
1052 |
label="Método de Entrada",
|
1053 |
value="Generar Guion con IA"
|
1054 |
)
|
1055 |
-
|
1056 |
-
# Contenedores para los campos de texto para controlar la visibilidad
|
1057 |
with gr.Column(visible=True) as ia_guion_column:
|
1058 |
prompt_ia = gr.Textbox(
|
1059 |
label="Tema para IA",
|
1060 |
lines=2,
|
1061 |
-
placeholder="Ej: Un paisaje natural con montañas y ríos al amanecer
|
1062 |
-
max_lines=4
|
1063 |
-
value=""
|
1064 |
)
|
1065 |
-
|
1066 |
with gr.Column(visible=False) as manual_guion_column:
|
1067 |
prompt_manual = gr.Textbox(
|
1068 |
label="Tu Guion Completo",
|
1069 |
lines=5,
|
1070 |
-
placeholder="Ej: En este video exploraremos los misterios del océano
|
1071 |
-
max_lines=10
|
1072 |
-
value=""
|
1073 |
)
|
1074 |
-
|
1075 |
musica_input = gr.Audio(
|
1076 |
label="Música de fondo (opcional)",
|
1077 |
type="filepath",
|
1078 |
-
interactive=True
|
1079 |
-
value=None
|
1080 |
)
|
1081 |
-
|
1082 |
-
# --- COMPONENTE: Selección de Voz ---
|
1083 |
voice_dropdown = gr.Dropdown(
|
1084 |
label="Seleccionar Voz para Guion",
|
1085 |
choices=AVAILABLE_VOICES,
|
1086 |
value=DEFAULT_VOICE_ID,
|
1087 |
interactive=True
|
1088 |
)
|
1089 |
-
# --- FIN COMPONENTE ---
|
1090 |
-
|
1091 |
generate_btn = gr.Button("✨ Generar Video", variant="primary")
|
1092 |
-
|
1093 |
with gr.Column():
|
1094 |
video_output = gr.Video(
|
1095 |
label="Previsualización del Video Generado",
|
@@ -1104,24 +643,19 @@ with gr.Blocks(title="Generador de Videos con IA", theme=gr.themes.Soft(), css="
|
|
1104 |
status_output = gr.Textbox(
|
1105 |
label="Estado",
|
1106 |
interactive=False,
|
1107 |
-
show_label=False,
|
1108 |
placeholder="Esperando acción...",
|
1109 |
value="Esperando entrada..."
|
1110 |
)
|
1111 |
|
1112 |
-
# Evento para mostrar/ocultar los campos de texto según el tipo de prompt
|
1113 |
prompt_type.change(
|
1114 |
-
lambda x: (gr.update(visible=x == "Generar Guion con IA"),
|
1115 |
-
gr.update(visible=x == "Usar Mi Guion")),
|
1116 |
inputs=prompt_type,
|
1117 |
outputs=[ia_guion_column, manual_guion_column]
|
1118 |
)
|
1119 |
|
1120 |
-
# Evento click del botón de generar video
|
1121 |
generate_btn.click(
|
1122 |
-
lambda: (None, None, gr.update(value="⏳ Procesando... Esto puede tomar varios minutos."
|
1123 |
-
outputs=[video_output, file_output, status_output]
|
1124 |
-
queue=True,
|
1125 |
).then(
|
1126 |
run_app,
|
1127 |
inputs=[prompt_type, prompt_ia, prompt_manual, musica_input, voice_dropdown],
|
@@ -1134,37 +668,29 @@ with gr.Blocks(title="Generador de Videos con IA", theme=gr.themes.Soft(), css="
|
|
1134 |
|
1135 |
gr.Markdown("### Instrucciones:")
|
1136 |
gr.Markdown("""
|
1137 |
-
1.
|
1138 |
-
2.
|
1139 |
-
3.
|
1140 |
-
4.
|
1141 |
-
5.
|
1142 |
-
6.
|
1143 |
-
7.
|
1144 |
-
8. Revisa `video_generator_full.log` para detalles si hay errores.
|
1145 |
""")
|
1146 |
-
gr.Markdown("---")
|
1147 |
-
gr.Markdown("Desarrollado por [Tu Nombre/Empresa/Alias - Opcional]")
|
1148 |
|
1149 |
if __name__ == "__main__":
|
1150 |
-
logger.info("Verificando dependencias
|
1151 |
try:
|
1152 |
from moviepy.editor import ColorClip
|
1153 |
-
|
1154 |
-
|
1155 |
-
|
1156 |
-
logger.info("Clips base de MoviePy creados y cerrados exitosamente. FFmpeg parece accesible.")
|
1157 |
-
except Exception as e:
|
1158 |
-
logger.critical(f"Fallo al crear clip base de MoviePy. A menudo indica problemas con FFmpeg/ImageMagick. Error: {e}", exc_info=True)
|
1159 |
except Exception as e:
|
1160 |
-
|
1161 |
-
|
1162 |
-
|
1163 |
-
os.environ['GRADIO_SERVER_TIMEOUT'] = '6000' # 600 segundos = 10 minutos
|
1164 |
-
|
1165 |
logger.info("Iniciando aplicación Gradio...")
|
1166 |
try:
|
1167 |
app.launch(server_name="0.0.0.0", server_port=7860, share=False)
|
1168 |
except Exception as e:
|
1169 |
-
logger.critical(f"No se pudo iniciar la app: {str(e)}"
|
1170 |
raise
|
|
|
9 |
import torch
|
10 |
from transformers import GPT2Tokenizer, GPT2LMHeadModel
|
11 |
from keybert import KeyBERT
|
|
|
12 |
from moviepy.editor import VideoFileClip, concatenate_videoclips, AudioFileClip, CompositeAudioClip, concatenate_audioclips, AudioClip
|
13 |
import re
|
14 |
import math
|
|
|
31 |
logger.info("="*80)
|
32 |
|
33 |
# Diccionario de voces TTS disponibles organizadas por idioma
|
|
|
34 |
VOCES_DISPONIBLES = {
|
35 |
"Español (España)": {
|
36 |
"es-ES-JuanNeural": "Juan (España) - Masculino",
|
|
|
98 |
choices = []
|
99 |
for region, voices in VOCES_DISPONIBLES.items():
|
100 |
for voice_id, voice_name in voices.items():
|
|
|
101 |
choices.append((f"{voice_name} ({region})", voice_id))
|
102 |
return choices
|
103 |
|
104 |
# Obtener las voces al inicio del script
|
105 |
+
AVAILABLE_VOICES = get_voice_choices()
|
106 |
+
DEFAULT_VOICE_ID = "es-ES-JuanNeural"
|
|
|
|
|
|
|
|
|
|
|
|
|
107 |
DEFAULT_VOICE_NAME = DEFAULT_VOICE_ID
|
108 |
for text, voice_id in AVAILABLE_VOICES:
|
109 |
if voice_id == DEFAULT_VOICE_ID:
|
110 |
DEFAULT_VOICE_NAME = text
|
111 |
break
|
|
|
112 |
if DEFAULT_VOICE_ID not in [v[1] for v in AVAILABLE_VOICES]:
|
113 |
DEFAULT_VOICE_ID = AVAILABLE_VOICES[0][1] if AVAILABLE_VOICES else "en-US-AriaNeural"
|
114 |
+
DEFAULT_VOICE_NAME = AVAILABLE_VOICES[0][0] if AVAILABLE_VOICES else "Aria (United States) - Female"
|
|
|
115 |
logger.info(f"Voz por defecto seleccionada (ID): {DEFAULT_VOICE_ID}")
|
116 |
|
|
|
117 |
# Clave API de Pexels
|
118 |
PEXELS_API_KEY = os.environ.get("PEXELS_API_KEY")
|
119 |
if not PEXELS_API_KEY:
|
120 |
logger.critical("NO SE ENCONTRÓ PEXELS_API_KEY EN VARIABLES DE ENTORNO")
|
|
|
121 |
|
122 |
# Inicialización de modelos
|
123 |
MODEL_NAME = "datificate/gpt2-small-spanish"
|
|
|
157 |
"orientation": "landscape",
|
158 |
"size": "medium"
|
159 |
}
|
|
|
160 |
response = requests.get(
|
161 |
"https://api.pexels.com/videos/search",
|
162 |
headers=headers,
|
|
|
164 |
timeout=20
|
165 |
)
|
166 |
response.raise_for_status()
|
|
|
167 |
data = response.json()
|
168 |
videos = data.get('videos', [])
|
169 |
logger.info(f"Pexels: {len(videos)} videos encontrados para '{query}'")
|
170 |
return videos
|
|
|
171 |
except requests.exceptions.RequestException as e:
|
172 |
logger.error(f"Error de conexión Pexels para '{query}': {str(e)}")
|
173 |
+
return []
|
174 |
except json.JSONDecodeError:
|
175 |
+
logger.error(f"Pexels: JSON inválido recibido | Status: {response.status_code}")
|
176 |
+
return []
|
177 |
except Exception as e:
|
178 |
+
logger.error(f"Error inesperado Pexels para '{query}': {str(e)}")
|
179 |
+
return []
|
|
|
180 |
|
181 |
def generate_script(prompt, max_length=150):
|
182 |
logger.info(f"Generando guión | Prompt: '{prompt[:50]}...' | Longitud máxima: {max_length}")
|
|
|
192 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
193 |
model.to(device)
|
194 |
inputs = {k: v.to(device) for k, v in inputs.items()}
|
|
|
195 |
outputs = model.generate(
|
196 |
**inputs,
|
197 |
max_length=max_length + inputs[list(inputs.keys())[0]].size(1),
|
|
|
204 |
eos_token_id=tokenizer.eos_token_id,
|
205 |
no_repeat_ngram_size=3
|
206 |
)
|
|
|
207 |
text = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
208 |
+
prompt_in_output_idx = text.lower().find(prompt.lower())
|
209 |
+
if prompt_in_output_idx != -1:
|
210 |
+
cleaned_text = text[prompt_in_output_idx + len(prompt):].strip()
|
211 |
+
logger.debug("Texto limpiado tomando parte después del prompt original.")
|
212 |
+
else:
|
213 |
+
instruction_start_idx = text.find(instruction_phrase_start)
|
214 |
+
if instruction_start_idx != -1:
|
215 |
+
cleaned_text = text[instruction_start_idx + len(instruction_phrase_start):].strip()
|
216 |
+
logger.debug("Texto limpiado tomando parte después de la frase de instrucción base.")
|
|
|
217 |
else:
|
218 |
+
logger.warning("No se pudo identificar el inicio del guión generado.")
|
219 |
+
cleaned_text = text.strip()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
220 |
cleaned_text = re.sub(r'<[^>]+>', '', cleaned_text).strip()
|
221 |
+
cleaned_text = cleaned_text.lstrip(':').lstrip('.').strip()
|
|
|
|
|
|
|
|
|
222 |
sentences = cleaned_text.split('.')
|
223 |
if sentences and sentences[0].strip():
|
224 |
final_text = sentences[0].strip() + '.'
|
225 |
+
if len(sentences) > 1 and sentences[1].strip() and len(final_text.split()) < max_length * 0.7:
|
226 |
+
final_text += " " + sentences[1].strip() + "."
|
227 |
+
final_text = final_text.replace("..", ".")
|
|
|
|
|
228 |
logger.info(f"Guion generado final (Truncado a 100 chars): '{final_text[:100]}...'")
|
229 |
return final_text.strip()
|
230 |
+
logger.info(f"Guion generado final (sin oraciones completas detectadas): '{cleaned_text[:100]}...'")
|
231 |
+
return cleaned_text.strip()
|
|
|
|
|
232 |
except Exception as e:
|
233 |
+
logger.error(f"Error generando guion con GPT-2: {str(e)}")
|
|
|
234 |
return prompt.strip()
|
235 |
|
|
|
236 |
async def text_to_speech(text, output_path, voice):
|
237 |
+
logger.info(f"Convirtiendo texto a voz | Caracteres: {len(text)} | Voz: {voice}")
|
238 |
if not text or not text.strip():
|
239 |
logger.warning("Texto vacío para TTS")
|
240 |
return False
|
|
|
241 |
try:
|
242 |
communicate = edge_tts.Communicate(text, voice)
|
243 |
await communicate.save(output_path)
|
|
|
244 |
if os.path.exists(output_path) and os.path.getsize(output_path) > 100:
|
245 |
+
logger.info(f"Audio guardado exitosamente en: {output_path}")
|
246 |
return True
|
247 |
+
logger.error(f"TTS guardó un archivo pequeño o vacío en: {output_path}")
|
248 |
+
return False
|
|
|
|
|
249 |
except Exception as e:
|
250 |
+
logger.error(f"Error en TTS con voz '{voice}': {str(e)}")
|
251 |
return False
|
252 |
|
253 |
def download_video_file(url, temp_dir):
|
254 |
if not url:
|
255 |
+
logger.warning("URL de video no proporcionada")
|
256 |
return None
|
|
|
257 |
try:
|
258 |
logger.info(f"Descargando video desde: {url[:80]}...")
|
259 |
os.makedirs(temp_dir, exist_ok=True)
|
260 |
file_name = f"video_dl_{datetime.now().strftime('%Y%m%d_%H%M%S_%f')}.mp4"
|
261 |
output_path = os.path.join(temp_dir, file_name)
|
|
|
262 |
with requests.get(url, stream=True, timeout=60) as r:
|
263 |
r.raise_for_status()
|
264 |
with open(output_path, 'wb') as f:
|
265 |
for chunk in r.iter_content(chunk_size=8192):
|
266 |
f.write(chunk)
|
|
|
267 |
if os.path.exists(output_path) and os.path.getsize(output_path) > 1000:
|
268 |
+
logger.info(f"Video descargado exitosamente: {output_path}")
|
269 |
+
return output_path
|
270 |
+
logger.warning(f"Descarga parece incompleta o vacía: {output_path}")
|
271 |
+
if os.path.exists(output_path):
|
272 |
+
os.remove(output_path)
|
273 |
+
return None
|
|
|
|
|
274 |
except requests.exceptions.RequestException as e:
|
275 |
+
logger.error(f"Error de descarga para {url[:80]}...: {str(e)}")
|
276 |
+
return None
|
277 |
except Exception as e:
|
278 |
+
logger.error(f"Error inesperado descargando {url[:80]}...: {str(e)}")
|
279 |
+
return None
|
|
|
280 |
|
281 |
def loop_audio_to_length(audio_clip, target_duration):
|
282 |
logger.debug(f"Ajustando audio | Duración actual: {audio_clip.duration:.2f}s | Objetivo: {target_duration:.2f}s")
|
|
|
283 |
if audio_clip is None or audio_clip.duration is None or audio_clip.duration <= 0:
|
284 |
+
logger.warning("Input audio clip is invalid")
|
285 |
+
sr = getattr(audio_clip, 'fps', 44100) if audio_clip else 44100
|
286 |
+
return AudioClip(lambda t: 0, duration=target_duration, fps=sr)
|
|
|
|
|
|
|
|
|
|
|
287 |
if audio_clip.duration >= target_duration:
|
288 |
+
logger.debug("Audio clip ya es suficientemente largo. Recortando.")
|
289 |
+
return audio_clip.subclip(0, target_duration)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
290 |
loops = math.ceil(target_duration / audio_clip.duration)
|
291 |
logger.debug(f"Creando {loops} loops de audio")
|
|
|
|
|
|
|
|
|
292 |
try:
|
293 |
+
looped_audio = concatenate_audioclips([audio_clip] * loops)
|
294 |
+
final_looped_audio = looped_audio.subclip(0, target_duration)
|
295 |
+
looped_audio.close()
|
296 |
+
return final_looped_audio
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
297 |
except Exception as e:
|
298 |
+
logger.error(f"Error concatenando audio: {str(e)}")
|
299 |
+
return audio_clip.subclip(0, min(audio_clip.duration, target_duration))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
300 |
|
301 |
def extract_visual_keywords_from_script(script_text):
|
302 |
logger.info("Extrayendo palabras clave del guion")
|
303 |
if not script_text or not script_text.strip():
|
304 |
+
logger.warning("Guion vacío")
|
305 |
return ["naturaleza", "ciudad", "paisaje"]
|
|
|
306 |
clean_text = re.sub(r'[^\w\sáéíóúñÁÉÍÓÚÑ]', '', script_text)
|
|
|
|
|
307 |
if kw_model:
|
308 |
try:
|
|
|
309 |
keywords1 = kw_model.extract_keywords(clean_text, keyphrase_ngram_range=(1, 1), stop_words='spanish', top_n=5)
|
310 |
keywords2 = kw_model.extract_keywords(clean_text, keyphrase_ngram_range=(2, 2), stop_words='spanish', top_n=3)
|
|
|
311 |
all_keywords = keywords1 + keywords2
|
312 |
all_keywords.sort(key=lambda item: item[1], reverse=True)
|
313 |
+
keywords_list = []
|
314 |
seen_keywords = set()
|
315 |
+
for keyword, _ in all_keywords:
|
316 |
formatted_keyword = keyword.lower().replace(" ", "+")
|
317 |
if formatted_keyword and formatted_keyword not in seen_keywords:
|
318 |
keywords_list.append(formatted_keyword)
|
319 |
seen_keywords.add(formatted_keyword)
|
320 |
if len(keywords_list) >= 5:
|
321 |
break
|
|
|
322 |
if keywords_list:
|
323 |
logger.debug(f"Palabras clave extraídas por KeyBERT: {keywords_list}")
|
324 |
return keywords_list
|
|
|
325 |
except Exception as e:
|
326 |
+
logger.warning(f"KeyBERT falló: {str(e)}. Usando método simple.")
|
|
|
327 |
logger.debug("Extrayendo palabras clave con método simple...")
|
328 |
words = clean_text.lower().split()
|
329 |
+
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"}
|
|
|
|
|
|
|
330 |
valid_words = [word for word in words if len(word) > 3 and word not in stop_words]
|
|
|
331 |
if not valid_words:
|
332 |
+
logger.warning("No se encontraron palabras clave válidas.")
|
333 |
+
return ["espiritual", "terror", "matrix", "arcontes", "galaxia"]
|
|
|
334 |
word_counts = Counter(valid_words)
|
335 |
top_keywords = [word.replace(" ", "+") for word, _ in word_counts.most_common(5)]
|
|
|
|
|
|
|
|
|
|
|
336 |
logger.info(f"Palabras clave finales: {top_keywords}")
|
337 |
return top_keywords
|
338 |
|
|
|
339 |
def crear_video(prompt_type, input_text, selected_voice, musica_file=None):
|
340 |
logger.info("="*80)
|
341 |
logger.info(f"INICIANDO CREACIÓN DE VIDEO | Tipo: {prompt_type}")
|
342 |
logger.debug(f"Input: '{input_text[:100]}...'")
|
343 |
logger.info(f"Voz seleccionada: {selected_voice}")
|
|
|
344 |
start_time = datetime.now()
|
345 |
+
temp_dir_intermediate = tempfile.mkdtemp(prefix="video_gen_intermediate_")
|
346 |
+
logger.info(f"Directorio temporal creado: {temp_dir_intermediate}")
|
347 |
+
temp_intermediate_files = []
|
348 |
audio_tts_original = None
|
349 |
musica_audio_original = None
|
350 |
audio_tts = None
|
|
|
356 |
|
357 |
try:
|
358 |
# 1. Generar o usar guion
|
359 |
+
guion = generate_script(input_text) if prompt_type == "Generar Guion con IA" else input_text.strip()
|
|
|
|
|
|
|
|
|
360 |
logger.info(f"Guion final ({len(guion)} chars): '{guion[:100]}...'")
|
|
|
361 |
if not guion.strip():
|
|
|
362 |
raise ValueError("El guion está vacío.")
|
363 |
|
364 |
+
# 2. Generar audio de voz
|
|
|
|
|
|
|
|
|
|
|
365 |
voz_path = os.path.join(temp_dir_intermediate, "voz.mp3")
|
366 |
+
tts_voices_to_try = [selected_voice, "es-ES-JuanNeural", "es-ES-ElviraNeural"]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
367 |
tts_success = False
|
368 |
tried_voices = set()
|
369 |
|
370 |
for current_voice in tts_voices_to_try:
|
371 |
+
if current_voice in tried_voices or not current_voice:
|
372 |
+
continue
|
373 |
tried_voices.add(current_voice)
|
|
|
374 |
logger.info(f"Intentando TTS con voz: {current_voice}...")
|
375 |
+
tts_success = asyncio.run(text_to_speech(guion, voz_path, current_voice))
|
376 |
+
if tts_success:
|
377 |
+
break
|
|
|
|
|
|
|
|
|
|
|
378 |
|
379 |
if not tts_success or not os.path.exists(voz_path) or os.path.getsize(voz_path) <= 100:
|
380 |
+
raise ValueError("Error generando voz a partir del guion.")
|
|
|
381 |
|
382 |
temp_intermediate_files.append(voz_path)
|
|
|
383 |
audio_tts_original = AudioFileClip(voz_path)
|
384 |
+
if audio_tts_original.duration is None or audio_tts_original.duration <= 0:
|
385 |
+
raise ValueError("Audio de voz generado es inválido.")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
386 |
audio_tts = audio_tts_original
|
387 |
audio_duration = audio_tts_original.duration
|
388 |
logger.info(f"Duración audio voz: {audio_duration:.2f} segundos")
|
|
|
389 |
if audio_duration < 1.0:
|
390 |
+
raise ValueError("Audio de voz demasiado corto.")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
391 |
|
392 |
+
# 3. Extraer palabras clave
|
393 |
+
keywords = extract_visual_keywords_from_script(guion)
|
394 |
if not keywords:
|
395 |
+
keywords = ["video", "background"]
|
396 |
+
logger.info(f"Palabras clave: {keywords}")
|
397 |
|
398 |
# 4. Buscar y descargar videos
|
|
|
399 |
videos_data = []
|
400 |
total_desired_videos = 10
|
401 |
per_page_per_keyword = max(1, total_desired_videos // len(keywords))
|
|
|
402 |
for keyword in keywords:
|
403 |
+
if len(videos_data) >= total_desired_videos:
|
404 |
+
break
|
405 |
+
videos = buscar_videos_pexels(keyword, PEXELS_API_KEY, per_page=per_page_per_keyword)
|
406 |
+
videos_data.extend(videos)
|
|
|
|
|
|
|
|
|
407 |
|
408 |
if len(videos_data) < total_desired_videos / 2:
|
409 |
+
generic_keywords = ["mystery", "alien", "ufo", "conspiracy", "paranormal"]
|
|
|
410 |
for keyword in generic_keywords:
|
411 |
+
if len(videos_data) >= total_desired_videos:
|
412 |
+
break
|
413 |
+
videos = buscar_videos_pexels(keyword, PEXELS_API_KEY, per_page=2)
|
414 |
+
videos_data.extend(videos)
|
|
|
|
|
|
|
|
|
415 |
|
416 |
if not videos_data:
|
417 |
+
raise ValueError("No se encontraron videos en Pexels.")
|
|
|
418 |
|
419 |
video_paths = []
|
|
|
420 |
for video in videos_data:
|
421 |
if 'video_files' not in video or not video['video_files']:
|
|
|
422 |
continue
|
423 |
+
best_quality = max(video['video_files'], key=lambda x: x.get('width', 0) * x.get('height', 0), default=None)
|
424 |
+
if best_quality and 'link' in best_quality:
|
425 |
+
path = download_video_file(best_quality['link'], temp_dir_intermediate)
|
426 |
+
if path:
|
427 |
+
video_paths.append(path)
|
428 |
+
temp_intermediate_files.append(path)
|
429 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
430 |
if not video_paths:
|
431 |
+
raise ValueError("No se descargaron videos utilizables.")
|
|
|
432 |
|
433 |
# 5. Procesar y concatenar clips de video
|
|
|
434 |
current_duration = 0
|
435 |
min_clip_duration = 0.5
|
436 |
max_clip_segment = 10.0
|
|
|
437 |
for i, path in enumerate(video_paths):
|
438 |
if current_duration >= audio_duration + max_clip_segment:
|
|
|
439 |
break
|
|
|
|
|
440 |
try:
|
|
|
441 |
clip = VideoFileClip(path)
|
442 |
source_clips.append(clip)
|
443 |
+
if clip.duration is None or clip.duration <= 0:
|
|
|
|
|
444 |
continue
|
|
|
445 |
remaining_needed = audio_duration - current_duration
|
446 |
+
segment_duration = min(clip.duration, max_clip_segment, remaining_needed + min_clip_duration)
|
447 |
+
if segment_duration >= min_clip_duration:
|
448 |
+
sub = clip.subclip(0, segment_duration)
|
449 |
+
clips_to_concatenate.append(sub)
|
450 |
+
current_duration += sub.duration
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
451 |
except Exception as e:
|
452 |
+
logger.warning(f"Error procesando video {path}: {str(e)}")
|
|
|
|
|
|
|
453 |
|
454 |
if not clips_to_concatenate:
|
455 |
+
raise ValueError("No hay segmentos de video válidos.")
|
|
|
456 |
|
457 |
+
video_base = concatenate_videoclips(clips_to_concatenate, method="chain")
|
458 |
+
if video_base.duration is None or video_base.duration <= 0:
|
459 |
+
raise ValueError("Video base inválido.")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
460 |
|
461 |
+
# Ajustar duración del video
|
462 |
+
if video_base.duration < audio_duration:
|
463 |
+
num_full_repeats = int(audio_duration // video_base.duration)
|
464 |
+
remaining_duration = audio_duration % video_base.duration
|
465 |
+
repeated_clips_list = [video_base] * num_full_repeats
|
|
|
|
|
466 |
if remaining_duration > 0:
|
467 |
+
remaining_clip = video_base.subclip(0, remaining_duration)
|
468 |
+
repeated_clips_list.append(remaining_clip)
|
469 |
+
video_base = concatenate_videoclips(repeated_clips_list, method="chain")
|
470 |
+
elif video_base.duration > audio_duration:
|
471 |
+
video_base = video_base.subclip(0, audio_duration)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
472 |
|
473 |
# 6. Manejar música de fondo
|
474 |
+
final_audio = audio_tts
|
|
|
|
|
|
|
|
|
|
|
475 |
if musica_file:
|
|
|
476 |
try:
|
477 |
music_path = os.path.join(temp_dir_intermediate, "musica_bg.mp3")
|
478 |
shutil.copyfile(musica_file, music_path)
|
479 |
temp_intermediate_files.append(music_path)
|
|
|
|
|
480 |
musica_audio_original = AudioFileClip(music_path)
|
481 |
+
if musica_audio_original.duration > 0:
|
482 |
+
musica_audio = loop_audio_to_length(musica_audio_original, video_base.duration)
|
483 |
+
final_audio = CompositeAudioClip([
|
484 |
+
musica_audio.volumex(0.2),
|
485 |
+
audio_tts.volumex(1.0)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
486 |
])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
487 |
except Exception as e:
|
488 |
+
logger.warning(f"Error procesando música: {str(e)}")
|
489 |
+
final_audio = audio_tts
|
490 |
|
491 |
+
if abs(final_audio.duration - video_base.duration) > 0.2:
|
492 |
+
final_audio = final_audio.subclip(0, video_base.duration)
|
493 |
|
494 |
+
# 7. Combinar audio y video
|
495 |
+
video_final = video_base.set_audio(final_audio)
|
496 |
+
output_filename = f"video_{int(datetime.now().timestamp())}.mp4"
|
497 |
output_path = os.path.join(temp_dir_intermediate, output_filename)
|
498 |
permanent_path = f"/tmp/{output_filename}"
|
499 |
+
|
|
|
500 |
video_final.write_videofile(
|
501 |
output_path,
|
502 |
fps=24,
|
|
|
504 |
codec="libx264",
|
505 |
audio_codec="aac",
|
506 |
preset="medium",
|
507 |
+
ffmpeg_params=['-vf', 'scale=1920:1080:force_original_aspect_ratio=decrease,pad=1920:1080:-1:-1:color=black', '-crf', '23'],
|
|
|
|
|
|
|
508 |
logger='bar'
|
509 |
)
|
510 |
|
511 |
+
shutil.copy(output_path, permanent_path)
|
512 |
+
logger.info(f"Video guardado en: {permanent_path}")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
513 |
total_time = (datetime.now() - start_time).total_seconds()
|
514 |
+
logger.info(f"Video generado en {total_time:.2f}s")
|
|
|
515 |
return permanent_path
|
516 |
|
517 |
except ValueError as ve:
|
518 |
+
logger.error(f"Error controlado: {str(ve)}")
|
519 |
+
raise
|
520 |
except Exception as e:
|
521 |
+
logger.critical(f"Error crítico: {str(e)}")
|
522 |
+
raise
|
523 |
finally:
|
524 |
+
for clip in source_clips + clips_to_concatenate:
|
|
|
|
|
525 |
try:
|
526 |
clip.close()
|
527 |
+
except:
|
528 |
+
pass
|
529 |
+
if audio_tts_original:
|
530 |
+
try:
|
531 |
+
audio_tts_original.close()
|
532 |
+
except:
|
533 |
+
pass
|
534 |
+
if musica_audio:
|
|
|
|
|
535 |
try:
|
536 |
musica_audio.close()
|
537 |
+
except:
|
538 |
+
pass
|
539 |
+
if musica_audio_original:
|
540 |
+
try:
|
541 |
+
musica_audio_original.close()
|
542 |
+
except:
|
543 |
+
pass
|
544 |
+
if video_base:
|
545 |
+
try:
|
546 |
+
video_base.close()
|
547 |
+
except:
|
548 |
+
pass
|
549 |
+
if video_final:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
550 |
try:
|
551 |
video_final.close()
|
552 |
+
except:
|
553 |
+
pass
|
554 |
+
for path in temp_intermediate_files:
|
555 |
+
if os.path.isfile(path) and path != permanent_path:
|
556 |
+
try:
|
557 |
+
os.remove(path)
|
558 |
+
except:
|
559 |
+
logger.warning(f"No se pudo eliminar {path}")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
560 |
|
561 |
+
def run_app(prompt_type, prompt_ia, prompt_manual, musica_file, selected_voice):
|
|
|
562 |
logger.info("="*80)
|
563 |
logger.info("SOLICITUD RECIBIDA EN INTERFAZ")
|
|
|
|
|
564 |
input_text = prompt_ia if prompt_type == "Generar Guion con IA" else prompt_manual
|
|
|
565 |
output_video = None
|
566 |
output_file = None
|
567 |
+
status_msg = gr.update(value="⏳ Procesando...")
|
568 |
|
569 |
if not input_text or not input_text.strip():
|
570 |
logger.warning("Texto de entrada vacío.")
|
571 |
+
return None, None, gr.update(value="⚠️ Ingresa texto para el guion o tema.")
|
|
|
572 |
|
|
|
|
|
573 |
voice_ids_disponibles = [v[1] for v in AVAILABLE_VOICES]
|
574 |
if selected_voice not in voice_ids_disponibles:
|
575 |
+
logger.warning(f"Voz inválida: '{selected_voice}'. Usando voz por defecto: {DEFAULT_VOICE_ID}")
|
576 |
+
selected_voice = DEFAULT_VOICE_ID
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
577 |
|
578 |
try:
|
579 |
+
video_path = crear_video(prompt_type, input_text, selected_voice, musica_file)
|
|
|
|
|
|
|
580 |
if video_path and os.path.exists(video_path):
|
581 |
+
output_video = video_path
|
582 |
+
output_file = video_path
|
583 |
+
status_msg = gr.update(value="✅ Video generado exitosamente.")
|
|
|
|
|
584 |
else:
|
585 |
+
status_msg = gr.update(value="❌ Error: Falló la generación del video.")
|
|
|
|
|
586 |
except ValueError as ve:
|
587 |
+
logger.warning(f"Error de validación: {str(ve)}")
|
588 |
+
status_msg = gr.update(value=f"⚠️ Error: {str(ve)}")
|
589 |
except Exception as e:
|
590 |
+
logger.critical(f"Error crítico: {str(e)}")
|
591 |
+
status_msg = gr.update(value=f"❌ Error inesperado: {str(e)}")
|
592 |
+
return output_video, output_file, status_msg
|
|
|
|
|
|
|
593 |
|
594 |
# Interfaz de Gradio
|
595 |
+
with gr.Blocks(title="Generador de Videos con IA", theme=gr.themes.Soft()) as app:
|
|
|
|
|
|
|
|
|
596 |
gr.Markdown("# 🎬 Generador Automático de Videos con IA")
|
597 |
gr.Markdown("Genera videos cortos a partir de un tema o guion, usando imágenes de archivo de Pexels y voz generada.")
|
598 |
|
|
|
603 |
label="Método de Entrada",
|
604 |
value="Generar Guion con IA"
|
605 |
)
|
|
|
|
|
606 |
with gr.Column(visible=True) as ia_guion_column:
|
607 |
prompt_ia = gr.Textbox(
|
608 |
label="Tema para IA",
|
609 |
lines=2,
|
610 |
+
placeholder="Ej: Un paisaje natural con montañas y ríos al amanecer...",
|
611 |
+
max_lines=4
|
|
|
612 |
)
|
|
|
613 |
with gr.Column(visible=False) as manual_guion_column:
|
614 |
prompt_manual = gr.Textbox(
|
615 |
label="Tu Guion Completo",
|
616 |
lines=5,
|
617 |
+
placeholder="Ej: En este video exploraremos los misterios del océano...",
|
618 |
+
max_lines=10
|
|
|
619 |
)
|
|
|
620 |
musica_input = gr.Audio(
|
621 |
label="Música de fondo (opcional)",
|
622 |
type="filepath",
|
623 |
+
interactive=True
|
|
|
624 |
)
|
|
|
|
|
625 |
voice_dropdown = gr.Dropdown(
|
626 |
label="Seleccionar Voz para Guion",
|
627 |
choices=AVAILABLE_VOICES,
|
628 |
value=DEFAULT_VOICE_ID,
|
629 |
interactive=True
|
630 |
)
|
|
|
|
|
631 |
generate_btn = gr.Button("✨ Generar Video", variant="primary")
|
|
|
632 |
with gr.Column():
|
633 |
video_output = gr.Video(
|
634 |
label="Previsualización del Video Generado",
|
|
|
643 |
status_output = gr.Textbox(
|
644 |
label="Estado",
|
645 |
interactive=False,
|
|
|
646 |
placeholder="Esperando acción...",
|
647 |
value="Esperando entrada..."
|
648 |
)
|
649 |
|
|
|
650 |
prompt_type.change(
|
651 |
+
lambda x: (gr.update(visible=x == "Generar Guion con IA"), gr.update(visible=x == "Usar Mi Guion")),
|
|
|
652 |
inputs=prompt_type,
|
653 |
outputs=[ia_guion_column, manual_guion_column]
|
654 |
)
|
655 |
|
|
|
656 |
generate_btn.click(
|
657 |
+
lambda: (None, None, gr.update(value="⏳ Procesando... Esto puede tomar varios minutos.")),
|
658 |
+
outputs=[video_output, file_output, status_output]
|
|
|
659 |
).then(
|
660 |
run_app,
|
661 |
inputs=[prompt_type, prompt_ia, prompt_manual, musica_input, voice_dropdown],
|
|
|
668 |
|
669 |
gr.Markdown("### Instrucciones:")
|
670 |
gr.Markdown("""
|
671 |
+
1. Configura la variable de entorno `PEXELS_API_KEY`.
|
672 |
+
2. Selecciona el tipo de entrada: "Generar Guion con IA" o "Usar Mi Guion".
|
673 |
+
3. Sube música (opcional).
|
674 |
+
4. Selecciona la voz.
|
675 |
+
5. Haz clic en "✨ Generar Video".
|
676 |
+
6. Revisa el estado y descarga el video generado.
|
677 |
+
7. Consulta `video_generator_full.log` para errores.
|
|
|
678 |
""")
|
|
|
|
|
679 |
|
680 |
if __name__ == "__main__":
|
681 |
+
logger.info("Verificando dependencias...")
|
682 |
try:
|
683 |
from moviepy.editor import ColorClip
|
684 |
+
temp_clip = ColorClip((100,100), color=(255,0,0), duration=0.1)
|
685 |
+
temp_clip.close()
|
686 |
+
logger.info("MoviePy y FFmpeg accesibles.")
|
|
|
|
|
|
|
687 |
except Exception as e:
|
688 |
+
logger.critical(f"Fallo en dependencias: {e}")
|
689 |
+
raise
|
690 |
+
os.environ['GRADIO_SERVER_TIMEOUT'] = '6000'
|
|
|
|
|
691 |
logger.info("Iniciando aplicación Gradio...")
|
692 |
try:
|
693 |
app.launch(server_name="0.0.0.0", server_port=7860, share=False)
|
694 |
except Exception as e:
|
695 |
+
logger.critical(f"No se pudo iniciar la app: {str(e)}")
|
696 |
raise
|