Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -7,21 +7,16 @@ from datetime import datetime
|
|
7 |
import edge_tts
|
8 |
import gradio as gr
|
9 |
import torch
|
10 |
-
# Corregir typo: GPT2Tokenizer en lugar de GPTizer
|
11 |
from transformers import GPT2Tokenizer, GPT2LMHeadModel
|
12 |
from keybert import KeyBERT
|
13 |
-
# Importar concatenate_audioclips específicamente si no está en editor por defecto
|
14 |
from moviepy.editor import VideoFileClip, concatenate_videoclips, AudioFileClip, CompositeAudioClip, concatenate_audioclips
|
15 |
import re
|
16 |
import math
|
17 |
-
# Pydub ya no parece usarse directamente en las funciones principales,
|
18 |
-
# pero se mantiene por si acaso o por si se usó en intentos previos.
|
19 |
-
# from pydub import AudioSegment # Comentado porque no se usa en el flujo principal corregido
|
20 |
-
from collections import Counter
|
21 |
import shutil
|
22 |
import json
|
|
|
23 |
|
24 |
-
#
|
25 |
logging.basicConfig(
|
26 |
level=logging.DEBUG,
|
27 |
format='%(asctime)s - %(levelname)s - %(message)s',
|
@@ -32,52 +27,46 @@ logging.basicConfig(
|
|
32 |
)
|
33 |
logger = logging.getLogger(__name__)
|
34 |
logger.info("="*80)
|
35 |
-
logger.info("
|
36 |
logger.info("="*80)
|
37 |
|
38 |
-
#
|
39 |
-
# Asegúrate de que esta variable de entorno esté configurada en tu entorno.
|
40 |
PEXELS_API_KEY = os.environ.get("PEXELS_API_KEY")
|
41 |
if not PEXELS_API_KEY:
|
42 |
-
logger.critical("
|
43 |
-
#
|
44 |
-
# raise ValueError("API key
|
45 |
-
|
46 |
-
|
47 |
-
logger.info("API key de Pexels configurada correctamente")
|
48 |
-
|
49 |
-
# Inicialización de modelos CON LOGS DETALLADOS
|
50 |
MODEL_NAME = "datificate/gpt2-small-spanish"
|
51 |
-
logger.info(f"
|
52 |
-
tokenizer = None
|
53 |
-
model = None
|
54 |
try:
|
55 |
-
# CORRECCIÓN: Usar GPT2Tokenizer
|
56 |
tokenizer = GPT2Tokenizer.from_pretrained(MODEL_NAME)
|
57 |
model = GPT2LMHeadModel.from_pretrained(MODEL_NAME).eval()
|
58 |
if tokenizer.pad_token is None:
|
59 |
tokenizer.pad_token = tokenizer.eos_token
|
60 |
-
logger.info(f"
|
61 |
except Exception as e:
|
62 |
-
logger.error(f"
|
63 |
-
tokenizer = model = None
|
64 |
|
65 |
-
logger.info("
|
66 |
-
kw_model = None
|
67 |
try:
|
68 |
kw_model = KeyBERT('distilbert-base-multilingual-cased')
|
69 |
-
logger.info("KeyBERT
|
70 |
except Exception as e:
|
71 |
-
logger.error(f"
|
72 |
-
kw_model = None
|
73 |
|
74 |
-
# [FUNCIÓN BUSCAR_VIDEOS_PEXELS ORIGINAL CON LOGS AÑADIDOS]
|
75 |
def buscar_videos_pexels(query, api_key, per_page=5):
|
76 |
if not api_key:
|
77 |
-
logger.warning("
|
78 |
return []
|
79 |
|
80 |
-
logger.debug(f"
|
81 |
headers = {"Authorization": api_key}
|
82 |
try:
|
83 |
params = {
|
@@ -87,368 +76,301 @@ def buscar_videos_pexels(query, api_key, per_page=5):
|
|
87 |
"size": "medium"
|
88 |
}
|
89 |
|
90 |
-
logger.debug(f"Params: {params}")
|
91 |
response = requests.get(
|
92 |
"https://api.pexels.com/videos/search",
|
93 |
headers=headers,
|
94 |
params=params,
|
95 |
timeout=20
|
96 |
)
|
97 |
-
response.raise_for_status()
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
return videos
|
104 |
-
except json.JSONDecodeError:
|
105 |
-
logger.error(f"Pexels: JSON inválido recibido | Status: {response.status_code} | Respuesta: {response.text[:200]}...")
|
106 |
-
return []
|
107 |
|
108 |
except requests.exceptions.RequestException as e:
|
109 |
-
logger.error(f"
|
|
|
|
|
110 |
except Exception as e:
|
111 |
-
logger.error(f"
|
112 |
|
113 |
return []
|
114 |
|
115 |
-
# [FUNCIÓN GENERATE_SCRIPT ORIGINAL CON LOGS]
|
116 |
def generate_script(prompt, max_length=150):
|
117 |
-
logger.info(f"
|
118 |
if not tokenizer or not model:
|
119 |
-
logger.warning("
|
120 |
return prompt
|
121 |
|
122 |
try:
|
123 |
-
|
124 |
-
# (Esto es una heurística simple, podrías mejorarlo)
|
125 |
-
if len(prompt.split()) < 5 or not re.search(r'[¿?.]', prompt):
|
126 |
-
enhanced_prompt = f"Escribe un guion corto, interesante y coherente sobre: {prompt}"
|
127 |
-
else:
|
128 |
-
enhanced_prompt = prompt # Si parece ya un guion o pregunta, úsalo directamente
|
129 |
-
logger.debug("Usando prompt original como base para la generación.")
|
130 |
-
|
131 |
inputs = tokenizer(enhanced_prompt, return_tensors="pt", truncation=True, max_length=512)
|
132 |
|
133 |
-
logger.debug("Generando texto con GPT-2...")
|
134 |
-
# Asegurar que el modelo esté en modo evaluación y en la CPU si no hay GPU
|
135 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
136 |
model.to(device)
|
137 |
inputs = {k: v.to(device) for k, v in inputs.items()}
|
138 |
|
139 |
-
|
140 |
outputs = model.generate(
|
141 |
**inputs,
|
142 |
max_length=max_length,
|
143 |
-
do_sample=True,
|
144 |
top_p=0.9,
|
145 |
top_k=40,
|
146 |
temperature=0.7,
|
147 |
-
repetition_penalty=1.2,
|
148 |
pad_token_id=tokenizer.pad_token_id,
|
149 |
eos_token_id=tokenizer.eos_token_id,
|
150 |
-
no_repeat_ngram_size=3
|
151 |
)
|
152 |
|
153 |
text = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
154 |
|
155 |
-
|
156 |
-
text = re.sub(r'<[^>]+>', '', text) # Elimina posibles tags HTML/XML si los hay
|
157 |
text = text.strip()
|
158 |
|
159 |
-
# Intenta obtener al menos una oración completa si es posible
|
160 |
sentences = text.split('.')
|
161 |
if sentences and sentences[0].strip():
|
162 |
final_text = sentences[0].strip() + '.'
|
163 |
-
|
164 |
-
if len(sentences) > 1 and sentences[1].strip() and len(final_text.split()) < max_length * 0.5: # Si la primera es corta y el guion no muy largo aún
|
165 |
final_text += " " + sentences[1].strip() + "."
|
166 |
-
final_text = final_text.replace("..", ".")
|
167 |
|
168 |
-
logger.info(f"
|
169 |
-
return final_text.strip()
|
170 |
|
171 |
-
logger.info(f"
|
172 |
-
return text.strip()
|
|
|
173 |
except Exception as e:
|
174 |
-
logger.error(f"Error
|
175 |
-
logger.warning("
|
176 |
-
return prompt.strip()
|
177 |
|
178 |
-
# [FUNCIÓN TEXT_TO_SPEECH ORIGINAL CON LOGS]
|
179 |
async def text_to_speech(text, output_path, voice="es-ES-ElviraNeural"):
|
180 |
-
logger.info(f"
|
181 |
if not text or not text.strip():
|
182 |
-
logger.warning("
|
183 |
return False
|
184 |
|
185 |
try:
|
186 |
communicate = edge_tts.Communicate(text, voice)
|
187 |
await communicate.save(output_path)
|
188 |
|
189 |
-
if os.path.exists(output_path) and os.path.getsize(output_path) > 100:
|
190 |
-
logger.info(f"Audio
|
191 |
return True
|
192 |
else:
|
193 |
-
logger.error(f"TTS
|
194 |
return False
|
195 |
|
196 |
except Exception as e:
|
197 |
-
logger.error(f"Error
|
198 |
return False
|
199 |
|
200 |
-
# [FUNCIÓN DOWNLOAD_VIDEO_FILE ORIGINAL CON LOGS]
|
201 |
def download_video_file(url, temp_dir):
|
202 |
if not url:
|
203 |
-
logger.warning("URL
|
204 |
return None
|
205 |
|
206 |
try:
|
207 |
-
logger.info(f"
|
208 |
-
# Usar un nombre más robusto y asegurar que el directorio exista
|
209 |
os.makedirs(temp_dir, exist_ok=True)
|
210 |
file_name = f"video_dl_{datetime.now().strftime('%Y%m%d_%H%M%S_%f')}.mp4"
|
211 |
output_path = os.path.join(temp_dir, file_name)
|
212 |
|
213 |
-
# Aumentar timeout para descargas grandes
|
214 |
with requests.get(url, stream=True, timeout=60) as r:
|
215 |
r.raise_for_status()
|
216 |
-
total_size = int(r.headers.get('content-length', 0))
|
217 |
-
downloaded_size = 0
|
218 |
-
logger.debug(f"Tamaño esperado: {total_size} bytes")
|
219 |
-
|
220 |
with open(output_path, 'wb') as f:
|
221 |
for chunk in r.iter_content(chunk_size=8192):
|
222 |
f.write(chunk)
|
223 |
-
downloaded_size += len(chunk)
|
224 |
-
# Opcional: loguear progreso si es muy lento
|
225 |
|
226 |
-
|
227 |
-
|
228 |
-
logger.info(f"Video descargado exitosamente: {output_path} | Tamaño: {os.path.getsize(output_path)} bytes")
|
229 |
return output_path
|
230 |
else:
|
231 |
-
logger.warning(f"
|
232 |
if os.path.exists(output_path):
|
233 |
-
os.remove(output_path)
|
234 |
return None
|
235 |
|
236 |
except requests.exceptions.RequestException as e:
|
237 |
-
logger.error(f"
|
238 |
except Exception as e:
|
239 |
-
logger.error(f"
|
240 |
|
241 |
return None
|
242 |
|
243 |
-
# [FUNCIÓN LOOP_AUDIO_TO_LENGTH ORIGINAL CON LOGS]
|
244 |
def loop_audio_to_length(audio_clip, target_duration):
|
245 |
-
logger.debug(f"
|
246 |
if audio_clip.duration <= 0:
|
247 |
-
logger.warning("
|
248 |
-
return AudioFileClip(filename="")
|
249 |
|
250 |
if audio_clip.duration >= target_duration:
|
251 |
-
logger.debug("
|
252 |
return audio_clip.subclip(0, target_duration)
|
253 |
|
254 |
-
# CORRECCIÓN: Usar concatenate_audioclips para audio
|
255 |
loops = math.ceil(target_duration / audio_clip.duration)
|
256 |
-
logger.debug(f"
|
257 |
|
258 |
-
# Crear una lista de clips de audio
|
259 |
audio_segments = [audio_clip] * loops
|
260 |
-
|
261 |
-
# Concatenar los clips de audio
|
262 |
looped_audio = concatenate_audioclips(audio_segments)
|
263 |
|
264 |
-
# Recortar al target_duration exacto
|
265 |
return looped_audio.subclip(0, target_duration)
|
266 |
|
267 |
-
# [FUNCIÓN EXTRACT_VISUAL_KEYWORDS_FROM_SCRIPT ORIGINAL CON LOGS]
|
268 |
def extract_visual_keywords_from_script(script_text):
|
269 |
-
logger.info("
|
270 |
if not script_text or not script_text.strip():
|
271 |
-
logger.warning("
|
272 |
-
return ["naturaleza", "ciudad", "paisaje"]
|
273 |
|
274 |
-
# Limpieza de texto más robusta, manteniendo espacios para keyphrases
|
275 |
clean_text = re.sub(r'[^\w\sáéíóúñÁÉÍÓÚÑ]', '', script_text)
|
276 |
-
|
277 |
keywords_list = []
|
278 |
|
279 |
if kw_model:
|
280 |
try:
|
281 |
-
logger.debug("
|
282 |
-
|
283 |
-
|
284 |
-
clean_text,
|
285 |
-
keyphrase_ngram_range=(1, 1),
|
286 |
-
stop_words='spanish',
|
287 |
-
top_n=5
|
288 |
-
)
|
289 |
-
keywords2 = kw_model.extract_keywords(
|
290 |
-
clean_text,
|
291 |
-
keyphrase_ngram_range=(2, 2),
|
292 |
-
stop_words='spanish',
|
293 |
-
top_n=3
|
294 |
-
)
|
295 |
|
296 |
-
# Combinar y priorizar palabras clave (ngram 1) y frases clave (ngram 2)
|
297 |
all_keywords = keywords1 + keywords2
|
298 |
-
# Ordenar por score descendente
|
299 |
all_keywords.sort(key=lambda item: item[1], reverse=True)
|
300 |
|
301 |
-
# Tomar los top N únicos
|
302 |
seen_keywords = set()
|
303 |
for keyword, score in all_keywords:
|
304 |
-
# Convertir a minúsculas y reemplazar espacios por + para URL
|
305 |
formatted_keyword = keyword.lower().replace(" ", "+")
|
306 |
-
if formatted_keyword not in seen_keywords:
|
307 |
keywords_list.append(formatted_keyword)
|
308 |
seen_keywords.add(formatted_keyword)
|
309 |
-
if len(keywords_list) >= 5:
|
310 |
break
|
311 |
|
312 |
if keywords_list:
|
313 |
-
logger.debug(f"KeyBERT keywords
|
314 |
return keywords_list
|
315 |
|
316 |
except Exception as e:
|
317 |
-
logger.warning(f"KeyBERT
|
318 |
-
# Continúa al método simple si KeyBERT falla
|
319 |
|
320 |
-
|
321 |
-
logger.debug("Extrayendo keywords con método simple...")
|
322 |
words = clean_text.lower().split()
|
323 |
-
|
324 |
-
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", "que", "porque", "si", "aunque", "mientras", "cuando", "como", "donde", "siempre", "nunca", "jamás", "muy", "más", "menos", "tan", "tanto", "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", "siempre", "nunca", "jamás", "quizá", "acaso", "tal vez"}
|
325 |
|
326 |
-
# Filtrar palabras, eliminar stop words y palabras cortas
|
327 |
valid_words = [word for word in words if len(word) > 3 and word not in stop_words]
|
328 |
|
329 |
if not valid_words:
|
330 |
-
logger.warning("No
|
331 |
return ["naturaleza", "ciudad", "paisaje"]
|
332 |
|
333 |
-
# Contar frecuencia y tomar las más comunes
|
334 |
word_counts = Counter(valid_words)
|
335 |
-
# Tomar hasta 5 palabras clave más comunes
|
336 |
top_keywords = [word.replace(" ", "+") for word, _ in word_counts.most_common(5)]
|
337 |
|
338 |
if not top_keywords:
|
339 |
-
logger.warning("
|
340 |
return ["naturaleza", "ciudad", "paisaje"]
|
341 |
|
342 |
-
logger.info(f"
|
343 |
return top_keywords
|
344 |
|
345 |
-
# [FUNCIÓN CREAR_VIDEO CORREGIDA]
|
346 |
def crear_video(prompt_type, input_text, musica_file=None):
|
347 |
logger.info("="*80)
|
348 |
-
logger.info(f"
|
349 |
logger.debug(f"Input: '{input_text[:100]}...'")
|
350 |
|
351 |
start_time = datetime.now()
|
352 |
-
temp_dir_intermediate = None
|
353 |
-
|
354 |
-
|
355 |
try:
|
356 |
-
# 1.
|
357 |
if prompt_type == "Generar Guion con IA":
|
358 |
-
logger.info("Generando guion con IA...")
|
359 |
guion = generate_script(input_text)
|
360 |
else:
|
361 |
-
|
362 |
-
guion = input_text.strip() # Asegurar que el guion manual no tenga espacios al inicio/fin
|
363 |
|
364 |
-
logger.info(f"
|
365 |
|
366 |
if not guion.strip():
|
367 |
-
logger.error("
|
368 |
-
raise ValueError("
|
369 |
|
370 |
-
# Crear directorio temporal para archivos intermedios (audio TTS, videos descargados)
|
371 |
temp_dir_intermediate = tempfile.mkdtemp(prefix="video_gen_intermediate_")
|
372 |
-
logger.info(f"
|
373 |
-
temp_intermediate_files = []
|
374 |
|
375 |
-
# 2.
|
376 |
-
logger.info("
|
377 |
voz_path = os.path.join(temp_dir_intermediate, "voz.mp3")
|
378 |
-
# No añadimos voz_path a temp_intermediate_files *antes* de la generación,
|
379 |
-
# solo si la generación es exitosa.
|
380 |
if not asyncio.run(text_to_speech(guion, voz_path)):
|
381 |
-
logger.error("
|
382 |
-
raise ValueError("Error
|
383 |
-
temp_intermediate_files.append(voz_path)
|
384 |
|
385 |
audio_tts = AudioFileClip(voz_path)
|
386 |
audio_duration = audio_tts.duration
|
387 |
-
logger.info(f"
|
388 |
|
389 |
if audio_duration < 1.0:
|
390 |
-
logger.warning(f"
|
391 |
-
# Podrías añadir un mínimo o generar un guion/voz alternativo
|
392 |
|
393 |
-
# 3.
|
394 |
-
logger.info("
|
395 |
try:
|
396 |
keywords = extract_visual_keywords_from_script(guion)
|
397 |
-
logger.info(f"
|
398 |
except Exception as e:
|
399 |
-
logger.error(f"Error
|
400 |
-
keywords = ["naturaleza", "paisaje"]
|
401 |
|
402 |
if not keywords:
|
403 |
-
|
404 |
-
keywords = ["video", "background"] # Fallback final
|
405 |
|
406 |
-
# 4.
|
407 |
-
logger.info("
|
408 |
videos_data = []
|
409 |
-
|
410 |
-
|
411 |
-
per_page_per_keyword = max(1, total_desired_videos // len(keywords)) # Distribuir la búsqueda
|
412 |
|
413 |
for keyword in keywords:
|
414 |
-
if len(videos_data) >= total_desired_videos: break
|
415 |
try:
|
416 |
videos = buscar_videos_pexels(keyword, PEXELS_API_KEY, per_page=per_page_per_keyword)
|
417 |
if videos:
|
418 |
videos_data.extend(videos)
|
419 |
-
logger.info(f"
|
420 |
except Exception as e:
|
421 |
-
logger.warning(f"Error
|
422 |
|
423 |
-
|
424 |
-
|
425 |
-
|
426 |
-
generic_keywords = ["nature", "city", "background", "abstract"] # Palabras clave genéricas en inglés para Pexels
|
427 |
for keyword in generic_keywords:
|
428 |
if len(videos_data) >= total_desired_videos: break
|
429 |
try:
|
430 |
-
|
431 |
-
videos = buscar_videos_pexels(keyword, PEXELS_API_KEY, per_page=2)
|
432 |
if videos:
|
433 |
videos_data.extend(videos)
|
434 |
-
logger.info(f"
|
435 |
except Exception as e:
|
436 |
-
logger.warning(f"Error
|
437 |
|
438 |
|
439 |
if not videos_data:
|
440 |
-
logger.error("No
|
441 |
-
raise ValueError("No
|
442 |
|
443 |
video_paths = []
|
444 |
-
logger.info(f"
|
445 |
for video in videos_data:
|
446 |
if 'video_files' not in video or not video['video_files']:
|
447 |
-
logger.debug(f"
|
448 |
continue
|
449 |
|
450 |
try:
|
451 |
-
# Encontrar la mejor calidad disponible con un enlace de descarga
|
452 |
best_quality = None
|
453 |
for vf in sorted(video['video_files'], key=lambda x: x.get('width', 0) * x.get('height', 0), reverse=True):
|
454 |
if 'link' in vf:
|
@@ -459,341 +381,287 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
459 |
path = download_video_file(best_quality['link'], temp_dir_intermediate)
|
460 |
if path:
|
461 |
video_paths.append(path)
|
462 |
-
temp_intermediate_files.append(path)
|
463 |
-
logger.info(f"Video
|
464 |
else:
|
465 |
-
logger.warning(f"
|
466 |
else:
|
467 |
-
logger.warning(f"No
|
468 |
|
469 |
except Exception as e:
|
470 |
-
logger.warning(f"Error
|
471 |
|
472 |
-
logger.info(f"
|
473 |
if not video_paths:
|
474 |
-
logger.error("
|
475 |
-
raise ValueError("
|
476 |
|
477 |
-
# 5.
|
478 |
-
logger.info("
|
479 |
clips = []
|
480 |
current_duration = 0
|
481 |
-
min_clip_duration = 1.0
|
482 |
-
max_clip_segment = 8.0
|
483 |
-
|
484 |
-
# Ordenar clips descargados por tamaño o duración si se desea, o usar el orden de descarga
|
485 |
-
# Por ahora, usamos el orden de descarga
|
486 |
|
487 |
-
for path in video_paths:
|
488 |
-
|
489 |
-
|
490 |
-
if current_duration >= audio_duration + max_clip_segment:
|
491 |
break
|
492 |
|
|
|
493 |
try:
|
494 |
-
logger.debug(f"
|
495 |
clip = VideoFileClip(path)
|
496 |
-
|
497 |
-
#
|
|
|
|
|
|
|
|
|
|
|
498 |
remaining_needed = audio_duration - current_duration
|
499 |
-
|
500 |
-
|
501 |
-
# Si necesitamos más duración, tomamos hasta el final del clip o el max_clip_segment
|
502 |
if remaining_needed > 0:
|
503 |
-
|
504 |
-
|
505 |
-
|
506 |
-
|
507 |
-
|
508 |
-
|
509 |
-
|
510 |
-
|
511 |
-
|
512 |
-
|
513 |
-
|
514 |
-
|
|
|
|
|
|
|
515 |
else:
|
516 |
-
logger.debug(f"
|
517 |
-
|
518 |
-
# Cerrar clip para liberar recursos
|
519 |
-
clip.close()
|
520 |
|
521 |
except Exception as e:
|
522 |
-
logger.warning(f"Error
|
523 |
-
# Continuar con el siguiente clip si hay un error con este
|
524 |
continue
|
525 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
526 |
if not clips:
|
527 |
-
logger.error("No
|
528 |
-
raise ValueError("No
|
529 |
|
530 |
-
logger.info(f"
|
531 |
video_base = concatenate_videoclips(clips, method="compose")
|
532 |
-
logger.info(f"
|
533 |
|
534 |
-
# Asegurar que el video base tenga al menos la duración del audio
|
535 |
if video_base.duration < audio_duration:
|
536 |
-
# Calcular cuántas veces se necesita repetir el video base o parte de él
|
537 |
-
# Un método simple es repetir el video base completo
|
538 |
num_repeats = math.ceil(audio_duration / video_base.duration)
|
539 |
-
logger.info(f"
|
540 |
repeated_clips = [video_base] * num_repeats
|
541 |
video_base = concatenate_videoclips(repeated_clips, method="compose").subclip(0, audio_duration)
|
542 |
-
logger.info(f"
|
543 |
|
544 |
-
# Si el video base es más largo que el audio (puede pasar si el último clip añadido fue largo)
|
545 |
-
# Recortarlo exactamente a la duración del audio
|
546 |
if video_base.duration > audio_duration:
|
547 |
-
logger.info(f"
|
548 |
video_base = video_base.subclip(0, audio_duration)
|
549 |
-
logger.info(f"
|
550 |
|
551 |
|
552 |
-
# 6.
|
553 |
-
logger.info("
|
554 |
-
final_audio = audio_tts
|
555 |
|
556 |
if musica_file:
|
557 |
try:
|
558 |
-
# Copiar archivo de música a un lugar temporal seguro
|
559 |
music_path = os.path.join(temp_dir_intermediate, "musica_bg.mp3")
|
560 |
shutil.copyfile(musica_file, music_path)
|
561 |
-
temp_intermediate_files.append(music_path)
|
562 |
-
logger.info(f"
|
563 |
|
564 |
musica_audio = AudioFileClip(music_path)
|
565 |
-
logger.debug(f"
|
566 |
|
567 |
-
# Ajustar duración de la música al video final (que ya tiene la duración del audio TTS)
|
568 |
musica_audio = loop_audio_to_length(musica_audio, video_base.duration)
|
569 |
-
logger.debug(f"
|
570 |
|
571 |
-
# Mezclar voz y música. Ajusta los volúmenes según prefieras.
|
572 |
final_audio = CompositeAudioClip([
|
573 |
-
musica_audio.volumex(0.2),
|
574 |
-
audio_tts.volumex(1.0)
|
575 |
])
|
576 |
-
logger.info("
|
577 |
except Exception as e:
|
578 |
-
logger.warning(f"Error
|
579 |
-
# Si falla la música, simplemente usamos el audio TTS
|
580 |
final_audio = audio_tts
|
581 |
-
logger.warning("
|
582 |
|
583 |
-
# Asegurar que el audio final tenga exactamente la misma duración que el video base
|
584 |
if final_audio.duration > video_base.duration:
|
585 |
final_audio = final_audio.subclip(0, video_base.duration)
|
586 |
-
|
587 |
-
|
588 |
-
|
589 |
-
|
590 |
-
# Para evitar errores en write_videofile, es mejor que el audio no sea más corto
|
591 |
-
# Si es más corto, moviepy suele extender el último frame de audio, lo cual es feo pero funciona.
|
592 |
-
# No intentaremos extenderlo aquí automáticamente para evitar complejidad, MoviePy lo manejará.
|
593 |
-
pass # MoviePy manejará la diferencia de duración
|
594 |
-
|
595 |
-
# 7. Crear video final
|
596 |
-
logger.info("Renderizando video final...")
|
597 |
video_final = video_base.set_audio(final_audio)
|
598 |
|
599 |
-
# Gradio maneja directorios temporales para los outputs.
|
600 |
-
# Podemos escribir directamente en un archivo temporal dentro de la carpeta temporal de Gradio.
|
601 |
-
# Si no estamos en un entorno Gradio, podríamos usar un tempfile normal.
|
602 |
-
# Por simplicidad y compatibilidad con Gradio, devolveremos el path absoluto.
|
603 |
-
# El directorio temporal creado por tempfile.mkdtemp() EXISTIRÁ cuando la función retorne
|
604 |
-
# porque hemos eliminado la línea que lo borraba. Gradio debería poder leerlo.
|
605 |
-
# Si queremos un archivo temporal más "formal" de Gradio, necesitaríamos que Gradio
|
606 |
-
# nos pasara un path temporal. La forma actual de retornar el path debería funcionar
|
607 |
-
# ahora que no borramos el directorio.
|
608 |
-
|
609 |
-
# Asegurar que el directorio para el output existe si no es temp_dir_intermediate
|
610 |
-
# output_filename = f"final_video_{datetime.now().strftime('%Y%m%d_%H%M%S')}.mp4"
|
611 |
-
# # Si queremos que el archivo no esté en temp_dir_intermediate, podríamos usar otro temp_dir
|
612 |
-
# # o un path fijo para prueba. Para Gradio, devolver el path en temp_dir_intermediate
|
613 |
-
# # FUNCIONARÁ si temp_dir_intermediate NO SE BORRA.
|
614 |
-
# output_path = os.path.join(temp_dir_intermediate, output_filename)
|
615 |
-
|
616 |
-
# Dejar el nombre y path original dentro del temp_dir_intermediate,
|
617 |
-
# pero ASEGURARSE de que el rmtree del finally está COMENTADO/ELIMINADO.
|
618 |
output_filename = "final_video.mp4"
|
619 |
output_path = os.path.join(temp_dir_intermediate, output_filename)
|
620 |
-
logger.info(f"
|
621 |
|
622 |
video_final.write_videofile(
|
623 |
output_path,
|
624 |
fps=24,
|
625 |
-
threads=4,
|
626 |
-
codec="libx264",
|
627 |
-
audio_codec="aac",
|
628 |
-
preset="medium",
|
629 |
-
#
|
630 |
-
logger='bar' # Muestra una barra de progreso y mensajes de FFmpeg
|
631 |
-
# logger='log' # Muestra mensajes detallados de FFmpeg
|
632 |
)
|
633 |
|
634 |
total_time = (datetime.now() - start_time).total_seconds()
|
635 |
-
logger.info(f"
|
636 |
|
637 |
-
#
|
638 |
-
|
639 |
-
|
640 |
-
|
641 |
-
|
|
|
|
|
|
|
642 |
|
643 |
-
return output_path # Retornar el path del archivo generado
|
644 |
|
|
|
|
|
645 |
except ValueError as ve:
|
646 |
-
logger.error(f"ERROR
|
647 |
-
|
648 |
-
# (Esto es mejor hacerlo en el finally, pero aquí es una capa extra)
|
649 |
-
try:
|
650 |
-
if 'video_base' in locals() and video_base is not None: video_base.close()
|
651 |
-
if 'audio_tts' in locals() and audio_tts is not None: audio_tts.close()
|
652 |
-
if 'musica_audio' in locals() and musica_audio is not None: musica_audio.close()
|
653 |
-
if 'video_final' in locals() and video_final is not None: video_final.close()
|
654 |
-
except: pass # Ignore errors during cleanup before re-raise
|
655 |
-
raise ve # Rellenar la excepción para que la atrape run_app
|
656 |
-
|
657 |
except Exception as e:
|
658 |
-
logger.critical(f"
|
659 |
-
|
660 |
-
try:
|
661 |
-
if 'video_base' in locals() and video_base is not None: video_base.close()
|
662 |
-
if 'audio_tts' in locals() and audio_tts is not None: audio_tts.close()
|
663 |
-
if 'musica_audio' in locals() and musica_audio is not None: musica_audio.close()
|
664 |
-
if 'video_final' in locals() and video_final is not None: video_final.close()
|
665 |
-
except: pass # Ignore errors during cleanup before re-raise
|
666 |
-
raise e # Rellenar la excepción
|
667 |
-
|
668 |
finally:
|
669 |
-
logger.info("
|
670 |
-
# Limpiar solo los archivos listados como temporales intermedios
|
671 |
if temp_dir_intermediate and os.path.exists(temp_dir_intermediate):
|
672 |
-
# Asegurarse de que el archivo de video final *no* sea borrado si está en esta lista por error
|
673 |
final_output_in_temp = os.path.join(temp_dir_intermediate, "final_video.mp4")
|
674 |
|
675 |
for path in temp_intermediate_files:
|
676 |
try:
|
677 |
-
if os.path.isfile(path) and path != final_output_in_temp:
|
678 |
-
logger.debug(f"
|
679 |
os.remove(path)
|
680 |
-
elif os.path.isdir(path): # Limpieza defensiva, aunque no deberíamos tener directorios aquí
|
681 |
-
logger.warning(f"Intentando eliminar directorio listado como archivo temporal: {path}")
|
682 |
-
shutil.rmtree(path, ignore_errors=True)
|
683 |
except Exception as e:
|
684 |
-
logger.warning(f"
|
685 |
|
686 |
-
#
|
687 |
-
#
|
688 |
-
|
689 |
-
# con el tiempo, o puedes implementar una limpieza más sofisticada si es necesario
|
690 |
-
# para archivos *muy* viejos, pero NO lo hagas inmediatamente después de crear el archivo.
|
691 |
-
# shutil.rmtree(temp_dir_intermediate, ignore_errors=True) # COMENTADO/ELIMINADO intencionadamente
|
692 |
-
logger.info(f"Directorio temporal intermedio {temp_dir_intermediate} persistirá para que Gradio lea el video final.")
|
693 |
|
694 |
|
695 |
-
# [FUNCIÓN RUN_APP ORIGINAL CON LOGS]
|
696 |
def run_app(prompt_type, prompt_ia, prompt_manual, musica_file):
|
697 |
logger.info("="*80)
|
698 |
-
logger.info("
|
699 |
|
700 |
input_text = prompt_ia if prompt_type == "Generar Guion con IA" else prompt_manual
|
701 |
|
702 |
if not input_text or not input_text.strip():
|
703 |
-
logger.warning("
|
704 |
-
# Asegurarse de devolver None para el video y un mensaje de error
|
705 |
return None, gr.update(value="⚠️ Por favor, ingresa texto para el guion o el tema.")
|
706 |
|
707 |
-
|
708 |
-
|
709 |
-
logger.info(f"Tipo de entrada: {prompt_type}")
|
710 |
-
logger.debug(f"Texto de entrada: '{input_text[:100]}...'")
|
711 |
if musica_file:
|
712 |
-
logger.info(f"
|
713 |
else:
|
714 |
-
logger.info("No
|
715 |
|
716 |
try:
|
717 |
-
logger.info("
|
718 |
video_path = crear_video(prompt_type, input_text, musica_file)
|
719 |
|
720 |
if video_path and os.path.exists(video_path):
|
721 |
-
logger.info(f"
|
722 |
-
logger.info(f"
|
723 |
return video_path, gr.update(value="✅ Video generado exitosamente.", interactive=False)
|
724 |
else:
|
725 |
-
logger.error(f"crear_video
|
726 |
return None, gr.update(value="❌ Error: La generación del video falló o el archivo no se creó correctamente.", interactive=False)
|
727 |
|
728 |
except ValueError as ve:
|
729 |
-
logger.warning(f"
|
730 |
return None, gr.update(value=f"⚠️ Error de validación: {str(ve)}", interactive=False)
|
731 |
except Exception as e:
|
732 |
-
logger.critical(f"
|
733 |
return None, gr.update(value=f"❌ Error inesperado: {str(e)}", interactive=False)
|
734 |
finally:
|
735 |
-
logger.info("
|
736 |
|
737 |
|
738 |
-
#
|
739 |
with gr.Blocks(title="Generador de Videos con IA", theme=gr.themes.Soft(), css="""
|
740 |
.gradio-container {max-width: 800px; margin: auto;}
|
741 |
h1 {text-align: center;}
|
742 |
""") as app:
|
743 |
|
744 |
-
gr.Markdown("# 🎬
|
745 |
-
gr.Markdown("
|
746 |
|
747 |
with gr.Row():
|
748 |
with gr.Column():
|
749 |
prompt_type = gr.Radio(
|
750 |
-
["Generar Guion con IA", "Usar Mi Guion"],
|
751 |
-
label="
|
752 |
-
value="Generar Guion con IA"
|
753 |
)
|
754 |
|
755 |
with gr.Column(visible=True) as ia_guion_column:
|
756 |
prompt_ia = gr.Textbox(
|
757 |
-
label="
|
758 |
lines=2,
|
759 |
-
placeholder="
|
760 |
max_lines=4,
|
761 |
-
value=""
|
762 |
)
|
763 |
|
764 |
with gr.Column(visible=False) as manual_guion_column:
|
765 |
prompt_manual = gr.Textbox(
|
766 |
-
label="
|
767 |
lines=5,
|
768 |
-
placeholder="
|
769 |
max_lines=10,
|
770 |
-
value=""
|
771 |
)
|
772 |
|
773 |
musica_input = gr.Audio(
|
774 |
-
label="
|
775 |
type="filepath",
|
776 |
interactive=True,
|
777 |
-
value=None
|
778 |
)
|
779 |
|
780 |
-
generate_btn = gr.Button("✨
|
781 |
|
782 |
with gr.Column():
|
783 |
video_output = gr.Video(
|
784 |
-
label="Video
|
785 |
-
interactive=False,
|
786 |
-
height=400
|
787 |
)
|
788 |
status_output = gr.Textbox(
|
789 |
-
label="
|
790 |
-
interactive=False,
|
791 |
-
show_label=False,
|
792 |
-
placeholder="
|
793 |
-
value="
|
794 |
)
|
795 |
|
796 |
-
# Lógica para mostrar/ocultar campos de texto según el tipo de prompt
|
797 |
prompt_type.change(
|
798 |
lambda x: (gr.update(visible=x == "Generar Guion con IA"),
|
799 |
gr.update(visible=x == "Usar Mi Guion")),
|
@@ -801,67 +669,41 @@ with gr.Blocks(title="Generador de Videos con IA", theme=gr.themes.Soft(), css="
|
|
801 |
outputs=[ia_guion_column, manual_guion_column]
|
802 |
)
|
803 |
|
804 |
-
# Lógica para el botón de generar video
|
805 |
-
# Primero, actualizar el estado a "Procesando..."
|
806 |
-
# Luego, llamar a la función run_app
|
807 |
generate_btn.click(
|
808 |
-
|
809 |
-
lambda: (None, gr.update(value="⏳ Procesando... Esto puede tomar 2-5 minutos dependiendo de la longitud y recursos.", interactive=False)),
|
810 |
outputs=[video_output, status_output],
|
811 |
-
queue=True,
|
812 |
-
# preprocess=False, # No pre-procesar para esta primera acción
|
813 |
).then(
|
814 |
-
# La segunda acción (asincrona) ejecuta la función principal de generación
|
815 |
run_app,
|
816 |
inputs=[prompt_type, prompt_ia, prompt_manual, musica_input],
|
817 |
outputs=[video_output, status_output]
|
818 |
)
|
819 |
|
820 |
-
gr.Markdown("###
|
821 |
gr.Markdown("""
|
822 |
-
1. **
|
823 |
-
2. **
|
824 |
-
- "
|
825 |
-
- "
|
826 |
-
3. **
|
827 |
-
4. **
|
828 |
-
5.
|
829 |
-
6.
|
830 |
""")
|
831 |
gr.Markdown("---")
|
832 |
-
gr.Markdown("
|
833 |
|
834 |
if __name__ == "__main__":
|
835 |
-
|
836 |
-
# Puedes añadir esta verificación aquí
|
837 |
-
logger.info("Verificando dependencias críticas...")
|
838 |
try:
|
839 |
-
|
840 |
-
|
841 |
-
# ffmpeg_path = get_setting("FFMPEG_BINARY")
|
842 |
-
# if not os.path.exists(ffmpeg_path):
|
843 |
-
# logger.critical(f"FFmpeg no encontrado en {ffmpeg_path}. Instálalo y/o configura la variable FFMPEG_BINARY.")
|
844 |
-
# # raise RuntimeError("FFmpeg no está instalado o configurado correctamente.")
|
845 |
-
# else:
|
846 |
-
# logger.info(f"FFmpeg encontrado en {ffmpeg_path}")
|
847 |
-
|
848 |
-
# Una forma más simple es intentar importar algo que use FFmpeg
|
849 |
-
try:
|
850 |
-
from moviepy.editor import VideoFileClip
|
851 |
-
# No cargamos un archivo real, solo verificamos que el import no falle por FFmpeg
|
852 |
-
logger.info("MoviePy importado correctamente. FFmpeg parece accesible.")
|
853 |
-
except Exception as e:
|
854 |
-
logger.critical(f"Fallo al importar MoviePy, a menudo indica problemas con FFmpeg. Asegúrate de tenerlo instalado y en el PATH. Error: {e}")
|
855 |
-
# raise RuntimeError("MoviePy/FFmpeg dependency issue.")
|
856 |
-
|
857 |
except Exception as e:
|
858 |
-
logger.
|
859 |
|
860 |
-
logger.info("
|
861 |
try:
|
862 |
-
# Usa share=True para obtener un enlace público si lo necesitas para probar fuera de localhost
|
863 |
-
# No uses share=True en producción sin medidas de seguridad adecuadas
|
864 |
app.launch(server_name="0.0.0.0", server_port=7860, share=False)
|
865 |
except Exception as e:
|
866 |
-
logger.critical(f"
|
867 |
-
raise
|
|
|
7 |
import edge_tts
|
8 |
import gradio as gr
|
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
|
13 |
import re
|
14 |
import math
|
|
|
|
|
|
|
|
|
15 |
import shutil
|
16 |
import json
|
17 |
+
from collections import Counter
|
18 |
|
19 |
+
# Logging configuration
|
20 |
logging.basicConfig(
|
21 |
level=logging.DEBUG,
|
22 |
format='%(asctime)s - %(levelname)s - %(message)s',
|
|
|
27 |
)
|
28 |
logger = logging.getLogger(__name__)
|
29 |
logger.info("="*80)
|
30 |
+
logger.info("STARTING VIDEO GENERATOR EXECUTION")
|
31 |
logger.info("="*80)
|
32 |
|
33 |
+
# Pexels API Key
|
|
|
34 |
PEXELS_API_KEY = os.environ.get("PEXELS_API_KEY")
|
35 |
if not PEXELS_API_KEY:
|
36 |
+
logger.critical("PEXELS_API_KEY environment variable not found.")
|
37 |
+
# logger.warning("Continuing without PEXELS_API_KEY. Video search will fail.")
|
38 |
+
# raise ValueError("Pexels API key not configured") # Uncomment to force fail if not set
|
39 |
+
|
40 |
+
# Model Initialization
|
|
|
|
|
|
|
41 |
MODEL_NAME = "datificate/gpt2-small-spanish"
|
42 |
+
logger.info(f"Initializing GPT-2 model: {MODEL_NAME}")
|
43 |
+
tokenizer = None
|
44 |
+
model = None
|
45 |
try:
|
|
|
46 |
tokenizer = GPT2Tokenizer.from_pretrained(MODEL_NAME)
|
47 |
model = GPT2LMHeadModel.from_pretrained(MODEL_NAME).eval()
|
48 |
if tokenizer.pad_token is None:
|
49 |
tokenizer.pad_token = tokenizer.eos_token
|
50 |
+
logger.info(f"GPT-2 model loaded | Vocab size: {len(tokenizer)}")
|
51 |
except Exception as e:
|
52 |
+
logger.error(f"CRITICAL FAILURE loading GPT-2: {str(e)}", exc_info=True)
|
53 |
+
tokenizer = model = None
|
54 |
|
55 |
+
logger.info("Loading KeyBERT model...")
|
56 |
+
kw_model = None
|
57 |
try:
|
58 |
kw_model = KeyBERT('distilbert-base-multilingual-cased')
|
59 |
+
logger.info("KeyBERT initialized successfully")
|
60 |
except Exception as e:
|
61 |
+
logger.error(f"FAILURE loading KeyBERT: {str(e)}", exc_info=True)
|
62 |
+
kw_model = None
|
63 |
|
|
|
64 |
def buscar_videos_pexels(query, api_key, per_page=5):
|
65 |
if not api_key:
|
66 |
+
logger.warning("Cannot search Pexels: API Key not configured.")
|
67 |
return []
|
68 |
|
69 |
+
logger.debug(f"Searching Pexels: '{query}' | Results per page: {per_page}")
|
70 |
headers = {"Authorization": api_key}
|
71 |
try:
|
72 |
params = {
|
|
|
76 |
"size": "medium"
|
77 |
}
|
78 |
|
|
|
79 |
response = requests.get(
|
80 |
"https://api.pexels.com/videos/search",
|
81 |
headers=headers,
|
82 |
params=params,
|
83 |
timeout=20
|
84 |
)
|
85 |
+
response.raise_for_status()
|
86 |
+
|
87 |
+
data = response.json()
|
88 |
+
videos = data.get('videos', [])
|
89 |
+
logger.info(f"Pexels: Found {len(videos)} videos for '{query}'")
|
90 |
+
return videos
|
|
|
|
|
|
|
|
|
91 |
|
92 |
except requests.exceptions.RequestException as e:
|
93 |
+
logger.error(f"Pexels connection error for '{query}': {str(e)}")
|
94 |
+
except json.JSONDecodeError:
|
95 |
+
logger.error(f"Pexels: Invalid JSON received | Status: {response.status_code} | Response: {response.text[:200]}...")
|
96 |
except Exception as e:
|
97 |
+
logger.error(f"Unexpected Pexels error for '{query}': {str(e)}", exc_info=True)
|
98 |
|
99 |
return []
|
100 |
|
|
|
101 |
def generate_script(prompt, max_length=150):
|
102 |
+
logger.info(f"Generating script | Prompt: '{prompt[:50]}...' | Max length: {max_length}")
|
103 |
if not tokenizer or not model:
|
104 |
+
logger.warning("GPT-2 models not available - Using original prompt as script.")
|
105 |
return prompt
|
106 |
|
107 |
try:
|
108 |
+
enhanced_prompt = f"Escribe un guion corto, interesante y coherente sobre: {prompt}"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
109 |
inputs = tokenizer(enhanced_prompt, return_tensors="pt", truncation=True, max_length=512)
|
110 |
|
|
|
|
|
111 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
112 |
model.to(device)
|
113 |
inputs = {k: v.to(device) for k, v in inputs.items()}
|
114 |
|
|
|
115 |
outputs = model.generate(
|
116 |
**inputs,
|
117 |
max_length=max_length,
|
118 |
+
do_sample=True,
|
119 |
top_p=0.9,
|
120 |
top_k=40,
|
121 |
temperature=0.7,
|
122 |
+
repetition_penalty=1.2,
|
123 |
pad_token_id=tokenizer.pad_token_id,
|
124 |
eos_token_id=tokenizer.eos_token_id,
|
125 |
+
no_repeat_ngram_size=3
|
126 |
)
|
127 |
|
128 |
text = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
129 |
|
130 |
+
text = re.sub(r'<[^>]+>', '', text)
|
|
|
131 |
text = text.strip()
|
132 |
|
|
|
133 |
sentences = text.split('.')
|
134 |
if sentences and sentences[0].strip():
|
135 |
final_text = sentences[0].strip() + '.'
|
136 |
+
if len(sentences) > 1 and sentences[1].strip() and len(final_text.split()) < max_length * 0.5:
|
|
|
137 |
final_text += " " + sentences[1].strip() + "."
|
138 |
+
final_text = final_text.replace("..", ".")
|
139 |
|
140 |
+
logger.info(f"Generated script (Truncated): '{final_text[:100]}...'")
|
141 |
+
return final_text.strip()
|
142 |
|
143 |
+
logger.info(f"Generated script (no full sentences detected): '{text[:100]}...'")
|
144 |
+
return text.strip()
|
145 |
+
|
146 |
except Exception as e:
|
147 |
+
logger.error(f"Error generating script with GPT-2: {str(e)}", exc_info=True)
|
148 |
+
logger.warning("Using original prompt as script due to generation error.")
|
149 |
+
return prompt.strip()
|
150 |
|
|
|
151 |
async def text_to_speech(text, output_path, voice="es-ES-ElviraNeural"):
|
152 |
+
logger.info(f"Converting text to speech | Chars: {len(text)} | Voice: {voice} | Output: {output_path}")
|
153 |
if not text or not text.strip():
|
154 |
+
logger.warning("Empty text for TTS")
|
155 |
return False
|
156 |
|
157 |
try:
|
158 |
communicate = edge_tts.Communicate(text, voice)
|
159 |
await communicate.save(output_path)
|
160 |
|
161 |
+
if os.path.exists(output_path) and os.path.getsize(output_path) > 100:
|
162 |
+
logger.info(f"Audio saved successfully to: {output_path} | Size: {os.path.getsize(output_path)} bytes")
|
163 |
return True
|
164 |
else:
|
165 |
+
logger.error(f"TTS saved small or empty file to: {output_path}")
|
166 |
return False
|
167 |
|
168 |
except Exception as e:
|
169 |
+
logger.error(f"Error in TTS: {str(e)}", exc_info=True)
|
170 |
return False
|
171 |
|
|
|
172 |
def download_video_file(url, temp_dir):
|
173 |
if not url:
|
174 |
+
logger.warning("Video URL not provided for download")
|
175 |
return None
|
176 |
|
177 |
try:
|
178 |
+
logger.info(f"Downloading video from: {url[:80]}...")
|
|
|
179 |
os.makedirs(temp_dir, exist_ok=True)
|
180 |
file_name = f"video_dl_{datetime.now().strftime('%Y%m%d_%H%M%S_%f')}.mp4"
|
181 |
output_path = os.path.join(temp_dir, file_name)
|
182 |
|
|
|
183 |
with requests.get(url, stream=True, timeout=60) as r:
|
184 |
r.raise_for_status()
|
185 |
+
# total_size = int(r.headers.get('content-length', 0)) # Uncomment for progress logging
|
186 |
+
# downloaded_size = 0
|
|
|
|
|
187 |
with open(output_path, 'wb') as f:
|
188 |
for chunk in r.iter_content(chunk_size=8192):
|
189 |
f.write(chunk)
|
190 |
+
# downloaded_size += len(chunk) # Uncomment for progress logging
|
|
|
191 |
|
192 |
+
if os.path.exists(output_path) and os.path.getsize(output_path) > 1000:
|
193 |
+
logger.info(f"Video downloaded successfully: {output_path} | Size: {os.path.getsize(output_path)} bytes")
|
|
|
194 |
return output_path
|
195 |
else:
|
196 |
+
logger.warning(f"Download seems incomplete or empty for {url[:80]}... File: {output_path} Size: {os.path.getsize(output_path) if os.path.exists(output_path) else 'N/A'} bytes")
|
197 |
if os.path.exists(output_path):
|
198 |
+
os.remove(output_path)
|
199 |
return None
|
200 |
|
201 |
except requests.exceptions.RequestException as e:
|
202 |
+
logger.error(f"Download error for {url[:80]}... : {str(e)}")
|
203 |
except Exception as e:
|
204 |
+
logger.error(f"Unexpected error downloading {url[:80]}... : {str(e)}", exc_info=True)
|
205 |
|
206 |
return None
|
207 |
|
|
|
208 |
def loop_audio_to_length(audio_clip, target_duration):
|
209 |
+
logger.debug(f"Adjusting audio | Current duration: {audio_clip.duration:.2f}s | Target: {target_duration:.2f}s")
|
210 |
if audio_clip.duration <= 0:
|
211 |
+
logger.warning("Audio clip has zero or negative duration, cannot loop.")
|
212 |
+
return AudioFileClip(filename="")
|
213 |
|
214 |
if audio_clip.duration >= target_duration:
|
215 |
+
logger.debug("Audio clip already longer or equal to target.")
|
216 |
return audio_clip.subclip(0, target_duration)
|
217 |
|
|
|
218 |
loops = math.ceil(target_duration / audio_clip.duration)
|
219 |
+
logger.debug(f"Creating {loops} audio loops")
|
220 |
|
|
|
221 |
audio_segments = [audio_clip] * loops
|
|
|
|
|
222 |
looped_audio = concatenate_audioclips(audio_segments)
|
223 |
|
|
|
224 |
return looped_audio.subclip(0, target_duration)
|
225 |
|
|
|
226 |
def extract_visual_keywords_from_script(script_text):
|
227 |
+
logger.info("Extracting keywords from script")
|
228 |
if not script_text or not script_text.strip():
|
229 |
+
logger.warning("Empty script, cannot extract keywords.")
|
230 |
+
return ["naturaleza", "ciudad", "paisaje"]
|
231 |
|
|
|
232 |
clean_text = re.sub(r'[^\w\sáéíóúñÁÉÍÓÚÑ]', '', script_text)
|
|
|
233 |
keywords_list = []
|
234 |
|
235 |
if kw_model:
|
236 |
try:
|
237 |
+
logger.debug("Attempting KeyBERT extraction...")
|
238 |
+
keywords1 = kw_model.extract_keywords(clean_text, keyphrase_ngram_range=(1, 1), stop_words='spanish', top_n=5)
|
239 |
+
keywords2 = kw_model.extract_keywords(clean_text, keyphrase_ngram_range=(2, 2), stop_words='spanish', top_n=3)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
240 |
|
|
|
241 |
all_keywords = keywords1 + keywords2
|
|
|
242 |
all_keywords.sort(key=lambda item: item[1], reverse=True)
|
243 |
|
|
|
244 |
seen_keywords = set()
|
245 |
for keyword, score in all_keywords:
|
|
|
246 |
formatted_keyword = keyword.lower().replace(" ", "+")
|
247 |
+
if formatted_keyword and formatted_keyword not in seen_keywords: # Ensure keyword is not empty
|
248 |
keywords_list.append(formatted_keyword)
|
249 |
seen_keywords.add(formatted_keyword)
|
250 |
+
if len(keywords_list) >= 5:
|
251 |
break
|
252 |
|
253 |
if keywords_list:
|
254 |
+
logger.debug(f"KeyBERT extracted keywords: {keywords_list}")
|
255 |
return keywords_list
|
256 |
|
257 |
except Exception as e:
|
258 |
+
logger.warning(f"KeyBERT failed: {str(e)}. Trying simple method.")
|
|
|
259 |
|
260 |
+
logger.debug("Extracting keywords with simple method...")
|
|
|
261 |
words = clean_text.lower().split()
|
262 |
+
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"}
|
|
|
263 |
|
|
|
264 |
valid_words = [word for word in words if len(word) > 3 and word not in stop_words]
|
265 |
|
266 |
if not valid_words:
|
267 |
+
logger.warning("No valid keywords found with simple method. Using default keywords.")
|
268 |
return ["naturaleza", "ciudad", "paisaje"]
|
269 |
|
|
|
270 |
word_counts = Counter(valid_words)
|
|
|
271 |
top_keywords = [word.replace(" ", "+") for word, _ in word_counts.most_common(5)]
|
272 |
|
273 |
if not top_keywords:
|
274 |
+
logger.warning("Simple method produced no keywords. Using default keywords.")
|
275 |
return ["naturaleza", "ciudad", "paisaje"]
|
276 |
|
277 |
+
logger.info(f"Final keywords: {top_keywords}")
|
278 |
return top_keywords
|
279 |
|
|
|
280 |
def crear_video(prompt_type, input_text, musica_file=None):
|
281 |
logger.info("="*80)
|
282 |
+
logger.info(f"STARTING VIDEO CREATION | Type: {prompt_type}")
|
283 |
logger.debug(f"Input: '{input_text[:100]}...'")
|
284 |
|
285 |
start_time = datetime.now()
|
286 |
+
temp_dir_intermediate = None
|
287 |
+
|
|
|
288 |
try:
|
289 |
+
# 1. Generate or use script
|
290 |
if prompt_type == "Generar Guion con IA":
|
|
|
291 |
guion = generate_script(input_text)
|
292 |
else:
|
293 |
+
guion = input_text.strip()
|
|
|
294 |
|
295 |
+
logger.info(f"Final script ({len(guion)} chars): '{guion[:100]}...'")
|
296 |
|
297 |
if not guion.strip():
|
298 |
+
logger.error("Resulting script is empty.")
|
299 |
+
raise ValueError("The script is empty.")
|
300 |
|
|
|
301 |
temp_dir_intermediate = tempfile.mkdtemp(prefix="video_gen_intermediate_")
|
302 |
+
logger.info(f"Intermediate temporary directory created: {temp_dir_intermediate}")
|
303 |
+
temp_intermediate_files = []
|
304 |
|
305 |
+
# 2. Generate voice audio
|
306 |
+
logger.info("Generating voice audio...")
|
307 |
voz_path = os.path.join(temp_dir_intermediate, "voz.mp3")
|
|
|
|
|
308 |
if not asyncio.run(text_to_speech(guion, voz_path)):
|
309 |
+
logger.error("Failed to generate voice audio.")
|
310 |
+
raise ValueError("Error generating voice audio.")
|
311 |
+
temp_intermediate_files.append(voz_path)
|
312 |
|
313 |
audio_tts = AudioFileClip(voz_path)
|
314 |
audio_duration = audio_tts.duration
|
315 |
+
logger.info(f"Voice audio duration: {audio_duration:.2f} seconds")
|
316 |
|
317 |
if audio_duration < 1.0:
|
318 |
+
logger.warning(f"Voice audio duration ({audio_duration:.2f}s) is very short.")
|
|
|
319 |
|
320 |
+
# 3. Extract keywords
|
321 |
+
logger.info("Extracting keywords...")
|
322 |
try:
|
323 |
keywords = extract_visual_keywords_from_script(guion)
|
324 |
+
logger.info(f"Identified keywords: {keywords}")
|
325 |
except Exception as e:
|
326 |
+
logger.error(f"Error extracting keywords: {str(e)}", exc_info=True)
|
327 |
+
keywords = ["naturaleza", "paisaje"]
|
328 |
|
329 |
if not keywords:
|
330 |
+
keywords = ["video", "background"]
|
|
|
331 |
|
332 |
+
# 4. Search and download videos
|
333 |
+
logger.info("Searching videos on Pexels...")
|
334 |
videos_data = []
|
335 |
+
total_desired_videos = 10
|
336 |
+
per_page_per_keyword = max(1, total_desired_videos // len(keywords))
|
|
|
337 |
|
338 |
for keyword in keywords:
|
339 |
+
if len(videos_data) >= total_desired_videos: break
|
340 |
try:
|
341 |
videos = buscar_videos_pexels(keyword, PEXELS_API_KEY, per_page=per_page_per_keyword)
|
342 |
if videos:
|
343 |
videos_data.extend(videos)
|
344 |
+
logger.info(f"Found {len(videos)} videos for '{keyword}'. Total data: {len(videos_data)}")
|
345 |
except Exception as e:
|
346 |
+
logger.warning(f"Error searching videos for '{keyword}': {str(e)}")
|
347 |
|
348 |
+
if len(videos_data) < total_desired_videos / 2:
|
349 |
+
logger.warning(f"Few videos found ({len(videos_data)}). Trying generic keywords.")
|
350 |
+
generic_keywords = ["nature", "city", "background", "abstract"]
|
|
|
351 |
for keyword in generic_keywords:
|
352 |
if len(videos_data) >= total_desired_videos: break
|
353 |
try:
|
354 |
+
videos = buscar_videos_pexels(keyword, PEXELS_API_KEY, per_page=2)
|
|
|
355 |
if videos:
|
356 |
videos_data.extend(videos)
|
357 |
+
logger.info(f"Found {len(videos)} videos for '{keyword}' (generic). Total data: {len(videos_data)}")
|
358 |
except Exception as e:
|
359 |
+
logger.warning(f"Error searching generic videos for '{keyword}': {str(e)}")
|
360 |
|
361 |
|
362 |
if not videos_data:
|
363 |
+
logger.error("No videos found on Pexels for any keyword.")
|
364 |
+
raise ValueError("No suitable videos found on Pexels.")
|
365 |
|
366 |
video_paths = []
|
367 |
+
logger.info(f"Attempting to download {len(videos_data)} found videos...")
|
368 |
for video in videos_data:
|
369 |
if 'video_files' not in video or not video['video_files']:
|
370 |
+
logger.debug(f"Skipping video without video files: {video.get('id')}")
|
371 |
continue
|
372 |
|
373 |
try:
|
|
|
374 |
best_quality = None
|
375 |
for vf in sorted(video['video_files'], key=lambda x: x.get('width', 0) * x.get('height', 0), reverse=True):
|
376 |
if 'link' in vf:
|
|
|
381 |
path = download_video_file(best_quality['link'], temp_dir_intermediate)
|
382 |
if path:
|
383 |
video_paths.append(path)
|
384 |
+
temp_intermediate_files.append(path)
|
385 |
+
logger.info(f"Video downloaded OK from {best_quality['link'][:50]}...")
|
386 |
else:
|
387 |
+
logger.warning(f"Could not download video from {best_quality['link'][:50]}...")
|
388 |
else:
|
389 |
+
logger.warning(f"No valid download link found for video {video.get('id')}.")
|
390 |
|
391 |
except Exception as e:
|
392 |
+
logger.warning(f"Error processing/downloading video {video.get('id')}: {str(e)}")
|
393 |
|
394 |
+
logger.info(f"Downloaded {len(video_paths)} usable video files.")
|
395 |
if not video_paths:
|
396 |
+
logger.error("Could not download any usable video file.")
|
397 |
+
raise ValueError("Could not download any usable video from Pexels.")
|
398 |
|
399 |
+
# 5. Process and concatenate video clips
|
400 |
+
logger.info("Processing and concatenating downloaded videos...")
|
401 |
clips = []
|
402 |
current_duration = 0
|
403 |
+
min_clip_duration = 1.0
|
404 |
+
max_clip_segment = 8.0
|
|
|
|
|
|
|
405 |
|
406 |
+
for i, path in enumerate(video_paths):
|
407 |
+
if current_duration >= audio_duration + max_clip_segment:
|
408 |
+
logger.debug(f"Video base sufficient ({current_duration:.1f}s >= {audio_duration:.1f}s + {max_clip_segment:.1f}s buffer). Stopping processing remaining source clips.")
|
|
|
409 |
break
|
410 |
|
411 |
+
clip = None
|
412 |
try:
|
413 |
+
logger.debug(f"[{i+1}/{len(video_paths)}] Opening clip: {path}")
|
414 |
clip = VideoFileClip(path)
|
415 |
+
|
416 |
+
# Check clip validity after opening
|
417 |
+
if clip.reader is None or clip.duration is None or clip.duration <= 0:
|
418 |
+
logger.warning(f"[{i+1}/{len(video_paths)}] Clip {path} seems invalid (reader is None or duration <= 0). Skipping.")
|
419 |
+
continue
|
420 |
+
|
421 |
+
# Calculate how much to take from this clip
|
422 |
remaining_needed = audio_duration - current_duration
|
423 |
+
potential_use_duration = min(clip.duration, max_clip_segment)
|
424 |
+
|
|
|
425 |
if remaining_needed > 0:
|
426 |
+
segment_duration = min(potential_use_duration, remaining_needed + min_clip_duration)
|
427 |
+
segment_duration = max(min_clip_duration, segment_duration)
|
428 |
+
segment_duration = min(segment_duration, clip.duration)
|
429 |
+
|
430 |
+
if segment_duration >= min_clip_duration:
|
431 |
+
try:
|
432 |
+
sub = clip.subclip(0, segment_duration)
|
433 |
+
clips.append(sub)
|
434 |
+
current_duration += sub.duration
|
435 |
+
logger.debug(f"[{i+1}/{len(video_paths)}] Segment added: {sub.duration:.1f}s (total video: {current_duration:.1f}/{audio_duration:.1f}s)")
|
436 |
+
except Exception as sub_e:
|
437 |
+
logger.warning(f"[{i+1}/{len(video_paths)}] Error creating subclip from {path} ({segment_duration:.1f}s): {str(sub_e)}")
|
438 |
+
continue
|
439 |
+
else:
|
440 |
+
logger.debug(f"[{i+1}/{len(video_paths)}] Clip {path} ({clip.duration:.1f}s) doesn't contribute sufficient segment ({segment_duration:.1f}s needed from it). Skipping.")
|
441 |
else:
|
442 |
+
logger.debug(f"[{i+1}/{len(video_paths)}] Video base duration already reached. Skipping clip.")
|
|
|
|
|
|
|
443 |
|
444 |
except Exception as e:
|
445 |
+
logger.warning(f"[{i+1}/{len(video_paths)}] Error processing video {path}: {str(e)}", exc_info=True)
|
|
|
446 |
continue
|
447 |
|
448 |
+
finally:
|
449 |
+
if clip is not None:
|
450 |
+
try:
|
451 |
+
clip.close()
|
452 |
+
logger.debug(f"[{i+1}/{len(video_paths)}] Clip {path} closed.")
|
453 |
+
except Exception as close_e:
|
454 |
+
logger.warning(f"[{i+1}/{len(video_paths)}] Error closing clip {path}: {str(close_e)}")
|
455 |
+
|
456 |
+
|
457 |
+
logger.info(f"Source clip processing finished. Obtained {len(clips)} valid clips.")
|
458 |
+
|
459 |
if not clips:
|
460 |
+
logger.error("No valid video clips available to create the sequence.")
|
461 |
+
raise ValueError("No valid video clips available to create the video.")
|
462 |
|
463 |
+
logger.info(f"Concatenating {len(clips)} video clips.")
|
464 |
video_base = concatenate_videoclips(clips, method="compose")
|
465 |
+
logger.info(f"Base video duration: {video_base.duration:.2f}s")
|
466 |
|
|
|
467 |
if video_base.duration < audio_duration:
|
|
|
|
|
468 |
num_repeats = math.ceil(audio_duration / video_base.duration)
|
469 |
+
logger.info(f"Repeating base video ({video_base.duration:.2f}s) {num_repeats} times to reach {audio_duration:.2f}s.")
|
470 |
repeated_clips = [video_base] * num_repeats
|
471 |
video_base = concatenate_videoclips(repeated_clips, method="compose").subclip(0, audio_duration)
|
472 |
+
logger.info(f"Adjusted base video duration: {video_base.duration:.2f}s")
|
473 |
|
|
|
|
|
474 |
if video_base.duration > audio_duration:
|
475 |
+
logger.info(f"Trimming base video ({video_base.duration:.2f}s) to match audio duration ({audio_duration:.2f}s).")
|
476 |
video_base = video_base.subclip(0, audio_duration)
|
477 |
+
logger.info(f"Final base video duration: {video_base.duration:.2f}s")
|
478 |
|
479 |
|
480 |
+
# 6. Handle background music
|
481 |
+
logger.info("Processing audio...")
|
482 |
+
final_audio = audio_tts
|
483 |
|
484 |
if musica_file:
|
485 |
try:
|
|
|
486 |
music_path = os.path.join(temp_dir_intermediate, "musica_bg.mp3")
|
487 |
shutil.copyfile(musica_file, music_path)
|
488 |
+
temp_intermediate_files.append(music_path)
|
489 |
+
logger.info(f"Background music copied to: {music_path}")
|
490 |
|
491 |
musica_audio = AudioFileClip(music_path)
|
492 |
+
logger.debug(f"Original music duration: {musica_audio.duration:.2f}s")
|
493 |
|
|
|
494 |
musica_audio = loop_audio_to_length(musica_audio, video_base.duration)
|
495 |
+
logger.debug(f"Music adjusted to video duration: {musica_audio.duration:.2f}s")
|
496 |
|
|
|
497 |
final_audio = CompositeAudioClip([
|
498 |
+
musica_audio.volumex(0.2),
|
499 |
+
audio_tts.volumex(1.0)
|
500 |
])
|
501 |
+
logger.info("Audio mix completed (voice + music).")
|
502 |
except Exception as e:
|
503 |
+
logger.warning(f"Error processing background music: {str(e)}", exc_info=True)
|
|
|
504 |
final_audio = audio_tts
|
505 |
+
logger.warning("Using voice audio only due to music processing error.")
|
506 |
|
|
|
507 |
if final_audio.duration > video_base.duration:
|
508 |
final_audio = final_audio.subclip(0, video_base.duration)
|
509 |
+
|
510 |
+
|
511 |
+
# 7. Create final video
|
512 |
+
logger.info("Rendering final video...")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
513 |
video_final = video_base.set_audio(final_audio)
|
514 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
515 |
output_filename = "final_video.mp4"
|
516 |
output_path = os.path.join(temp_dir_intermediate, output_filename)
|
517 |
+
logger.info(f"Writing final video to: {output_path}")
|
518 |
|
519 |
video_final.write_videofile(
|
520 |
output_path,
|
521 |
fps=24,
|
522 |
+
threads=4,
|
523 |
+
codec="libx264",
|
524 |
+
audio_codec="aac",
|
525 |
+
preset="medium",
|
526 |
+
logger='bar' # Show MoviePy progress bar
|
|
|
|
|
527 |
)
|
528 |
|
529 |
total_time = (datetime.now() - start_time).total_seconds()
|
530 |
+
logger.info(f"VIDEO PROCESS FINISHED | Output: {output_path} | Total time: {total_time:.2f}s")
|
531 |
|
532 |
+
# Close main clips
|
533 |
+
try:
|
534 |
+
video_base.close()
|
535 |
+
audio_tts.close()
|
536 |
+
if 'musica_audio' in locals() and musica_audio is not None: musica_audio.close()
|
537 |
+
video_final.close()
|
538 |
+
except Exception as e:
|
539 |
+
logger.warning(f"Error closing final clips: {str(e)}")
|
540 |
|
|
|
541 |
|
542 |
+
return output_path
|
543 |
+
|
544 |
except ValueError as ve:
|
545 |
+
logger.error(f"CONTROLLED ERROR in crear_video: {str(ve)}")
|
546 |
+
raise ve
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
547 |
except Exception as e:
|
548 |
+
logger.critical(f"CRITICAL UNHANDLED ERROR in crear_video: {str(e)}", exc_info=True)
|
549 |
+
raise e
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
550 |
finally:
|
551 |
+
logger.info("Starting cleanup of intermediate temporary files...")
|
|
|
552 |
if temp_dir_intermediate and os.path.exists(temp_dir_intermediate):
|
|
|
553 |
final_output_in_temp = os.path.join(temp_dir_intermediate, "final_video.mp4")
|
554 |
|
555 |
for path in temp_intermediate_files:
|
556 |
try:
|
557 |
+
if os.path.isfile(path) and path != final_output_in_temp:
|
558 |
+
logger.debug(f"Deleting temporary file: {path}")
|
559 |
os.remove(path)
|
|
|
|
|
|
|
560 |
except Exception as e:
|
561 |
+
logger.warning(f"Could not delete temporary file {path}: {str(e)}")
|
562 |
|
563 |
+
# IMPORTANT: DO NOT remove the temp_dir_intermediate itself.
|
564 |
+
# It contains the final video file needed by Gradio.
|
565 |
+
logger.info(f"Intermediate temporary directory {temp_dir_intermediate} will persist for Gradio to read the final video.")
|
|
|
|
|
|
|
|
|
566 |
|
567 |
|
|
|
568 |
def run_app(prompt_type, prompt_ia, prompt_manual, musica_file):
|
569 |
logger.info("="*80)
|
570 |
+
logger.info("REQUEST RECEIVED IN INTERFACE")
|
571 |
|
572 |
input_text = prompt_ia if prompt_type == "Generar Guion con IA" else prompt_manual
|
573 |
|
574 |
if not input_text or not input_text.strip():
|
575 |
+
logger.warning("Empty input text.")
|
|
|
576 |
return None, gr.update(value="⚠️ Por favor, ingresa texto para el guion o el tema.")
|
577 |
|
578 |
+
logger.info(f"Input Type: {prompt_type}")
|
579 |
+
logger.debug(f"Input Text: '{input_text[:100]}...'")
|
|
|
|
|
580 |
if musica_file:
|
581 |
+
logger.info(f"Music file received: {musica_file}")
|
582 |
else:
|
583 |
+
logger.info("No music file provided.")
|
584 |
|
585 |
try:
|
586 |
+
logger.info("Calling crear_video...")
|
587 |
video_path = crear_video(prompt_type, input_text, musica_file)
|
588 |
|
589 |
if video_path and os.path.exists(video_path):
|
590 |
+
logger.info(f"crear_video returned path: {video_path}")
|
591 |
+
logger.info(f"Size of returned video file: {os.path.getsize(video_path)} bytes")
|
592 |
return video_path, gr.update(value="✅ Video generado exitosamente.", interactive=False)
|
593 |
else:
|
594 |
+
logger.error(f"crear_video did not return a valid path or file does not exist: {video_path}")
|
595 |
return None, gr.update(value="❌ Error: La generación del video falló o el archivo no se creó correctamente.", interactive=False)
|
596 |
|
597 |
except ValueError as ve:
|
598 |
+
logger.warning(f"Validation error during video creation: {str(ve)}")
|
599 |
return None, gr.update(value=f"⚠️ Error de validación: {str(ve)}", interactive=False)
|
600 |
except Exception as e:
|
601 |
+
logger.critical(f"Critical error during video creation: {str(e)}", exc_info=True)
|
602 |
return None, gr.update(value=f"❌ Error inesperado: {str(e)}", interactive=False)
|
603 |
finally:
|
604 |
+
logger.info("End of run_app handler.")
|
605 |
|
606 |
|
607 |
+
# Gradio Interface
|
608 |
with gr.Blocks(title="Generador de Videos con IA", theme=gr.themes.Soft(), css="""
|
609 |
.gradio-container {max-width: 800px; margin: auto;}
|
610 |
h1 {text-align: center;}
|
611 |
""") as app:
|
612 |
|
613 |
+
gr.Markdown("# 🎬 Automatic AI Video Generator")
|
614 |
+
gr.Markdown("Generate short videos from a topic or script, using stock footage from Pexels and generated voice.")
|
615 |
|
616 |
with gr.Row():
|
617 |
with gr.Column():
|
618 |
prompt_type = gr.Radio(
|
619 |
+
["Generar Guion con IA", "Usar Mi Guion"],
|
620 |
+
label="Input Method",
|
621 |
+
value="Generar Guion con IA"
|
622 |
)
|
623 |
|
624 |
with gr.Column(visible=True) as ia_guion_column:
|
625 |
prompt_ia = gr.Textbox(
|
626 |
+
label="Topic for AI",
|
627 |
lines=2,
|
628 |
+
placeholder="Ex: A natural landscape with mountains and rivers at sunrise, showing the beauty of nature...",
|
629 |
max_lines=4,
|
630 |
+
value=""
|
631 |
)
|
632 |
|
633 |
with gr.Column(visible=False) as manual_guion_column:
|
634 |
prompt_manual = gr.Textbox(
|
635 |
+
label="Your Full Script",
|
636 |
lines=5,
|
637 |
+
placeholder="Ex: In this video, we will explore the mysteries of the ocean. We will see fascinating marine life and vibrant coral reefs. Join us on this underwater adventure!",
|
638 |
max_lines=10,
|
639 |
+
value=""
|
640 |
)
|
641 |
|
642 |
musica_input = gr.Audio(
|
643 |
+
label="Background Music (optional)",
|
644 |
type="filepath",
|
645 |
interactive=True,
|
646 |
+
value=None
|
647 |
)
|
648 |
|
649 |
+
generate_btn = gr.Button("✨ Generate Video", variant="primary")
|
650 |
|
651 |
with gr.Column():
|
652 |
video_output = gr.Video(
|
653 |
+
label="Generated Video",
|
654 |
+
interactive=False,
|
655 |
+
height=400
|
656 |
)
|
657 |
status_output = gr.Textbox(
|
658 |
+
label="Status",
|
659 |
+
interactive=False,
|
660 |
+
show_label=False,
|
661 |
+
placeholder="Waiting for action...",
|
662 |
+
value="Waiting for input..."
|
663 |
)
|
664 |
|
|
|
665 |
prompt_type.change(
|
666 |
lambda x: (gr.update(visible=x == "Generar Guion con IA"),
|
667 |
gr.update(visible=x == "Usar Mi Guion")),
|
|
|
669 |
outputs=[ia_guion_column, manual_guion_column]
|
670 |
)
|
671 |
|
|
|
|
|
|
|
672 |
generate_btn.click(
|
673 |
+
lambda: (None, gr.update(value="⏳ Processing... This can take 2-5 minutes depending on length and resources.", interactive=False)),
|
|
|
674 |
outputs=[video_output, status_output],
|
675 |
+
queue=True,
|
|
|
676 |
).then(
|
|
|
677 |
run_app,
|
678 |
inputs=[prompt_type, prompt_ia, prompt_manual, musica_input],
|
679 |
outputs=[video_output, status_output]
|
680 |
)
|
681 |
|
682 |
+
gr.Markdown("### Instructions:")
|
683 |
gr.Markdown("""
|
684 |
+
1. **Pexels API Key:** Ensure you have set the `PEXELS_API_KEY` environment variable.
|
685 |
+
2. **Select Input Method**:
|
686 |
+
- "Generate Script with AI": Describe a topic (e.g., "The beauty of mountains"). AI will generate a short script.
|
687 |
+
- "Use My Script": Write the full script for your video.
|
688 |
+
3. **Upload Music** (optional): Select an audio file (MP3, WAV, etc.) for background music.
|
689 |
+
4. **Click "✨ Generate Video"**.
|
690 |
+
5. Wait for the video to process. Processing time may vary. Check the status box.
|
691 |
+
6. If there are errors, check the `video_generator_full.log` file for details.
|
692 |
""")
|
693 |
gr.Markdown("---")
|
694 |
+
gr.Markdown("Developed by [Your Name/Company/Alias - Optional]")
|
695 |
|
696 |
if __name__ == "__main__":
|
697 |
+
logger.info("Verifying critical dependencies...")
|
|
|
|
|
698 |
try:
|
699 |
+
from moviepy.editor import VideoFileClip
|
700 |
+
logger.info("MoviePy imported correctly. FFmpeg seems accessible.")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
701 |
except Exception as e:
|
702 |
+
logger.critical(f"Failed to import MoviePy, often indicates FFmpeg issues. Ensure it is installed and in PATH. Error: {e}")
|
703 |
|
704 |
+
logger.info("Starting Gradio app...")
|
705 |
try:
|
|
|
|
|
706 |
app.launch(server_name="0.0.0.0", server_port=7860, share=False)
|
707 |
except Exception as e:
|
708 |
+
logger.critical(f"Could not launch app: {str(e)}", exc_info=True)
|
709 |
+
raise
|