Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -9,14 +9,14 @@ 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, AudioClip
|
13 |
import re
|
14 |
import math
|
15 |
import shutil
|
16 |
import json
|
17 |
from collections import Counter
|
18 |
|
19 |
-
#
|
20 |
logging.basicConfig(
|
21 |
level=logging.DEBUG,
|
22 |
format='%(asctime)s - %(levelname)s - %(message)s',
|
@@ -27,19 +27,18 @@ logging.basicConfig(
|
|
27 |
)
|
28 |
logger = logging.getLogger(__name__)
|
29 |
logger.info("="*80)
|
30 |
-
logger.info("
|
31 |
logger.info("="*80)
|
32 |
|
33 |
-
#
|
34 |
PEXELS_API_KEY = os.environ.get("PEXELS_API_KEY")
|
35 |
if not PEXELS_API_KEY:
|
36 |
-
logger.critical("PEXELS_API_KEY
|
37 |
-
#
|
38 |
-
# raise ValueError("Pexels API key not configured")
|
39 |
|
40 |
-
#
|
41 |
MODEL_NAME = "datificate/gpt2-small-spanish"
|
42 |
-
logger.info(f"
|
43 |
tokenizer = None
|
44 |
model = None
|
45 |
try:
|
@@ -47,26 +46,26 @@ try:
|
|
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
|
51 |
except Exception as e:
|
52 |
-
logger.error(f"
|
53 |
tokenizer = model = None
|
54 |
|
55 |
-
logger.info("
|
56 |
kw_model = None
|
57 |
try:
|
58 |
kw_model = KeyBERT('distilbert-base-multilingual-cased')
|
59 |
-
logger.info("KeyBERT
|
60 |
except Exception as e:
|
61 |
-
logger.error(f"
|
62 |
kw_model = None
|
63 |
|
64 |
def buscar_videos_pexels(query, api_key, per_page=5):
|
65 |
if not api_key:
|
66 |
-
logger.warning("
|
67 |
return []
|
68 |
|
69 |
-
logger.debug(f"
|
70 |
headers = {"Authorization": api_key}
|
71 |
try:
|
72 |
params = {
|
@@ -75,7 +74,7 @@ def buscar_videos_pexels(query, api_key, per_page=5):
|
|
75 |
"orientation": "landscape",
|
76 |
"size": "medium"
|
77 |
}
|
78 |
-
|
79 |
response = requests.get(
|
80 |
"https://api.pexels.com/videos/search",
|
81 |
headers=headers,
|
@@ -83,38 +82,43 @@ def buscar_videos_pexels(query, api_key, per_page=5):
|
|
83 |
timeout=20
|
84 |
)
|
85 |
response.raise_for_status()
|
86 |
-
|
87 |
data = response.json()
|
88 |
videos = data.get('videos', [])
|
89 |
-
logger.info(f"Pexels:
|
90 |
return videos
|
91 |
-
|
92 |
except requests.exceptions.RequestException as e:
|
93 |
-
logger.error(f"
|
94 |
except json.JSONDecodeError:
|
95 |
-
logger.error(f"Pexels:
|
96 |
except Exception as e:
|
97 |
-
logger.error(f"
|
98 |
-
|
99 |
return []
|
100 |
|
101 |
def generate_script(prompt, max_length=150):
|
102 |
-
logger.info(f"Generando guión | Prompt: '{prompt[:50]}...' |
|
103 |
if not tokenizer or not model:
|
104 |
-
logger.warning("GPT-2
|
105 |
-
return prompt
|
106 |
-
|
|
|
|
|
|
|
|
|
|
|
107 |
try:
|
108 |
-
|
109 |
-
inputs = tokenizer(
|
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,
|
@@ -124,87 +128,127 @@ def generate_script(prompt, max_length=150):
|
|
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 |
-
|
131 |
-
|
132 |
-
|
133 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
134 |
if sentences and sentences[0].strip():
|
135 |
final_text = sentences[0].strip() + '.'
|
136 |
-
|
|
|
137 |
final_text += " " + sentences[1].strip() + "."
|
138 |
-
final_text = final_text.replace("..", ".")
|
139 |
|
140 |
-
logger.info(f"Guion generado (Truncado): '{final_text[:100]}...'")
|
141 |
return final_text.strip()
|
142 |
-
|
143 |
-
logger.info(f"Guion generado (sin oraciones completas): '{
|
144 |
-
return
|
145 |
|
146 |
except Exception as e:
|
147 |
-
logger.error(f"Error
|
148 |
-
logger.warning("
|
149 |
return prompt.strip()
|
150 |
|
151 |
-
|
152 |
-
|
|
|
153 |
if not text or not text.strip():
|
154 |
-
logger.warning("
|
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
|
163 |
return True
|
164 |
else:
|
165 |
-
logger.error(f"TTS
|
166 |
return False
|
167 |
-
|
168 |
except Exception as e:
|
169 |
-
logger.error(f"Error
|
170 |
return False
|
171 |
|
172 |
def download_video_file(url, temp_dir):
|
173 |
if not url:
|
174 |
-
logger.warning("
|
175 |
return None
|
176 |
-
|
177 |
try:
|
178 |
-
logger.info(f"
|
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 |
with open(output_path, 'wb') as f:
|
186 |
for chunk in r.iter_content(chunk_size=8192):
|
187 |
f.write(chunk)
|
188 |
|
189 |
if os.path.exists(output_path) and os.path.getsize(output_path) > 1000:
|
190 |
-
logger.info(f"Video
|
191 |
return output_path
|
192 |
else:
|
193 |
-
logger.warning(f"
|
194 |
if os.path.exists(output_path):
|
195 |
os.remove(output_path)
|
196 |
return None
|
197 |
|
198 |
except requests.exceptions.RequestException as e:
|
199 |
-
logger.error(f"
|
200 |
except Exception as e:
|
201 |
-
logger.error(f"
|
202 |
-
|
203 |
return None
|
204 |
|
205 |
def loop_audio_to_length(audio_clip, target_duration):
|
206 |
-
logger.debug(f"
|
207 |
-
|
208 |
if audio_clip is None or audio_clip.duration is None or audio_clip.duration <= 0:
|
209 |
logger.warning("Input audio clip is invalid (None or zero duration), cannot loop.")
|
210 |
try:
|
@@ -223,16 +267,16 @@ def loop_audio_to_length(audio_clip, target_duration):
|
|
223 |
except: pass
|
224 |
return AudioFileClip(filename="")
|
225 |
return trimmed_clip
|
226 |
-
|
227 |
loops = math.ceil(target_duration / audio_clip.duration)
|
228 |
-
logger.debug(f"
|
229 |
-
|
230 |
audio_segments = [audio_clip] * loops
|
231 |
looped_audio = None
|
232 |
final_looped_audio = None
|
233 |
try:
|
234 |
looped_audio = concatenate_audioclips(audio_segments)
|
235 |
-
|
236 |
if looped_audio.duration is None or looped_audio.duration <= 0:
|
237 |
logger.error("Concatenated audio clip is invalid (None or zero duration).")
|
238 |
raise ValueError("Invalid concatenated audio.")
|
@@ -263,9 +307,9 @@ def loop_audio_to_length(audio_clip, target_duration):
|
|
263 |
|
264 |
|
265 |
def extract_visual_keywords_from_script(script_text):
|
266 |
-
logger.info("
|
267 |
if not script_text or not script_text.strip():
|
268 |
-
logger.warning("
|
269 |
return ["naturaleza", "ciudad", "paisaje"]
|
270 |
|
271 |
clean_text = re.sub(r'[^\w\sáéíóúñÁÉÍÓÚÑ]', '', script_text)
|
@@ -273,13 +317,13 @@ def extract_visual_keywords_from_script(script_text):
|
|
273 |
|
274 |
if kw_model:
|
275 |
try:
|
276 |
-
logger.debug("
|
277 |
keywords1 = kw_model.extract_keywords(clean_text, keyphrase_ngram_range=(1, 1), stop_words='spanish', top_n=5)
|
278 |
keywords2 = kw_model.extract_keywords(clean_text, keyphrase_ngram_range=(2, 2), stop_words='spanish', top_n=3)
|
279 |
-
|
280 |
all_keywords = keywords1 + keywords2
|
281 |
all_keywords.sort(key=lambda item: item[1], reverse=True)
|
282 |
-
|
283 |
seen_keywords = set()
|
284 |
for keyword, score in all_keywords:
|
285 |
formatted_keyword = keyword.lower().replace(" ", "+")
|
@@ -290,76 +334,78 @@ def extract_visual_keywords_from_script(script_text):
|
|
290 |
break
|
291 |
|
292 |
if keywords_list:
|
293 |
-
logger.debug(f"
|
294 |
return keywords_list
|
295 |
|
296 |
except Exception as e:
|
297 |
-
logger.warning(f"KeyBERT
|
298 |
-
|
299 |
-
logger.debug("
|
300 |
words = clean_text.lower().split()
|
301 |
-
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",
|
|
|
302 |
|
303 |
valid_words = [word for word in words if len(word) > 3 and word not in stop_words]
|
304 |
-
|
305 |
if not valid_words:
|
306 |
-
logger.warning("No
|
307 |
return ["naturaleza", "ciudad", "paisaje"]
|
308 |
-
|
309 |
word_counts = Counter(valid_words)
|
310 |
top_keywords = [word.replace(" ", "+") for word, _ in word_counts.most_common(5)]
|
311 |
-
|
312 |
if not top_keywords:
|
313 |
-
logger.warning("
|
314 |
return ["naturaleza", "ciudad", "paisaje"]
|
315 |
|
316 |
-
logger.info(f"
|
317 |
return top_keywords
|
318 |
|
319 |
def crear_video(prompt_type, input_text, musica_file=None):
|
320 |
logger.info("="*80)
|
321 |
-
logger.info(f"
|
322 |
logger.debug(f"Input: '{input_text[:100]}...'")
|
323 |
-
|
324 |
start_time = datetime.now()
|
325 |
temp_dir_intermediate = None
|
326 |
-
|
327 |
audio_tts_original = None
|
328 |
musica_audio_original = None
|
329 |
-
audio_tts = None
|
330 |
-
musica_audio = None
|
331 |
video_base = None
|
332 |
video_final = None
|
333 |
source_clips = []
|
334 |
clips_to_concatenate = []
|
335 |
|
336 |
try:
|
337 |
-
# 1.
|
338 |
if prompt_type == "Generar Guion con IA":
|
339 |
guion = generate_script(input_text)
|
340 |
else:
|
341 |
guion = input_text.strip()
|
342 |
|
343 |
-
logger.info(f"
|
344 |
-
|
345 |
if not guion.strip():
|
346 |
-
logger.error("
|
347 |
-
raise ValueError("
|
348 |
-
|
349 |
temp_dir_intermediate = tempfile.mkdtemp(prefix="video_gen_intermediate_")
|
350 |
-
logger.info(f"
|
351 |
temp_intermediate_files = []
|
352 |
|
353 |
-
# 2.
|
354 |
-
logger.info("
|
355 |
voz_path = os.path.join(temp_dir_intermediate, "voz.mp3")
|
356 |
-
|
357 |
-
|
358 |
-
|
|
|
359 |
temp_intermediate_files.append(voz_path)
|
360 |
|
361 |
audio_tts_original = AudioFileClip(voz_path)
|
362 |
-
|
363 |
if audio_tts_original.reader is None or audio_tts_original.duration is None or audio_tts_original.duration <= 0:
|
364 |
logger.critical("Initial TTS audio clip is invalid (reader is None or duration <= 0).")
|
365 |
try: audio_tts_original.close()
|
@@ -368,27 +414,27 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
368 |
raise ValueError("Generated voice audio is invalid.")
|
369 |
|
370 |
audio_tts = audio_tts_original
|
371 |
-
audio_duration =
|
372 |
-
logger.info(f"
|
373 |
|
374 |
if audio_duration < 1.0:
|
375 |
-
logger.error(f"
|
376 |
raise ValueError("Generated voice audio is too short (min 1 second required).")
|
377 |
|
378 |
-
# 3.
|
379 |
-
logger.info("
|
380 |
try:
|
381 |
keywords = extract_visual_keywords_from_script(guion)
|
382 |
-
logger.info(f"
|
383 |
except Exception as e:
|
384 |
-
logger.error(f"Error
|
385 |
keywords = ["naturaleza", "paisaje"]
|
386 |
|
387 |
if not keywords:
|
388 |
keywords = ["video", "background"]
|
389 |
|
390 |
-
# 4.
|
391 |
-
logger.info("
|
392 |
videos_data = []
|
393 |
total_desired_videos = 10
|
394 |
per_page_per_keyword = max(1, total_desired_videos // len(keywords))
|
@@ -399,12 +445,12 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
399 |
videos = buscar_videos_pexels(keyword, PEXELS_API_KEY, per_page=per_page_per_keyword)
|
400 |
if videos:
|
401 |
videos_data.extend(videos)
|
402 |
-
logger.info(f"
|
403 |
except Exception as e:
|
404 |
-
logger.warning(f"Error
|
405 |
|
406 |
if len(videos_data) < total_desired_videos / 2:
|
407 |
-
logger.warning(f"
|
408 |
generic_keywords = ["nature", "city", "background", "abstract"]
|
409 |
for keyword in generic_keywords:
|
410 |
if len(videos_data) >= total_desired_videos: break
|
@@ -412,21 +458,21 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
412 |
videos = buscar_videos_pexels(keyword, PEXELS_API_KEY, per_page=2)
|
413 |
if videos:
|
414 |
videos_data.extend(videos)
|
415 |
-
logger.info(f"
|
416 |
except Exception as e:
|
417 |
-
logger.warning(f"Error
|
418 |
|
419 |
if not videos_data:
|
420 |
-
logger.error("No videos
|
421 |
-
raise ValueError("No
|
422 |
-
|
423 |
video_paths = []
|
424 |
-
logger.info(f"
|
425 |
for video in videos_data:
|
426 |
if 'video_files' not in video or not video['video_files']:
|
427 |
-
logger.debug(f"
|
428 |
continue
|
429 |
-
|
430 |
try:
|
431 |
best_quality = None
|
432 |
for vf in sorted(video['video_files'], key=lambda x: x.get('width', 0) * x.get('height', 0), reverse=True):
|
@@ -439,39 +485,39 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
439 |
if path:
|
440 |
video_paths.append(path)
|
441 |
temp_intermediate_files.append(path)
|
442 |
-
logger.info(f"Video
|
443 |
else:
|
444 |
-
logger.warning(f"
|
445 |
else:
|
446 |
-
logger.warning(f"No
|
447 |
|
448 |
except Exception as e:
|
449 |
-
logger.warning(f"Error
|
450 |
|
451 |
-
logger.info(f"
|
452 |
if not video_paths:
|
453 |
-
logger.error("
|
454 |
-
raise ValueError("
|
455 |
-
|
456 |
-
# 5.
|
457 |
-
logger.info("
|
458 |
current_duration = 0
|
459 |
min_clip_duration = 0.5
|
460 |
max_clip_segment = 10.0
|
461 |
|
462 |
for i, path in enumerate(video_paths):
|
463 |
if current_duration >= audio_duration + max_clip_segment:
|
464 |
-
logger.debug(f"Video base
|
465 |
break
|
466 |
|
467 |
clip = None
|
468 |
try:
|
469 |
-
logger.debug(f"[{i+1}/{len(video_paths)}]
|
470 |
clip = VideoFileClip(path)
|
471 |
source_clips.append(clip)
|
472 |
|
473 |
if clip.reader is None or clip.duration is None or clip.duration <= 0:
|
474 |
-
logger.warning(f"[{i+1}/{len(video_paths)}]
|
475 |
continue
|
476 |
|
477 |
remaining_needed = audio_duration - current_duration
|
@@ -486,46 +532,46 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
486 |
try:
|
487 |
sub = clip.subclip(0, segment_duration)
|
488 |
if sub.reader is None or sub.duration is None or sub.duration <= 0:
|
489 |
-
logger.warning(f"[{i+1}/{len(video_paths)}]
|
490 |
try: sub.close()
|
491 |
except: pass
|
492 |
continue
|
493 |
-
|
494 |
clips_to_concatenate.append(sub)
|
495 |
-
current_duration += sub.duration
|
496 |
-
logger.debug(f"[{i+1}/{len(video_paths)}]
|
497 |
-
|
498 |
except Exception as sub_e:
|
499 |
-
logger.warning(f"[{i+1}/{len(video_paths)}] Error
|
500 |
continue
|
501 |
else:
|
502 |
-
logger.debug(f"[{i+1}/{len(video_paths)}] Clip {path} ({clip.duration:.1f}s)
|
503 |
else:
|
504 |
-
logger.debug(f"[{i+1}/{len(video_paths)}]
|
505 |
|
506 |
except Exception as e:
|
507 |
-
logger.warning(f"[{i+1}/{len(video_paths)}] Error
|
508 |
continue
|
509 |
|
510 |
-
logger.info(f"
|
511 |
|
512 |
if not clips_to_concatenate:
|
513 |
-
logger.error("No
|
514 |
-
raise ValueError("No
|
515 |
-
|
516 |
-
logger.info(f"
|
517 |
concatenated_base = None
|
518 |
try:
|
519 |
concatenated_base = concatenate_videoclips(clips_to_concatenate, method="chain")
|
520 |
-
logger.info(f"
|
521 |
|
522 |
if concatenated_base is None or concatenated_base.duration is None or concatenated_base.duration <= 0:
|
523 |
-
logger.critical("
|
524 |
-
raise ValueError("
|
525 |
|
526 |
except Exception as e:
|
527 |
-
logger.critical(f"Error
|
528 |
-
raise ValueError("
|
529 |
finally:
|
530 |
for clip_segment in clips_to_concatenate:
|
531 |
try: clip_segment.close()
|
@@ -537,47 +583,47 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
537 |
final_video_base = video_base
|
538 |
|
539 |
if final_video_base.duration < audio_duration:
|
540 |
-
logger.info(f"
|
541 |
-
|
542 |
num_full_repeats = int(audio_duration // final_video_base.duration)
|
543 |
remaining_duration = audio_duration % final_video_base.duration
|
544 |
-
|
545 |
repeated_clips_list = [final_video_base] * num_full_repeats
|
546 |
-
|
547 |
if remaining_duration > 0:
|
548 |
try:
|
549 |
remaining_clip = final_video_base.subclip(0, remaining_duration)
|
550 |
if remaining_clip is None or remaining_clip.duration is None or remaining_clip.duration <= 0:
|
551 |
-
logger.warning(f"
|
552 |
try: remaining_clip.close()
|
553 |
except: pass
|
554 |
else:
|
555 |
repeated_clips_list.append(remaining_clip)
|
556 |
-
logger.debug(f"
|
557 |
|
558 |
except Exception as e:
|
559 |
-
logger.warning(f"Error
|
560 |
|
561 |
if repeated_clips_list:
|
562 |
-
logger.info(f"
|
563 |
video_base_repeated = None
|
564 |
try:
|
565 |
video_base_repeated = concatenate_videoclips(repeated_clips_list, method="chain")
|
566 |
-
logger.info(f"
|
567 |
-
|
568 |
if video_base_repeated is None or video_base_repeated.duration is None or video_base_repeated.duration <= 0:
|
569 |
-
logger.critical("
|
570 |
-
raise ValueError("
|
571 |
|
572 |
if final_video_base is not video_base_repeated:
|
573 |
try: final_video_base.close()
|
574 |
except: pass
|
575 |
-
|
576 |
final_video_base = video_base_repeated
|
577 |
|
578 |
except Exception as e:
|
579 |
-
logger.critical(f"Error
|
580 |
-
raise ValueError("
|
581 |
finally:
|
582 |
if 'repeated_clips_list' in locals():
|
583 |
for clip in repeated_clips_list:
|
@@ -587,13 +633,13 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
587 |
|
588 |
|
589 |
if final_video_base.duration > audio_duration:
|
590 |
-
logger.info(f"
|
591 |
trimmed_video_base = None
|
592 |
try:
|
593 |
trimmed_video_base = final_video_base.subclip(0, audio_duration)
|
594 |
if trimmed_video_base is None or trimmed_video_base.duration is None or trimmed_video_base.duration <= 0:
|
595 |
-
logger.critical("
|
596 |
-
raise ValueError("
|
597 |
|
598 |
if final_video_base is not trimmed_video_base:
|
599 |
try: final_video_base.close()
|
@@ -602,23 +648,23 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
602 |
final_video_base = trimmed_video_base
|
603 |
|
604 |
except Exception as e:
|
605 |
-
logger.critical(f"Error
|
606 |
-
raise ValueError("
|
607 |
|
608 |
|
609 |
if final_video_base is None or final_video_base.duration is None or final_video_base.duration <= 0:
|
610 |
-
logger.critical("
|
611 |
-
raise ValueError("
|
612 |
|
613 |
if final_video_base.size is None or final_video_base.size[0] <= 0 or final_video_base.size[1] <= 0:
|
614 |
-
logger.critical(f"
|
615 |
-
raise ValueError("
|
616 |
|
617 |
video_base = final_video_base
|
618 |
|
619 |
-
# 6.
|
620 |
-
logger.info("
|
621 |
-
|
622 |
final_audio = audio_tts_original
|
623 |
|
624 |
musica_audio_looped = None
|
@@ -629,56 +675,59 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
629 |
music_path = os.path.join(temp_dir_intermediate, "musica_bg.mp3")
|
630 |
shutil.copyfile(musica_file, music_path)
|
631 |
temp_intermediate_files.append(music_path)
|
632 |
-
logger.info(f"
|
633 |
-
|
634 |
musica_audio_original = AudioFileClip(music_path)
|
635 |
-
|
636 |
if musica_audio_original.reader is None or musica_audio_original.duration is None or musica_audio_original.duration <= 0:
|
637 |
-
logger.warning("
|
638 |
try: musica_audio_original.close()
|
639 |
except: pass
|
640 |
musica_audio_original = None
|
641 |
else:
|
|
|
642 |
musica_audio_looped = loop_audio_to_length(musica_audio_original, video_base.duration)
|
643 |
-
logger.debug(f"
|
644 |
|
645 |
if musica_audio_looped is None or musica_audio_looped.duration is None or musica_audio_looped.duration <= 0:
|
646 |
-
logger.warning("
|
647 |
try: musica_audio_looped.close()
|
648 |
except: pass
|
649 |
musica_audio_looped = None
|
650 |
|
651 |
|
652 |
if musica_audio_looped:
|
|
|
|
|
653 |
composite_audio = CompositeAudioClip([
|
654 |
musica_audio_looped.volumex(0.2),
|
655 |
audio_tts_original.volumex(1.0)
|
656 |
])
|
657 |
-
|
658 |
if composite_audio.duration is None or composite_audio.duration <= 0:
|
659 |
-
logger.warning("
|
660 |
try: composite_audio.close()
|
661 |
except: pass
|
662 |
final_audio = audio_tts_original
|
663 |
else:
|
664 |
-
logger.info("
|
665 |
final_audio = composite_audio
|
666 |
musica_audio = musica_audio_looped
|
667 |
|
668 |
except Exception as e:
|
669 |
-
logger.warning(f"Error
|
670 |
final_audio = audio_tts_original
|
671 |
musica_audio = None
|
672 |
-
logger.warning("
|
673 |
|
674 |
|
675 |
if final_audio.duration is not None and abs(final_audio.duration - video_base.duration) > 0.2:
|
676 |
-
logger.warning(f"
|
677 |
try:
|
678 |
if final_audio.duration > video_base.duration:
|
679 |
trimmed_final_audio = final_audio.subclip(0, video_base.duration)
|
680 |
if trimmed_final_audio.duration is None or trimmed_final_audio.duration <= 0:
|
681 |
-
logger.warning("
|
682 |
try: trimmed_final_audio.close()
|
683 |
except: pass
|
684 |
else:
|
@@ -686,21 +735,21 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
686 |
try: final_audio.close()
|
687 |
except: pass
|
688 |
final_audio = trimmed_final_audio
|
689 |
-
logger.warning("
|
690 |
except Exception as e:
|
691 |
-
logger.warning(f"Error
|
692 |
-
|
693 |
|
|
|
|
|
694 |
video_final = video_base.set_audio(final_audio)
|
695 |
|
696 |
if video_final is None or video_final.duration is None or video_final.duration <= 0:
|
697 |
-
logger.critical("
|
698 |
-
raise ValueError("
|
699 |
-
|
700 |
|
701 |
output_filename = "final_video.mp4"
|
702 |
output_path = os.path.join(temp_dir_intermediate, output_filename)
|
703 |
-
logger.info(f"
|
704 |
|
705 |
video_final.write_videofile(
|
706 |
output_path,
|
@@ -711,253 +760,274 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
711 |
preset="medium",
|
712 |
logger='bar'
|
713 |
)
|
714 |
-
|
715 |
total_time = (datetime.now() - start_time).total_seconds()
|
716 |
-
logger.info(f"VIDEO
|
717 |
-
|
718 |
return output_path
|
719 |
-
|
720 |
except ValueError as ve:
|
721 |
-
logger.error(f"
|
722 |
raise ve
|
723 |
except Exception as e:
|
724 |
-
logger.critical(f"
|
725 |
raise e
|
726 |
finally:
|
727 |
-
logger.info("
|
728 |
-
|
|
|
729 |
for clip in source_clips:
|
730 |
try:
|
731 |
clip.close()
|
732 |
except Exception as e:
|
733 |
-
logger.warning(f"Error
|
734 |
-
|
735 |
-
for clip_segment in clips_to_concatenate:
|
736 |
-
try:
|
737 |
-
clip_segment.close()
|
738 |
-
except Exception as e:
|
739 |
-
logger.warning(f"Error closing video segment clip in finally: {str(e)}")
|
740 |
|
741 |
-
#
|
742 |
-
|
743 |
-
try:
|
744 |
-
musica_audio.close()
|
745 |
-
except Exception as e:
|
746 |
-
logger.warning(f"Error closing musica_audio (processed) in finally: {str(e)}")
|
747 |
-
|
748 |
-
if musica_audio_original is not None and musica_audio_original is not musica_audio:
|
749 |
try:
|
750 |
-
|
751 |
except Exception as e:
|
752 |
-
logger.warning(f"Error
|
753 |
|
754 |
-
#
|
755 |
-
|
756 |
-
|
757 |
-
|
758 |
-
|
759 |
-
|
760 |
-
|
|
|
761 |
|
762 |
-
|
763 |
-
|
764 |
-
|
765 |
-
|
766 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
767 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
768 |
|
769 |
-
# Close video clips: final video (should cascade), then video base if it wasn't the final
|
770 |
-
if video_final is not None:
|
771 |
-
try:
|
772 |
-
video_final.close()
|
773 |
-
except Exception as e:
|
774 |
-
logger.warning(f"Error closing video_final in finally: {str(e)}")
|
775 |
-
# Note: video_base variable is reassigned to final_video_base. Close only if it exists and is different from video_final
|
776 |
-
elif video_base is not None and video_base is not video_final: # Check if video_base holds a different clip
|
777 |
-
try:
|
778 |
-
video_base.close()
|
779 |
-
except Exception as e:
|
780 |
-
logger.warning(f"Error closing video_base in finally: {str(e)}")
|
781 |
|
|
|
|
|
782 |
|
783 |
-
#
|
784 |
if temp_dir_intermediate and os.path.exists(temp_dir_intermediate):
|
785 |
final_output_in_temp = os.path.join(temp_dir_intermediate, "final_video.mp4")
|
786 |
-
|
787 |
for path in temp_intermediate_files:
|
788 |
try:
|
|
|
789 |
if os.path.isfile(path) and path != final_output_in_temp:
|
790 |
-
logger.debug(f"
|
791 |
os.remove(path)
|
|
|
|
|
792 |
except Exception as e:
|
793 |
-
logger.warning(f"
|
794 |
-
|
795 |
-
logger.info(f"Intermediate temporary directory {temp_dir_intermediate} will persist for Gradio to read the final video.")
|
796 |
|
|
|
|
|
797 |
|
798 |
-
|
|
|
799 |
logger.info("="*80)
|
800 |
-
logger.info("
|
801 |
-
|
802 |
input_text = prompt_ia if prompt_type == "Generar Guion con IA" else prompt_manual
|
803 |
-
|
804 |
output_video = None
|
805 |
-
output_file = None
|
806 |
status_msg = gr.update(value="⏳ Procesando...", interactive=False)
|
807 |
|
808 |
if not input_text or not input_text.strip():
|
809 |
-
logger.warning("
|
810 |
return None, None, gr.update(value="⚠️ Por favor, ingresa texto para el guion o el tema.", interactive=False)
|
811 |
|
812 |
-
logger.info(f"
|
813 |
-
logger.debug(f"
|
814 |
if musica_file:
|
815 |
-
logger.info(f"
|
816 |
else:
|
817 |
-
logger.info("No
|
818 |
|
819 |
try:
|
820 |
-
logger.info("
|
821 |
video_path = crear_video(prompt_type, input_text, musica_file)
|
822 |
-
|
823 |
if video_path and os.path.exists(video_path):
|
824 |
-
logger.info(f"crear_video
|
825 |
-
logger.info(f"
|
826 |
-
output_video = video_path
|
827 |
-
output_file = video_path
|
828 |
status_msg = gr.update(value="✅ Video generado exitosamente.", interactive=False)
|
829 |
else:
|
830 |
-
logger.error(f"crear_video
|
|
|
831 |
status_msg = gr.update(value="❌ Error: La generación del video falló o el archivo no se creó correctamente.", interactive=False)
|
832 |
|
833 |
except ValueError as ve:
|
834 |
-
logger.warning(f"
|
|
|
835 |
status_msg = gr.update(value=f"⚠️ Error de validación: {str(ve)}", interactive=False)
|
836 |
except Exception as e:
|
837 |
-
logger.critical(f"
|
|
|
838 |
status_msg = gr.update(value=f"❌ Error inesperado: {str(e)}", interactive=False)
|
839 |
finally:
|
840 |
-
logger.info("
|
|
|
841 |
return output_video, output_file, status_msg
|
842 |
|
843 |
|
844 |
-
# Gradio
|
845 |
with gr.Blocks(title="Generador de Videos con IA", theme=gr.themes.Soft(), css="""
|
846 |
.gradio-container {max-width: 800px; margin: auto;}
|
847 |
h1 {text-align: center;}
|
848 |
""") as app:
|
849 |
-
|
850 |
-
gr.Markdown("# 🎬
|
851 |
-
gr.Markdown("
|
852 |
-
|
853 |
with gr.Row():
|
854 |
with gr.Column():
|
855 |
prompt_type = gr.Radio(
|
856 |
-
["Generar Guion con IA", "Usar Mi Guion"],
|
857 |
-
label="
|
858 |
value="Generar Guion con IA"
|
859 |
)
|
860 |
-
|
861 |
with gr.Column(visible=True) as ia_guion_column:
|
862 |
prompt_ia = gr.Textbox(
|
863 |
-
label="
|
864 |
lines=2,
|
865 |
-
placeholder="
|
866 |
max_lines=4,
|
867 |
value=""
|
868 |
)
|
869 |
-
|
870 |
with gr.Column(visible=False) as manual_guion_column:
|
871 |
prompt_manual = gr.Textbox(
|
872 |
-
label="
|
873 |
lines=5,
|
874 |
-
placeholder="
|
875 |
max_lines=10,
|
876 |
value=""
|
877 |
)
|
878 |
-
|
879 |
musica_input = gr.Audio(
|
880 |
-
label="
|
881 |
type="filepath",
|
882 |
interactive=True,
|
883 |
value=None
|
884 |
)
|
885 |
-
|
886 |
-
generate_btn = gr.Button("✨
|
887 |
-
|
888 |
with gr.Column():
|
889 |
video_output = gr.Video(
|
890 |
-
label="
|
891 |
interactive=False,
|
892 |
height=400
|
893 |
)
|
|
|
894 |
file_output = gr.File(
|
895 |
-
label="
|
896 |
interactive=False,
|
897 |
-
visible=False
|
898 |
)
|
899 |
status_output = gr.Textbox(
|
900 |
-
label="
|
901 |
interactive=False,
|
902 |
show_label=False,
|
903 |
-
placeholder="
|
904 |
-
value="
|
905 |
)
|
906 |
|
907 |
prompt_type.change(
|
908 |
-
lambda x: (gr.update(visible=x == "Generar Guion con IA"),
|
909 |
gr.update(visible=x == "Usar Mi Guion")),
|
910 |
inputs=prompt_type,
|
911 |
outputs=[ia_guion_column, manual_guion_column]
|
912 |
)
|
913 |
|
|
|
914 |
generate_btn.click(
|
915 |
-
|
|
|
916 |
outputs=[video_output, file_output, status_output],
|
917 |
-
queue=True,
|
918 |
).then(
|
|
|
919 |
run_app,
|
920 |
inputs=[prompt_type, prompt_ia, prompt_manual, musica_input],
|
921 |
-
outputs=[video_output, file_output, status_output]
|
922 |
).then(
|
923 |
-
|
924 |
-
|
925 |
-
|
|
|
|
|
926 |
)
|
927 |
|
928 |
|
929 |
-
gr.Markdown("###
|
930 |
gr.Markdown("""
|
931 |
-
1. **
|
932 |
-
2. **
|
933 |
-
- "
|
934 |
-
- "
|
935 |
-
3. **
|
936 |
-
4. **
|
937 |
-
5.
|
938 |
-
6.
|
939 |
-
7.
|
940 |
""")
|
941 |
gr.Markdown("---")
|
942 |
-
gr.Markdown("
|
943 |
|
944 |
if __name__ == "__main__":
|
945 |
-
logger.info("
|
946 |
try:
|
947 |
from moviepy.editor import ColorClip
|
948 |
try:
|
949 |
temp_clip = ColorClip((100,100), color=(255,0,0), duration=0.1)
|
950 |
temp_clip.close()
|
951 |
-
logger.info("
|
952 |
except Exception as e:
|
953 |
-
logger.critical(f"
|
954 |
|
955 |
except Exception as e:
|
956 |
-
logger.critical(f"
|
957 |
|
958 |
-
logger.info("
|
959 |
try:
|
|
|
|
|
960 |
app.launch(server_name="0.0.0.0", server_port=7860, share=False)
|
961 |
except Exception as e:
|
962 |
-
logger.critical(f"
|
963 |
raise
|
|
|
9 |
import torch
|
10 |
from transformers import GPT2Tokenizer, GPT2LMHeadModel
|
11 |
from keybert import KeyBERT
|
12 |
+
from moviepy.editor import VideoFileClip, concatenate_videoclip, concatenate_videoclips, AudioFileClip, CompositeAudioClip, concatenate_audioclips, AudioClip
|
13 |
import re
|
14 |
import math
|
15 |
import shutil
|
16 |
import json
|
17 |
from collections import Counter
|
18 |
|
19 |
+
# Configuración de logging
|
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("INICIO DE EJECUCIÓN - GENERADOR DE VIDEOS")
|
31 |
logger.info("="*80)
|
32 |
|
33 |
+
# Clave API de Pexels
|
34 |
PEXELS_API_KEY = os.environ.get("PEXELS_API_KEY")
|
35 |
if not PEXELS_API_KEY:
|
36 |
+
logger.critical("NO SE ENCONTRÓ PEXELS_API_KEY EN VARIABLES DE ENTORNO")
|
37 |
+
# raise ValueError("API key de Pexels no configurada") # Descomentar para forzar fallo si no está
|
|
|
38 |
|
39 |
+
# Inicialización de modelos
|
40 |
MODEL_NAME = "datificate/gpt2-small-spanish"
|
41 |
+
logger.info(f"Inicializando modelo GPT-2: {MODEL_NAME}")
|
42 |
tokenizer = None
|
43 |
model = None
|
44 |
try:
|
|
|
46 |
model = GPT2LMHeadModel.from_pretrained(MODEL_NAME).eval()
|
47 |
if tokenizer.pad_token is None:
|
48 |
tokenizer.pad_token = tokenizer.eos_token
|
49 |
+
logger.info(f"Modelo GPT-2 cargado | Vocabulario: {len(tokenizer)} tokens")
|
50 |
except Exception as e:
|
51 |
+
logger.error(f"FALLA CRÍTICA al cargar GPT-2: {str(e)}", exc_info=True)
|
52 |
tokenizer = model = None
|
53 |
|
54 |
+
logger.info("Cargando modelo KeyBERT...")
|
55 |
kw_model = None
|
56 |
try:
|
57 |
kw_model = KeyBERT('distilbert-base-multilingual-cased')
|
58 |
+
logger.info("KeyBERT inicializado correctamente")
|
59 |
except Exception as e:
|
60 |
+
logger.error(f"FALLA al cargar KeyBERT: {str(e)}", exc_info=True)
|
61 |
kw_model = None
|
62 |
|
63 |
def buscar_videos_pexels(query, api_key, per_page=5):
|
64 |
if not api_key:
|
65 |
+
logger.warning("No se puede buscar en Pexels: API Key no configurada.")
|
66 |
return []
|
67 |
|
68 |
+
logger.debug(f"Buscando en Pexels: '{query}' | Resultados: {per_page}")
|
69 |
headers = {"Authorization": api_key}
|
70 |
try:
|
71 |
params = {
|
|
|
74 |
"orientation": "landscape",
|
75 |
"size": "medium"
|
76 |
}
|
77 |
+
|
78 |
response = requests.get(
|
79 |
"https://api.pexels.com/videos/search",
|
80 |
headers=headers,
|
|
|
82 |
timeout=20
|
83 |
)
|
84 |
response.raise_for_status()
|
85 |
+
|
86 |
data = response.json()
|
87 |
videos = data.get('videos', [])
|
88 |
+
logger.info(f"Pexels: {len(videos)} videos encontrados para '{query}'")
|
89 |
return videos
|
90 |
+
|
91 |
except requests.exceptions.RequestException as e:
|
92 |
+
logger.error(f"Error de conexión Pexels para '{query}': {str(e)}")
|
93 |
except json.JSONDecodeError:
|
94 |
+
logger.error(f"Pexels: JSON inválido recibido | Status: {response.status_code} | Respuesta: {response.text[:200]}...")
|
95 |
except Exception as e:
|
96 |
+
logger.error(f"Error inesperado Pexels para '{query}': {str(e)}", exc_info=True)
|
97 |
+
|
98 |
return []
|
99 |
|
100 |
def generate_script(prompt, max_length=150):
|
101 |
+
logger.info(f"Generando guión | Prompt: '{prompt[:50]}...' | Longitud máxima: {max_length}")
|
102 |
if not tokenizer or not model:
|
103 |
+
logger.warning("Modelos GPT-2 no disponibles - Usando prompt original como guion.")
|
104 |
+
return prompt.strip()
|
105 |
+
|
106 |
+
# Frase de instrucción que se le da a la IA
|
107 |
+
instruction_phrase_start = "Escribe un guion corto, interesante y coherente sobre:"
|
108 |
+
# Construir el prompt exacto que se le pasará a la IA
|
109 |
+
ai_prompt = f"{instruction_phrase_start} {prompt}"
|
110 |
+
|
111 |
try:
|
112 |
+
# Generar texto usando el prompt completo
|
113 |
+
inputs = tokenizer(ai_prompt, return_tensors="pt", truncation=True, max_length=512)
|
114 |
+
|
115 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
116 |
model.to(device)
|
117 |
inputs = {k: v.to(device) for k, v in inputs.items()}
|
118 |
|
119 |
outputs = model.generate(
|
120 |
**inputs,
|
121 |
+
max_length=max_length + inputs[list(inputs.keys())[0]].size(1), # Longitud máxima incluyendo los tokens de entrada
|
122 |
do_sample=True,
|
123 |
top_p=0.9,
|
124 |
top_k=40,
|
|
|
128 |
eos_token_id=tokenizer.eos_token_id,
|
129 |
no_repeat_ngram_size=3
|
130 |
)
|
131 |
+
|
132 |
text = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
133 |
+
|
134 |
+
# --- Limpiar la frase de instrucción inicial de la salida de la IA ---
|
135 |
+
cleaned_text = text.strip()
|
136 |
+
# Intentar encontrar el inicio de la respuesta real después de la instrucción
|
137 |
+
# A veces la IA repite el prompt o la instrucción
|
138 |
+
try:
|
139 |
+
# Buscar el final de la frase de instrucción literal en la salida
|
140 |
+
instruction_end_idx = text.find(instruction_phrase)
|
141 |
+
if instruction_end_idx != -1:
|
142 |
+
# Tomar el texto que viene *después* de la instrucción exacta
|
143 |
+
cleaned_text = text[instruction_end_idx + len(instruction_phrase):].strip()
|
144 |
+
logger.debug("Instrucción inicial encontrada y eliminada del guión generado.")
|
145 |
+
else:
|
146 |
+
# Si no se encuentra la frase exacta, buscar solo el inicio de la instrucción base
|
147 |
+
instruction_start_idx = text.find(instruction_phrase_start)
|
148 |
+
if instruction_start_idx != -1:
|
149 |
+
# Tomar texto después de la frase base + prompt (heurística)
|
150 |
+
prompt_in_output_idx = text.find(prompt, instruction_start_idx)
|
151 |
+
if prompt_in_output_idx != -1:
|
152 |
+
cleaned_text = text[prompt_in_output_idx + len(prompt):].strip()
|
153 |
+
logger.debug("Instrucción base y prompt encontrados y eliminados del guión generado.")
|
154 |
+
else:
|
155 |
+
# Fallback: si la instrucción base está pero no el prompt después, tomar después de la instrucción base
|
156 |
+
cleaned_text = text[instruction_start_idx + len(instruction_phrase_start):].strip()
|
157 |
+
logger.debug("Instrucción base encontrada, eliminada del guión generado (sin prompt detectado).")
|
158 |
+
|
159 |
+
except Exception as e:
|
160 |
+
logger.warning(f"Error durante la limpieza heurística del guión de IA: {e}. Usando texto generado sin limpieza adicional.")
|
161 |
+
cleaned_text = re.sub(r'<[^>]+>', '', text).strip() # Limpieza básica como fallback
|
162 |
+
|
163 |
+
# Asegurarse de que el texto resultante no sea solo la instrucción o vacío
|
164 |
+
if not cleaned_text or len(cleaned_text) < 10: # Umbral de longitud mínima
|
165 |
+
logger.warning("El guión generado parece muy corto o vacío después de la limpieza. Usando el texto generado original (sin limpieza heurística).")
|
166 |
+
cleaned_text = re.sub(r'<[^>]+>', '', text).strip() # Fallback al texto original limpio
|
167 |
+
|
168 |
+
# Limpieza final de caracteres especiales y espacios
|
169 |
+
cleaned_text = re.sub(r'<[^>]+>', '', cleaned_text).strip()
|
170 |
+
cleaned_text = cleaned_text.lstrip(':').strip() # Quitar posibles ':' al inicio
|
171 |
+
cleaned_text = cleaned_text.lstrip('.').strip() # Quitar posibles '.' al inicio
|
172 |
+
|
173 |
+
|
174 |
+
# Intentar obtener al menos una oración completa si es posible
|
175 |
+
sentences = cleaned_text.split('.')
|
176 |
if sentences and sentences[0].strip():
|
177 |
final_text = sentences[0].strip() + '.'
|
178 |
+
# Añadir la segunda oración si existe y es razonable
|
179 |
+
if len(sentences) > 1 and sentences[1].strip() and len(final_text.split()) < max_length * 0.7: # Usar un 70% de max_length como umbral
|
180 |
final_text += " " + sentences[1].strip() + "."
|
181 |
+
final_text = final_text.replace("..", ".") # Limpiar doble punto
|
182 |
|
183 |
+
logger.info(f"Guion generado final (Truncado a 100 chars): '{final_text[:100]}...'")
|
184 |
return final_text.strip()
|
185 |
+
|
186 |
+
logger.info(f"Guion generado final (sin oraciones completas detectadas - Truncado): '{cleaned_text[:100]}...'")
|
187 |
+
return cleaned_text.strip() # Si no se puede formar una oración, devolver el texto limpio tal cual
|
188 |
|
189 |
except Exception as e:
|
190 |
+
logger.error(f"Error generando guion con GPT-2 (fuera del bloque de limpieza): {str(e)}", exc_info=True)
|
191 |
+
logger.warning("Usando prompt original como guion debido al error de generación.")
|
192 |
return prompt.strip()
|
193 |
|
194 |
+
# CAMBIO: Voz por defecto a "es-ES-JuanNeural"
|
195 |
+
async def text_to_speech(text, output_path, voice="es-ES-JuanNeural"):
|
196 |
+
logger.info(f"Convirtiendo texto a voz | Caracteres: {len(text)} | Voz: {voice} | Salida: {output_path}")
|
197 |
if not text or not text.strip():
|
198 |
+
logger.warning("Texto vacío para TTS")
|
199 |
return False
|
200 |
+
|
201 |
try:
|
202 |
communicate = edge_tts.Communicate(text, voice)
|
203 |
await communicate.save(output_path)
|
204 |
+
|
205 |
if os.path.exists(output_path) and os.path.getsize(output_path) > 100:
|
206 |
+
logger.info(f"Audio guardado exitosamente en: {output_path} | Tamaño: {os.path.getsize(output_path)} bytes")
|
207 |
return True
|
208 |
else:
|
209 |
+
logger.error(f"TTS guardó un archivo pequeño o vacío en: {output_path}")
|
210 |
return False
|
211 |
+
|
212 |
except Exception as e:
|
213 |
+
logger.error(f"Error en TTS: {str(e)}", exc_info=True)
|
214 |
return False
|
215 |
|
216 |
def download_video_file(url, temp_dir):
|
217 |
if not url:
|
218 |
+
logger.warning("URL de video no proporcionada para descargar")
|
219 |
return None
|
220 |
+
|
221 |
try:
|
222 |
+
logger.info(f"Descargando video desde: {url[:80]}...")
|
223 |
os.makedirs(temp_dir, exist_ok=True)
|
224 |
file_name = f"video_dl_{datetime.now().strftime('%Y%m%d_%H%M%S_%f')}.mp4"
|
225 |
output_path = os.path.join(temp_dir, file_name)
|
226 |
+
|
227 |
+
with requests.get(url, stream=True, timeout=60) as r:
|
228 |
r.raise_for_status()
|
229 |
with open(output_path, 'wb') as f:
|
230 |
for chunk in r.iter_content(chunk_size=8192):
|
231 |
f.write(chunk)
|
232 |
|
233 |
if os.path.exists(output_path) and os.path.getsize(output_path) > 1000:
|
234 |
+
logger.info(f"Video descargado exitosamente: {output_path} | Tamaño: {os.path.getsize(output_path)} bytes")
|
235 |
return output_path
|
236 |
else:
|
237 |
+
logger.warning(f"Descarga parece incompleta o vacía para {url[:80]}... Archivo: {output_path} Tamaño: {os.path.getsize(output_path) if os.path.exists(output_path) else 'N/A'} bytes")
|
238 |
if os.path.exists(output_path):
|
239 |
os.remove(output_path)
|
240 |
return None
|
241 |
|
242 |
except requests.exceptions.RequestException as e:
|
243 |
+
logger.error(f"Error de descarga para {url[:80]}... : {str(e)}")
|
244 |
except Exception as e:
|
245 |
+
logger.error(f"Error inesperado descargando {url[:80]}... : {str(e)}", exc_info=True)
|
246 |
+
|
247 |
return None
|
248 |
|
249 |
def loop_audio_to_length(audio_clip, target_duration):
|
250 |
+
logger.debug(f"Ajustando audio | Duración actual: {audio_clip.duration:.2f}s | Objetivo: {target_duration:.2f}s")
|
251 |
+
|
252 |
if audio_clip is None or audio_clip.duration is None or audio_clip.duration <= 0:
|
253 |
logger.warning("Input audio clip is invalid (None or zero duration), cannot loop.")
|
254 |
try:
|
|
|
267 |
except: pass
|
268 |
return AudioFileClip(filename="")
|
269 |
return trimmed_clip
|
270 |
+
|
271 |
loops = math.ceil(target_duration / audio_clip.duration)
|
272 |
+
logger.debug(f"Creando {loops} loops de audio")
|
273 |
+
|
274 |
audio_segments = [audio_clip] * loops
|
275 |
looped_audio = None
|
276 |
final_looped_audio = None
|
277 |
try:
|
278 |
looped_audio = concatenate_audioclips(audio_segments)
|
279 |
+
|
280 |
if looped_audio.duration is None or looped_audio.duration <= 0:
|
281 |
logger.error("Concatenated audio clip is invalid (None or zero duration).")
|
282 |
raise ValueError("Invalid concatenated audio.")
|
|
|
307 |
|
308 |
|
309 |
def extract_visual_keywords_from_script(script_text):
|
310 |
+
logger.info("Extrayendo palabras clave del guion")
|
311 |
if not script_text or not script_text.strip():
|
312 |
+
logger.warning("Guion vacío, no se pueden extraer palabras clave.")
|
313 |
return ["naturaleza", "ciudad", "paisaje"]
|
314 |
|
315 |
clean_text = re.sub(r'[^\w\sáéíóúñÁÉÍÓÚÑ]', '', script_text)
|
|
|
317 |
|
318 |
if kw_model:
|
319 |
try:
|
320 |
+
logger.debug("Intentando extracción con KeyBERT...")
|
321 |
keywords1 = kw_model.extract_keywords(clean_text, keyphrase_ngram_range=(1, 1), stop_words='spanish', top_n=5)
|
322 |
keywords2 = kw_model.extract_keywords(clean_text, keyphrase_ngram_range=(2, 2), stop_words='spanish', top_n=3)
|
323 |
+
|
324 |
all_keywords = keywords1 + keywords2
|
325 |
all_keywords.sort(key=lambda item: item[1], reverse=True)
|
326 |
+
|
327 |
seen_keywords = set()
|
328 |
for keyword, score in all_keywords:
|
329 |
formatted_keyword = keyword.lower().replace(" ", "+")
|
|
|
334 |
break
|
335 |
|
336 |
if keywords_list:
|
337 |
+
logger.debug(f"Palabras clave extraídas por KeyBERT: {keywords_list}")
|
338 |
return keywords_list
|
339 |
|
340 |
except Exception as e:
|
341 |
+
logger.warning(f"KeyBERT falló: {str(e)}. Intentando método simple.")
|
342 |
+
|
343 |
+
logger.debug("Extrayendo palabras clave con método simple...")
|
344 |
words = clean_text.lower().split()
|
345 |
+
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",
|
346 |
+
"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"}
|
347 |
|
348 |
valid_words = [word for word in words if len(word) > 3 and word not in stop_words]
|
349 |
+
|
350 |
if not valid_words:
|
351 |
+
logger.warning("No se encontraron palabras clave válidas con método simple. Usando palabras clave predeterminadas.")
|
352 |
return ["naturaleza", "ciudad", "paisaje"]
|
353 |
+
|
354 |
word_counts = Counter(valid_words)
|
355 |
top_keywords = [word.replace(" ", "+") for word, _ in word_counts.most_common(5)]
|
356 |
+
|
357 |
if not top_keywords:
|
358 |
+
logger.warning("El método simple no produjo keywords. Usando palabras clave predeterminadas.")
|
359 |
return ["naturaleza", "ciudad", "paisaje"]
|
360 |
|
361 |
+
logger.info(f"Palabras clave finales: {top_keywords}")
|
362 |
return top_keywords
|
363 |
|
364 |
def crear_video(prompt_type, input_text, musica_file=None):
|
365 |
logger.info("="*80)
|
366 |
+
logger.info(f"INICIANDO CREACIÓN DE VIDEO | Tipo: {prompt_type}")
|
367 |
logger.debug(f"Input: '{input_text[:100]}...'")
|
368 |
+
|
369 |
start_time = datetime.now()
|
370 |
temp_dir_intermediate = None
|
371 |
+
|
372 |
audio_tts_original = None
|
373 |
musica_audio_original = None
|
374 |
+
audio_tts = None # Variable to hold potentially modified TTS clip (currently not used, but kept for structure)
|
375 |
+
musica_audio = None # Variable to hold potentially modified music clip (looped/trimmed)
|
376 |
video_base = None
|
377 |
video_final = None
|
378 |
source_clips = []
|
379 |
clips_to_concatenate = []
|
380 |
|
381 |
try:
|
382 |
+
# 1. Generar o usar guion
|
383 |
if prompt_type == "Generar Guion con IA":
|
384 |
guion = generate_script(input_text)
|
385 |
else:
|
386 |
guion = input_text.strip()
|
387 |
|
388 |
+
logger.info(f"Guion final ({len(guion)} chars): '{guion[:100]}...'")
|
389 |
+
|
390 |
if not guion.strip():
|
391 |
+
logger.error("El guion resultante está vacío o solo contiene espacios.")
|
392 |
+
raise ValueError("El guion está vacío.")
|
393 |
+
|
394 |
temp_dir_intermediate = tempfile.mkdtemp(prefix="video_gen_intermediate_")
|
395 |
+
logger.info(f"Directorio temporal intermedio creado: {temp_dir_intermediate}")
|
396 |
temp_intermediate_files = []
|
397 |
|
398 |
+
# 2. Generar audio de voz
|
399 |
+
logger.info("Generando audio de voz...")
|
400 |
voz_path = os.path.join(temp_dir_intermediate, "voz.mp3")
|
401 |
+
# CAMBIO: Usar voz de Juan por defecto en la llamada
|
402 |
+
if not asyncio.run(text_to_speech(guion, voz_path, voice="es-ES-JuanNeural")):
|
403 |
+
logger.error("Fallo en generación de voz")
|
404 |
+
raise ValueError("Error generando voz a partir del guion.")
|
405 |
temp_intermediate_files.append(voz_path)
|
406 |
|
407 |
audio_tts_original = AudioFileClip(voz_path)
|
408 |
+
|
409 |
if audio_tts_original.reader is None or audio_tts_original.duration is None or audio_tts_original.duration <= 0:
|
410 |
logger.critical("Initial TTS audio clip is invalid (reader is None or duration <= 0).")
|
411 |
try: audio_tts_original.close()
|
|
|
414 |
raise ValueError("Generated voice audio is invalid.")
|
415 |
|
416 |
audio_tts = audio_tts_original
|
417 |
+
audio_duration = audio_tts_original.duration
|
418 |
+
logger.info(f"Duración audio voz: {audio_duration:.2f} segundos")
|
419 |
|
420 |
if audio_duration < 1.0:
|
421 |
+
logger.error(f"Duración audio voz ({audio_duration:.2f}s) es muy corta.")
|
422 |
raise ValueError("Generated voice audio is too short (min 1 second required).")
|
423 |
|
424 |
+
# 3. Extraer palabras clave
|
425 |
+
logger.info("Extrayendo palabras clave...")
|
426 |
try:
|
427 |
keywords = extract_visual_keywords_from_script(guion)
|
428 |
+
logger.info(f"Palabras clave identificadas: {keywords}")
|
429 |
except Exception as e:
|
430 |
+
logger.error(f"Error extrayendo keywords: {str(e)}", exc_info=True)
|
431 |
keywords = ["naturaleza", "paisaje"]
|
432 |
|
433 |
if not keywords:
|
434 |
keywords = ["video", "background"]
|
435 |
|
436 |
+
# 4. Buscar y descargar videos
|
437 |
+
logger.info("Buscando videos en Pexels...")
|
438 |
videos_data = []
|
439 |
total_desired_videos = 10
|
440 |
per_page_per_keyword = max(1, total_desired_videos // len(keywords))
|
|
|
445 |
videos = buscar_videos_pexels(keyword, PEXELS_API_KEY, per_page=per_page_per_keyword)
|
446 |
if videos:
|
447 |
videos_data.extend(videos)
|
448 |
+
logger.info(f"Encontrados {len(videos)} videos para '{keyword}'. Total data: {len(videos_data)}")
|
449 |
except Exception as e:
|
450 |
+
logger.warning(f"Error buscando videos para '{keyword}': {str(e)}")
|
451 |
|
452 |
if len(videos_data) < total_desired_videos / 2:
|
453 |
+
logger.warning(f"Pocos videos encontrados ({len(videos_data)}). Intentando con palabras clave genéricas.")
|
454 |
generic_keywords = ["nature", "city", "background", "abstract"]
|
455 |
for keyword in generic_keywords:
|
456 |
if len(videos_data) >= total_desired_videos: break
|
|
|
458 |
videos = buscar_videos_pexels(keyword, PEXELS_API_KEY, per_page=2)
|
459 |
if videos:
|
460 |
videos_data.extend(videos)
|
461 |
+
logger.info(f"Encontrados {len(videos)} videos para '{keyword}' (genérico). Total data: {len(videos_data)}")
|
462 |
except Exception as e:
|
463 |
+
logger.warning(f"Error buscando videos genéricos para '{keyword}': {str(e)}")
|
464 |
|
465 |
if not videos_data:
|
466 |
+
logger.error("No se encontraron videos en Pexels para ninguna palabra clave.")
|
467 |
+
raise ValueError("No se encontraron videos adecuados en Pexels.")
|
468 |
+
|
469 |
video_paths = []
|
470 |
+
logger.info(f"Intentando descargar {len(videos_data)} videos encontrados...")
|
471 |
for video in videos_data:
|
472 |
if 'video_files' not in video or not video['video_files']:
|
473 |
+
logger.debug(f"Saltando video sin archivos de video: {video.get('id')}")
|
474 |
continue
|
475 |
+
|
476 |
try:
|
477 |
best_quality = None
|
478 |
for vf in sorted(video['video_files'], key=lambda x: x.get('width', 0) * x.get('height', 0), reverse=True):
|
|
|
485 |
if path:
|
486 |
video_paths.append(path)
|
487 |
temp_intermediate_files.append(path)
|
488 |
+
logger.info(f"Video descargado OK desde {best_quality['link'][:50]}...")
|
489 |
else:
|
490 |
+
logger.warning(f"No se pudo descargar video desde {best_quality['link'][:50]}...")
|
491 |
else:
|
492 |
+
logger.warning(f"No se encontró enlace de descarga válido para video {video.get('id')}.")
|
493 |
|
494 |
except Exception as e:
|
495 |
+
logger.warning(f"Error procesando/descargando video {video.get('id')}: {str(e)}")
|
496 |
|
497 |
+
logger.info(f"Descargados {len(video_paths)} archivos de video utilizables.")
|
498 |
if not video_paths:
|
499 |
+
logger.error("No se pudo descargar ningún archivo de video utilizable.")
|
500 |
+
raise ValueError("No se pudo descargar ningún video utilizable de Pexels.")
|
501 |
+
|
502 |
+
# 5. Procesar y concatenar clips de video
|
503 |
+
logger.info("Procesando y concatenando videos descargados...")
|
504 |
current_duration = 0
|
505 |
min_clip_duration = 0.5
|
506 |
max_clip_segment = 10.0
|
507 |
|
508 |
for i, path in enumerate(video_paths):
|
509 |
if current_duration >= audio_duration + max_clip_segment:
|
510 |
+
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.")
|
511 |
break
|
512 |
|
513 |
clip = None
|
514 |
try:
|
515 |
+
logger.debug(f"[{i+1}/{len(video_paths)}] Abriendo clip: {path}")
|
516 |
clip = VideoFileClip(path)
|
517 |
source_clips.append(clip)
|
518 |
|
519 |
if clip.reader is None or clip.duration is None or clip.duration <= 0:
|
520 |
+
logger.warning(f"[{i+1}/{len(video_paths)}] Clip fuente {path} parece inválido (reader is None o duration <= 0). Saltando.")
|
521 |
continue
|
522 |
|
523 |
remaining_needed = audio_duration - current_duration
|
|
|
532 |
try:
|
533 |
sub = clip.subclip(0, segment_duration)
|
534 |
if sub.reader is None or sub.duration is None or sub.duration <= 0:
|
535 |
+
logger.warning(f"[{i+1}/{len(video_paths)}] Subclip generado de {path} es inválido. Saltando.")
|
536 |
try: sub.close()
|
537 |
except: pass
|
538 |
continue
|
539 |
+
|
540 |
clips_to_concatenate.append(sub)
|
541 |
+
current_duration += sub.duration
|
542 |
+
logger.debug(f"[{i+1}/{len(video_paths)}] Segmento añadido: {sub.duration:.1f}s (total video: {current_duration:.1f}/{audio_duration:.1f}s)")
|
543 |
+
|
544 |
except Exception as sub_e:
|
545 |
+
logger.warning(f"[{i+1}/{len(video_paths)}] Error creando subclip de {path} ({segment_duration:.1f}s): {str(sub_e)}")
|
546 |
continue
|
547 |
else:
|
548 |
+
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.")
|
549 |
else:
|
550 |
+
logger.debug(f"[{i+1}/{len(video_paths)}] Duración de video base ya alcanzada. Saltando clip.")
|
551 |
|
552 |
except Exception as e:
|
553 |
+
logger.warning(f"[{i+1}/{len(video_paths)}] Error procesando video {path}: {str(e)}", exc_info=True)
|
554 |
continue
|
555 |
|
556 |
+
logger.info(f"Procesamiento de clips fuente finalizado. Se obtuvieron {len(clips_to_concatenate)} segmentos válidos.")
|
557 |
|
558 |
if not clips_to_concatenate:
|
559 |
+
logger.error("No hay segmentos de video válidos disponibles para crear la secuencia.")
|
560 |
+
raise ValueError("No hay segmentos de video válidos disponibles para crear el video.")
|
561 |
+
|
562 |
+
logger.info(f"Concatenando {len(clips_to_concatenate)} segmentos de video.")
|
563 |
concatenated_base = None
|
564 |
try:
|
565 |
concatenated_base = concatenate_videoclips(clips_to_concatenate, method="chain")
|
566 |
+
logger.info(f"Duración video base después de concatenación inicial: {concatenated_base.duration:.2f}s")
|
567 |
|
568 |
if concatenated_base is None or concatenated_base.duration is None or concatenated_base.duration <= 0:
|
569 |
+
logger.critical("Video base concatenado es inválido después de la primera concatenación (None o duración cero).")
|
570 |
+
raise ValueError("Fallo al crear video base válido a partir de segmentos.")
|
571 |
|
572 |
except Exception as e:
|
573 |
+
logger.critical(f"Error durante la concatenación inicial: {str(e)}", exc_info=True)
|
574 |
+
raise ValueError("Fallo durante la concatenación de video inicial.")
|
575 |
finally:
|
576 |
for clip_segment in clips_to_concatenate:
|
577 |
try: clip_segment.close()
|
|
|
583 |
final_video_base = video_base
|
584 |
|
585 |
if final_video_base.duration < audio_duration:
|
586 |
+
logger.info(f"Video base ({final_video_base.duration:.2f}s) es más corto que el audio ({audio_duration:.2f}s). Repitiendo...")
|
587 |
+
|
588 |
num_full_repeats = int(audio_duration // final_video_base.duration)
|
589 |
remaining_duration = audio_duration % final_video_base.duration
|
590 |
+
|
591 |
repeated_clips_list = [final_video_base] * num_full_repeats
|
592 |
+
|
593 |
if remaining_duration > 0:
|
594 |
try:
|
595 |
remaining_clip = final_video_base.subclip(0, remaining_duration)
|
596 |
if remaining_clip is None or remaining_clip.duration is None or remaining_clip.duration <= 0:
|
597 |
+
logger.warning(f"Subclip generado para duración restante {remaining_duration:.2f}s es inválido. Saltando.")
|
598 |
try: remaining_clip.close()
|
599 |
except: pass
|
600 |
else:
|
601 |
repeated_clips_list.append(remaining_clip)
|
602 |
+
logger.debug(f"Añadiendo segmento restante: {remaining_duration:.2f}s")
|
603 |
|
604 |
except Exception as e:
|
605 |
+
logger.warning(f"Error creando subclip para duración restante {remaining_duration:.2f}s: {str(e)}")
|
606 |
|
607 |
if repeated_clips_list:
|
608 |
+
logger.info(f"Concatenando {len(repeated_clips_list)} partes para repetición.")
|
609 |
video_base_repeated = None
|
610 |
try:
|
611 |
video_base_repeated = concatenate_videoclips(repeated_clips_list, method="chain")
|
612 |
+
logger.info(f"Duración del video base repetido: {video_base_repeated.duration:.2f}s")
|
613 |
+
|
614 |
if video_base_repeated is None or video_base_repeated.duration is None or video_base_repeated.duration <= 0:
|
615 |
+
logger.critical("Video base repetido concatenado es inválido.")
|
616 |
+
raise ValueError("Fallo al crear video base repetido válido.")
|
617 |
|
618 |
if final_video_base is not video_base_repeated:
|
619 |
try: final_video_base.close()
|
620 |
except: pass
|
621 |
+
|
622 |
final_video_base = video_base_repeated
|
623 |
|
624 |
except Exception as e:
|
625 |
+
logger.critical(f"Error durante la concatenación de repetición: {str(e)}", exc_info=True)
|
626 |
+
raise ValueError("Fallo durante la repetición de video.")
|
627 |
finally:
|
628 |
if 'repeated_clips_list' in locals():
|
629 |
for clip in repeated_clips_list:
|
|
|
633 |
|
634 |
|
635 |
if final_video_base.duration > audio_duration:
|
636 |
+
logger.info(f"Recortando video base ({final_video_base.duration:.2f}s) para que coincida con la duración del audio ({audio_duration:.2f}s).")
|
637 |
trimmed_video_base = None
|
638 |
try:
|
639 |
trimmed_video_base = final_video_base.subclip(0, audio_duration)
|
640 |
if trimmed_video_base is None or trimmed_video_base.duration is None or trimmed_video_base.duration <= 0:
|
641 |
+
logger.critical("Video base recortado es inválido.")
|
642 |
+
raise ValueError("Fallo al crear video base recortado válido.")
|
643 |
|
644 |
if final_video_base is not trimmed_video_base:
|
645 |
try: final_video_base.close()
|
|
|
648 |
final_video_base = trimmed_video_base
|
649 |
|
650 |
except Exception as e:
|
651 |
+
logger.critical(f"Error durante el recorte: {str(e)}", exc_info=True)
|
652 |
+
raise ValueError("Fallo durante el recorte de video.")
|
653 |
|
654 |
|
655 |
if final_video_base is None or final_video_base.duration is None or final_video_base.duration <= 0:
|
656 |
+
logger.critical("Video base final es inválido antes de audio/escritura (None o duración cero).")
|
657 |
+
raise ValueError("Video base final es inválido.")
|
658 |
|
659 |
if final_video_base.size is None or final_video_base.size[0] <= 0 or final_video_base.size[1] <= 0:
|
660 |
+
logger.critical(f"Video base final tiene tamaño inválido: {final_video_base.size}. No se puede escribir video.")
|
661 |
+
raise ValueError("Video base final tiene tamaño inválido antes de escribir.")
|
662 |
|
663 |
video_base = final_video_base
|
664 |
|
665 |
+
# 6. Manejar música de fondo
|
666 |
+
logger.info("Procesando audio...")
|
667 |
+
|
668 |
final_audio = audio_tts_original
|
669 |
|
670 |
musica_audio_looped = None
|
|
|
675 |
music_path = os.path.join(temp_dir_intermediate, "musica_bg.mp3")
|
676 |
shutil.copyfile(musica_file, music_path)
|
677 |
temp_intermediate_files.append(music_path)
|
678 |
+
logger.info(f"Música de fondo copiada a: {music_path}")
|
679 |
+
|
680 |
musica_audio_original = AudioFileClip(music_path)
|
681 |
+
|
682 |
if musica_audio_original.reader is None or musica_audio_original.duration is None or musica_audio_original.duration <= 0:
|
683 |
+
logger.warning("Clip de música de fondo parece inválido o tiene duración cero. Saltando música.")
|
684 |
try: musica_audio_original.close()
|
685 |
except: pass
|
686 |
musica_audio_original = None
|
687 |
else:
|
688 |
+
# CAMBIO: Asegurar que loop_audio_to_length recibe la duración correcta del video base
|
689 |
musica_audio_looped = loop_audio_to_length(musica_audio_original, video_base.duration)
|
690 |
+
logger.debug(f"Música ajustada a duración del video: {musica_audio_looped.duration:.2f}s")
|
691 |
|
692 |
if musica_audio_looped is None or musica_audio_looped.duration is None or musica_audio_looped.duration <= 0:
|
693 |
+
logger.warning("Clip de música de fondo loopeado es inválido. Saltando música.")
|
694 |
try: musica_audio_looped.close()
|
695 |
except: pass
|
696 |
musica_audio_looped = None
|
697 |
|
698 |
|
699 |
if musica_audio_looped:
|
700 |
+
# CAMBIO: Usar volumex(0.2) para la música (ya estaba en 0.2, confirmamos)
|
701 |
+
# CAMBIO: Usar volumex(1.0) para la voz (ya estaba en 1.0, confirmamos)
|
702 |
composite_audio = CompositeAudioClip([
|
703 |
musica_audio_looped.volumex(0.2),
|
704 |
audio_tts_original.volumex(1.0)
|
705 |
])
|
706 |
+
|
707 |
if composite_audio.duration is None or composite_audio.duration <= 0:
|
708 |
+
logger.warning("Clip de audio compuesto es inválido (None o duración cero). Usando solo audio de voz.")
|
709 |
try: composite_audio.close()
|
710 |
except: pass
|
711 |
final_audio = audio_tts_original
|
712 |
else:
|
713 |
+
logger.info("Mezcla de audio completada (voz + música).")
|
714 |
final_audio = composite_audio
|
715 |
musica_audio = musica_audio_looped
|
716 |
|
717 |
except Exception as e:
|
718 |
+
logger.warning(f"Error procesando música de fondo: {str(e)}", exc_info=True)
|
719 |
final_audio = audio_tts_original
|
720 |
musica_audio = None
|
721 |
+
logger.warning("Usando solo audio de voz debido a un error con la música.")
|
722 |
|
723 |
|
724 |
if final_audio.duration is not None and abs(final_audio.duration - video_base.duration) > 0.2:
|
725 |
+
logger.warning(f"Duración del audio final ({final_audio.duration:.2f}s) difiere significativamente del video base ({video_base.duration:.2f}s). Intentando recorte.")
|
726 |
try:
|
727 |
if final_audio.duration > video_base.duration:
|
728 |
trimmed_final_audio = final_audio.subclip(0, video_base.duration)
|
729 |
if trimmed_final_audio.duration is None or trimmed_final_audio.duration <= 0:
|
730 |
+
logger.warning("Audio final recortado es inválido. Usando audio final original.")
|
731 |
try: trimmed_final_audio.close()
|
732 |
except: pass
|
733 |
else:
|
|
|
735 |
try: final_audio.close()
|
736 |
except: pass
|
737 |
final_audio = trimmed_final_audio
|
738 |
+
logger.warning("Audio final recortado para que coincida con la duración del video.")
|
739 |
except Exception as e:
|
740 |
+
logger.warning(f"Error ajustando duración del audio final: {str(e)}")
|
|
|
741 |
|
742 |
+
# 7. Crear video final
|
743 |
+
logger.info("Renderizando video final...")
|
744 |
video_final = video_base.set_audio(final_audio)
|
745 |
|
746 |
if video_final is None or video_final.duration is None or video_final.duration <= 0:
|
747 |
+
logger.critical("Clip de video final (con audio) es inválido antes de escribir (None o duración cero).")
|
748 |
+
raise ValueError("Clip de video final es inválido antes de escribir.")
|
|
|
749 |
|
750 |
output_filename = "final_video.mp4"
|
751 |
output_path = os.path.join(temp_dir_intermediate, output_filename)
|
752 |
+
logger.info(f"Escribiendo video final a: {output_path}")
|
753 |
|
754 |
video_final.write_videofile(
|
755 |
output_path,
|
|
|
760 |
preset="medium",
|
761 |
logger='bar'
|
762 |
)
|
763 |
+
|
764 |
total_time = (datetime.now() - start_time).total_seconds()
|
765 |
+
logger.info(f"PROCESO DE VIDEO FINALIZADO | Output: {output_path} | Tiempo total: {total_time:.2f}s")
|
766 |
+
|
767 |
return output_path
|
768 |
+
|
769 |
except ValueError as ve:
|
770 |
+
logger.error(f"ERROR CONTROLADO en crear_video: {str(ve)}")
|
771 |
raise ve
|
772 |
except Exception as e:
|
773 |
+
logger.critical(f"ERROR CRÍTICO NO CONTROLADO en crear_video: {str(e)}", exc_info=True)
|
774 |
raise e
|
775 |
finally:
|
776 |
+
logger.info("Iniciando limpieza de clips y archivos temporales intermedios...")
|
777 |
+
|
778 |
+
# Cerrar todos los clips de video fuente iniciales abiertos
|
779 |
for clip in source_clips:
|
780 |
try:
|
781 |
clip.close()
|
782 |
except Exception as e:
|
783 |
+
logger.warning(f"Error cerrando clip de video fuente en finally: {str(e)}")
|
|
|
|
|
|
|
|
|
|
|
|
|
784 |
|
785 |
+
# Cerrar cualquier segmento de video que quede en la lista (debería estar vacía si tuvo éxito)
|
786 |
+
for clip_segment in clips_to_concatenate:
|
|
|
|
|
|
|
|
|
|
|
|
|
787 |
try:
|
788 |
+
clip_segment.close()
|
789 |
except Exception as e:
|
790 |
+
logger.warning(f"Error cerrando segmento de video en finally: {str(e)}")
|
791 |
|
792 |
+
# Cerrar los objetos principales de MoviePy si se crearon correctamente
|
793 |
+
try:
|
794 |
+
# Cerrar clips de audio en orden: música loopeada, música original (si es diferente), TTS original
|
795 |
+
if musica_audio is not None: # musica_audio holds the potentially looped clip
|
796 |
+
try:
|
797 |
+
musica_audio.close()
|
798 |
+
except Exception as e:
|
799 |
+
logger.warning(f"Error cerrando musica_audio (procesada) en finally: {str(e)}")
|
800 |
|
801 |
+
if musica_audio_original is not None and musica_audio_original is not musica_audio:
|
802 |
+
try:
|
803 |
+
musica_audio_original.close()
|
804 |
+
except Exception as e:
|
805 |
+
logger.warning(f"Error cerrando musica_audio_original en finally: {str(e)}")
|
806 |
+
|
807 |
+
# audio_tts currently only holds audio_tts_original, but keep structure
|
808 |
+
if audio_tts is not None and audio_tts is not audio_tts_original:
|
809 |
+
try:
|
810 |
+
audio_tts.close()
|
811 |
+
except Exception as e:
|
812 |
+
logger.warning(f"Error cerrando audio_tts (procesada) en finally: {str(e)}")
|
813 |
|
814 |
+
if audio_tts_original is not None:
|
815 |
+
try:
|
816 |
+
audio_tts_original.close()
|
817 |
+
except Exception as e:
|
818 |
+
logger.warning(f"Error cerrando audio_tts_original en finally: {str(e)}")
|
819 |
+
|
820 |
+
|
821 |
+
# Cerrar clips de video en orden: video_final (debería cerrar sus componentes), luego video_base (si es diferente de video_final)
|
822 |
+
if video_final is not None:
|
823 |
+
try:
|
824 |
+
video_final.close()
|
825 |
+
except Exception as e:
|
826 |
+
logger.warning(f"Error cerrando video_final en finally: {str(e)}")
|
827 |
+
elif video_base is not None and video_base is not video_final: # Asegurarse de que video_base no es el mismo objeto que video_final
|
828 |
+
try:
|
829 |
+
video_base.close()
|
830 |
+
except Exception as e:
|
831 |
+
logger.warning(f"Error cerrando video_base en finally: {str(e)}")
|
832 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
833 |
|
834 |
+
except Exception as e:
|
835 |
+
logger.warning(f"Error durante el cierre de clips finales en finally: {str(e)}")
|
836 |
|
837 |
+
# Limpiar archivos intermedios, pero NO el archivo de video final
|
838 |
if temp_dir_intermediate and os.path.exists(temp_dir_intermediate):
|
839 |
final_output_in_temp = os.path.join(temp_dir_intermediate, "final_video.mp4")
|
840 |
+
|
841 |
for path in temp_intermediate_files:
|
842 |
try:
|
843 |
+
# Verificar explícitamente que la ruta no sea la ruta de salida del video final antes de eliminar
|
844 |
if os.path.isfile(path) and path != final_output_in_temp:
|
845 |
+
logger.debug(f"Eliminando archivo temporal intermedio: {path}")
|
846 |
os.remove(path)
|
847 |
+
elif os.path.isfile(path) and path == final_output_in_temp:
|
848 |
+
logger.debug(f"Saltando eliminación del archivo de video final: {path}")
|
849 |
except Exception as e:
|
850 |
+
logger.warning(f"No se pudo eliminar archivo temporal intermedio {path}: {str(e)}")
|
|
|
|
|
851 |
|
852 |
+
# El directorio temporal *persistirá* porque contiene el archivo final
|
853 |
+
logger.info(f"Directorio temporal intermedio {temp_dir_intermediate} persistirá para que Gradio lea el video final.")
|
854 |
|
855 |
+
|
856 |
+
def run_app(prompt_type, input_text, musica_file):
|
857 |
logger.info("="*80)
|
858 |
+
logger.info("SOLICITUD RECIBIDA EN INTERFAZ")
|
859 |
+
|
860 |
input_text = prompt_ia if prompt_type == "Generar Guion con IA" else prompt_manual
|
861 |
+
|
862 |
output_video = None
|
863 |
+
output_file = None # Inicializar a None
|
864 |
status_msg = gr.update(value="⏳ Procesando...", interactive=False)
|
865 |
|
866 |
if not input_text or not input_text.strip():
|
867 |
+
logger.warning("Texto de entrada vacío.")
|
868 |
return None, None, gr.update(value="⚠️ Por favor, ingresa texto para el guion o el tema.", interactive=False)
|
869 |
|
870 |
+
logger.info(f"Tipo de entrada: {prompt_type}")
|
871 |
+
logger.debug(f"Texto de entrada: '{input_text[:100]}...'")
|
872 |
if musica_file:
|
873 |
+
logger.info(f"Archivo de música recibido: {musica_file}")
|
874 |
else:
|
875 |
+
logger.info("No se proporcionó archivo de música.")
|
876 |
|
877 |
try:
|
878 |
+
logger.info("Llamando a crear_video...")
|
879 |
video_path = crear_video(prompt_type, input_text, musica_file)
|
880 |
+
|
881 |
if video_path and os.path.exists(video_path):
|
882 |
+
logger.info(f"crear_video retornó path: {video_path}")
|
883 |
+
logger.info(f"Tamaño del archivo de video retornado: {os.path.getsize(video_path)} bytes")
|
884 |
+
output_video = video_path # Establecer valor del componente de video
|
885 |
+
output_file = video_path # Establecer valor del componente de archivo para descarga
|
886 |
status_msg = gr.update(value="✅ Video generado exitosamente.", interactive=False)
|
887 |
else:
|
888 |
+
logger.error(f"crear_video no retornó un path válido o el archivo no existe: {video_path}")
|
889 |
+
# Dejar las salidas de video y archivo como None
|
890 |
status_msg = gr.update(value="❌ Error: La generación del video falló o el archivo no se creó correctamente.", interactive=False)
|
891 |
|
892 |
except ValueError as ve:
|
893 |
+
logger.warning(f"Error de validación durante la creación del video: {str(ve)}")
|
894 |
+
# Dejar las salidas de video y archivo como None
|
895 |
status_msg = gr.update(value=f"⚠️ Error de validación: {str(ve)}", interactive=False)
|
896 |
except Exception as e:
|
897 |
+
logger.critical(f"Error crítico durante la creación del video: {str(e)}", exc_info=True)
|
898 |
+
# Dejar las salidas de video y archivo como None
|
899 |
status_msg = gr.update(value=f"❌ Error inesperado: {str(e)}", interactive=False)
|
900 |
finally:
|
901 |
+
logger.info("Fin del handler run_app.")
|
902 |
+
# Retornar las tres salidas
|
903 |
return output_video, output_file, status_msg
|
904 |
|
905 |
|
906 |
+
# Interfaz de Gradio
|
907 |
with gr.Blocks(title="Generador de Videos con IA", theme=gr.themes.Soft(), css="""
|
908 |
.gradio-container {max-width: 800px; margin: auto;}
|
909 |
h1 {text-align: center;}
|
910 |
""") as app:
|
911 |
+
|
912 |
+
gr.Markdown("# 🎬 Generador Automático de Videos con IA")
|
913 |
+
gr.Markdown("Genera videos cortos a partir de un tema o guion, usando imágenes de archivo de Pexels y voz generada.")
|
914 |
+
|
915 |
with gr.Row():
|
916 |
with gr.Column():
|
917 |
prompt_type = gr.Radio(
|
918 |
+
["Generar Guion con IA", "Usar Mi Guion"],
|
919 |
+
label="Método de Entrada",
|
920 |
value="Generar Guion con IA"
|
921 |
)
|
922 |
+
|
923 |
with gr.Column(visible=True) as ia_guion_column:
|
924 |
prompt_ia = gr.Textbox(
|
925 |
+
label="Tema para IA",
|
926 |
lines=2,
|
927 |
+
placeholder="Ej: Un paisaje natural con montañas y ríos al amanecer, mostrando la belleza de la naturaleza...",
|
928 |
max_lines=4,
|
929 |
value=""
|
930 |
)
|
931 |
+
|
932 |
with gr.Column(visible=False) as manual_guion_column:
|
933 |
prompt_manual = gr.Textbox(
|
934 |
+
label="Tu Guion Completo",
|
935 |
lines=5,
|
936 |
+
placeholder="Ej: En este video exploraremos los misterios del océano. Veremos la vida marina fascinante y los arrecifes de coral vibrantes. ¡Acompáñanos en esta aventura subacuática!",
|
937 |
max_lines=10,
|
938 |
value=""
|
939 |
)
|
940 |
+
|
941 |
musica_input = gr.Audio(
|
942 |
+
label="Música de fondo (opcional)",
|
943 |
type="filepath",
|
944 |
interactive=True,
|
945 |
value=None
|
946 |
)
|
947 |
+
|
948 |
+
generate_btn = gr.Button("✨ Generar Video", variant="primary")
|
949 |
+
|
950 |
with gr.Column():
|
951 |
video_output = gr.Video(
|
952 |
+
label="Previsualización del Video Generado", # Etiqueta cambiada
|
953 |
interactive=False,
|
954 |
height=400
|
955 |
)
|
956 |
+
# Añadir el componente File para la descarga
|
957 |
file_output = gr.File(
|
958 |
+
label="Descargar Archivo de Video", # Etiqueta cambiada
|
959 |
interactive=False,
|
960 |
+
visible=False # Ocultar inicialmente
|
961 |
)
|
962 |
status_output = gr.Textbox(
|
963 |
+
label="Estado",
|
964 |
interactive=False,
|
965 |
show_label=False,
|
966 |
+
placeholder="Esperando acción...",
|
967 |
+
value="Esperando entrada..."
|
968 |
)
|
969 |
|
970 |
prompt_type.change(
|
971 |
+
lambda x: (gr.update(visible=x == "Generar Guion con IA"),
|
972 |
gr.update(visible=x == "Usar Mi Guion")),
|
973 |
inputs=prompt_type,
|
974 |
outputs=[ia_guion_column, manual_guion_column]
|
975 |
)
|
976 |
|
977 |
+
# Modificar el evento click para retornar 3 salidas
|
978 |
generate_btn.click(
|
979 |
+
# Acción 1: Resetear salidas y establecer estado a procesando
|
980 |
+
lambda: (None, None, gr.update(value="⏳ Procesando... Esto puede tomar 2-5 minutos o más para videos largos.", interactive=False)), # Mensaje de estado mejorado
|
981 |
outputs=[video_output, file_output, status_output],
|
982 |
+
queue=True, # Mantener la cola habilitada
|
983 |
).then(
|
984 |
+
# Acción 2: Llamar a la función principal de procesamiento
|
985 |
run_app,
|
986 |
inputs=[prompt_type, prompt_ia, prompt_manual, musica_input],
|
987 |
+
outputs=[video_output, file_output, status_output] # Coincidir las 3 salidas
|
988 |
).then(
|
989 |
+
# Acción 3: Hacer visible el enlace de descarga si se retornó un archivo
|
990 |
+
# Verificar si file_output tiene un valor
|
991 |
+
lambda video_path, file_path: gr.update(visible=file_path is not None), # Usar file_path para visibilidad
|
992 |
+
inputs=[video_output, file_output], # Ambas salidas como entrada
|
993 |
+
outputs=[file_output] # Actualizar visibilidad de file_output
|
994 |
)
|
995 |
|
996 |
|
997 |
+
gr.Markdown("### Instrucciones:")
|
998 |
gr.Markdown("""
|
999 |
+
1. **Clave API de Pexels:** Asegúrate de haber configurado la variable de entorno `PEXELS_API_KEY` con tu clave.
|
1000 |
+
2. **Selecciona el tipo de entrada**:
|
1001 |
+
- "Generar Guion con IA": Describe brevemente un tema (ej. "La belleza de las montañas"). La IA generará un guion corto.
|
1002 |
+
- "Usar Mi Guion": Escribe el guion completo que quieres para el video.
|
1003 |
+
3. **Sube música** (opcional): Selecciona un archivo de audio (MP3, WAV, etc.) para usar como música de fondo.
|
1004 |
+
4. **Haz clic en "✨ Generar Video"**.
|
1005 |
+
5. Espera a que se procese el video. El tiempo de espera puede variar. Verás el estado en el cuadro de texto.
|
1006 |
+
6. La previsualización del video aparecerá arriba (puede fallar para archivos grandes), y un enlace **Descargar Archivo de Video** se mostrará si la generación fue exitosa.
|
1007 |
+
7. Si hay errores, revisa el log `video_generator_full.log` para más detalles.
|
1008 |
""")
|
1009 |
gr.Markdown("---")
|
1010 |
+
gr.Markdown("Desarrollado por [Tu Nombre/Empresa/Alias - Opcional]")
|
1011 |
|
1012 |
if __name__ == "__main__":
|
1013 |
+
logger.info("Verificando dependencias críticas...")
|
1014 |
try:
|
1015 |
from moviepy.editor import ColorClip
|
1016 |
try:
|
1017 |
temp_clip = ColorClip((100,100), color=(255,0,0), duration=0.1)
|
1018 |
temp_clip.close()
|
1019 |
+
logger.info("Clips base de MoviePy (como ColorClip) creados y cerrados exitosamente. FFmpeg parece accesible.")
|
1020 |
except Exception as e:
|
1021 |
+
logger.critical(f"Fallo al crear clip base de MoviePy. A menudo indica problemas con FFmpeg/ImageMagick. Error: {e}", exc_info=True)
|
1022 |
|
1023 |
except Exception as e:
|
1024 |
+
logger.critical(f"Fallo al importar MoviePy. Asegúrate de que está instalado. Error: {e}", exc_info=True)
|
1025 |
|
1026 |
+
logger.info("Iniciando aplicación Gradio...")
|
1027 |
try:
|
1028 |
+
# Gradio Queue maneja tareas largas, no es necesario un ajuste global de timeout aquí.
|
1029 |
+
# El timeout se gestiona por solicitud o por el límite del worker de la cola.
|
1030 |
app.launch(server_name="0.0.0.0", server_port=7860, share=False)
|
1031 |
except Exception as e:
|
1032 |
+
logger.critical(f"No se pudo iniciar la app: {str(e)}", exc_info=True)
|
1033 |
raise
|