Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -17,15 +17,52 @@ from collections import Counter
|
|
| 17 |
import shutil
|
| 18 |
import json
|
| 19 |
|
| 20 |
-
# Configuración de logging
|
| 21 |
-
logging.basicConfig(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
logger = logging.getLogger(__name__)
|
|
|
|
|
|
|
|
|
|
| 23 |
|
| 24 |
-
# Clave API de Pexels
|
| 25 |
PEXELS_API_KEY = os.environ.get("PEXELS_API_KEY")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
|
| 27 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
def buscar_videos_pexels(query, api_key, per_page=5):
|
|
|
|
| 29 |
headers = {"Authorization": api_key}
|
| 30 |
try:
|
| 31 |
params = {
|
|
@@ -35,6 +72,7 @@ def buscar_videos_pexels(query, api_key, per_page=5):
|
|
| 35 |
"size": "medium"
|
| 36 |
}
|
| 37 |
|
|
|
|
| 38 |
response = requests.get(
|
| 39 |
"https://api.pexels.com/videos/search",
|
| 40 |
headers=headers,
|
|
@@ -43,52 +81,33 @@ def buscar_videos_pexels(query, api_key, per_page=5):
|
|
| 43 |
)
|
| 44 |
response.raise_for_status()
|
| 45 |
|
| 46 |
-
# Intentar parsear la respuesta
|
| 47 |
try:
|
| 48 |
data = response.json()
|
|
|
|
| 49 |
return data.get('videos', [])
|
| 50 |
except json.JSONDecodeError:
|
| 51 |
-
logger.error("
|
| 52 |
return []
|
| 53 |
|
| 54 |
except requests.exceptions.RequestException as e:
|
| 55 |
-
logger.error(f"Error de conexión
|
| 56 |
except Exception as e:
|
| 57 |
-
logger.error(f"Error inesperado: {e}")
|
| 58 |
|
| 59 |
return []
|
| 60 |
|
| 61 |
-
#
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
tokenizer = GPT2Tokenizer.from_pretrained(MODEL_NAME)
|
| 65 |
-
model = GPT2LMHeadModel.from_pretrained(MODEL_NAME).eval()
|
| 66 |
-
if tokenizer.pad_token is None:
|
| 67 |
-
tokenizer.pad_token = tokenizer.eos_token
|
| 68 |
-
logger.info("Modelo GPT-2 en español cargado")
|
| 69 |
-
except Exception as e:
|
| 70 |
-
logger.error(f"Error al cargar modelo GPT-2: {e}")
|
| 71 |
-
tokenizer = model = None
|
| 72 |
-
|
| 73 |
-
try:
|
| 74 |
-
kw_model = KeyBERT('distilbert-base-multilingual-cased')
|
| 75 |
-
logger.info("KeyBERT cargado")
|
| 76 |
-
except Exception as e:
|
| 77 |
-
logger.error(f"Error al cargar KeyBERT: {e}")
|
| 78 |
-
kw_model = None
|
| 79 |
-
|
| 80 |
-
# Función mejorada para generar guiones
|
| 81 |
-
def generate_script(prompt, max_length=150):
|
| 82 |
if not tokenizer or not model:
|
| 83 |
-
|
|
|
|
| 84 |
|
| 85 |
try:
|
| 86 |
-
# Prompt mejorado con instrucciones claras
|
| 87 |
enhanced_prompt = f"Escribe un guion corto y coherente sobre: {prompt}"
|
| 88 |
-
|
| 89 |
inputs = tokenizer(enhanced_prompt, return_tensors="pt", truncation=True, max_length=512)
|
| 90 |
|
| 91 |
-
|
| 92 |
outputs = model.generate(
|
| 93 |
**inputs,
|
| 94 |
max_length=max_length,
|
|
@@ -100,63 +119,71 @@ def generate_script(prompt, max_length=150):
|
|
| 100 |
pad_token_id=tokenizer.pad_token_id,
|
| 101 |
eos_token_id=tokenizer.eos_token_id
|
| 102 |
)
|
| 103 |
-
text = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 104 |
|
| 105 |
-
|
| 106 |
-
text = re.sub(r'<[^>]+>', '', text)
|
| 107 |
sentences = text.split('.')
|
|
|
|
| 108 |
if sentences:
|
| 109 |
-
|
|
|
|
|
|
|
| 110 |
return text
|
| 111 |
except Exception as e:
|
| 112 |
-
logger.error(f"Error generando guion: {e}")
|
| 113 |
-
return prompt
|
| 114 |
|
| 115 |
-
#
|
| 116 |
async def text_to_speech(text, output_path, voice="es-ES-ElviraNeural"):
|
|
|
|
| 117 |
try:
|
| 118 |
communicate = edge_tts.Communicate(text, voice)
|
| 119 |
await communicate.save(output_path)
|
|
|
|
| 120 |
return True
|
| 121 |
except Exception as e:
|
| 122 |
-
logger.error(f"Error en TTS: {e}")
|
| 123 |
return False
|
| 124 |
|
| 125 |
-
#
|
| 126 |
-
def download_video_file(url, temp_dir):
|
| 127 |
if not url:
|
|
|
|
| 128 |
return None
|
| 129 |
|
| 130 |
try:
|
| 131 |
-
|
| 132 |
-
response.raise_for_status()
|
| 133 |
-
|
| 134 |
file_name = f"video_{datetime.now().strftime('%H%M%S%f')}.mp4"
|
| 135 |
output_path = os.path.join(temp_dir, file_name)
|
| 136 |
|
| 137 |
-
with
|
| 138 |
-
|
| 139 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 140 |
return output_path
|
| 141 |
except Exception as e:
|
| 142 |
-
logger.error(f"Error descargando video: {e}")
|
| 143 |
return None
|
| 144 |
|
| 145 |
-
#
|
| 146 |
def loop_audio_to_length(audio_clip, target_duration):
|
|
|
|
| 147 |
if audio_clip.duration >= target_duration:
|
| 148 |
return audio_clip.subclip(0, target_duration)
|
| 149 |
|
| 150 |
-
loops =
|
|
|
|
| 151 |
audios = [audio_clip] * loops
|
| 152 |
return concatenate_videoclips(audios).subclip(0, target_duration)
|
| 153 |
|
| 154 |
-
#
|
| 155 |
def extract_visual_keywords_from_script(script_text):
|
| 156 |
-
|
| 157 |
clean_text = re.sub(r'[^\w\sáéíóúñ]', '', script_text.lower())
|
| 158 |
|
| 159 |
-
# Método 1: KeyBERT si está disponible
|
| 160 |
if kw_model:
|
| 161 |
try:
|
| 162 |
keywords = kw_model.extract_keywords(
|
|
@@ -166,64 +193,74 @@ def extract_visual_keywords_from_script(script_text):
|
|
| 166 |
top_n=3
|
| 167 |
)
|
| 168 |
if keywords:
|
|
|
|
| 169 |
return [kw[0].replace(" ", "+") for kw in keywords]
|
| 170 |
except Exception as e:
|
| 171 |
-
logger.warning(f"KeyBERT falló: {e}")
|
| 172 |
|
| 173 |
-
# Método 2: Frecuencia de palabras (fallback)
|
| 174 |
words = clean_text.split()
|
| 175 |
stop_words = {"el", "la", "los", "las", "de", "en", "y", "a", "que", "es", "un", "una", "con"}
|
| 176 |
keywords = [word for word in words if len(word) > 3 and word not in stop_words]
|
| 177 |
|
| 178 |
-
# Si aún no hay palabras clave, usar palabras predeterminadas
|
| 179 |
if not keywords:
|
| 180 |
logger.warning("Usando palabras clave predeterminadas")
|
| 181 |
return ["naturaleza", "ciudad", "paisaje"]
|
| 182 |
|
| 183 |
-
# Contar frecuencia y seleccionar las 3 más comunes
|
| 184 |
word_counts = Counter(keywords)
|
| 185 |
-
|
|
|
|
|
|
|
| 186 |
|
| 187 |
-
#
|
| 188 |
def crear_video(prompt_type, input_text, musica_file=None):
|
| 189 |
-
logger.info(
|
|
|
|
|
|
|
| 190 |
|
| 191 |
# 1. Generar o usar guion
|
|
|
|
| 192 |
if prompt_type == "Generar Guion con IA":
|
|
|
|
| 193 |
guion = generate_script(input_text)
|
| 194 |
else:
|
|
|
|
| 195 |
guion = input_text
|
| 196 |
|
| 197 |
-
logger.info(f"Guion: {guion[:100]}...")
|
| 198 |
|
| 199 |
-
# Validar guion
|
| 200 |
if not guion.strip():
|
|
|
|
| 201 |
raise ValueError("El guion está vacío")
|
| 202 |
|
| 203 |
# Directorio temporal
|
| 204 |
temp_dir = tempfile.mkdtemp()
|
|
|
|
| 205 |
temp_files = []
|
| 206 |
|
| 207 |
try:
|
| 208 |
# 2. Generar audio de voz
|
|
|
|
| 209 |
voz_path = os.path.join(temp_dir, "voz.mp3")
|
| 210 |
if not asyncio.run(text_to_speech(guion, voz_path)):
|
|
|
|
| 211 |
raise ValueError("Error generando voz")
|
| 212 |
temp_files.append(voz_path)
|
| 213 |
|
| 214 |
audio_tts = AudioFileClip(voz_path)
|
| 215 |
audio_duration = audio_tts.duration
|
|
|
|
| 216 |
|
| 217 |
-
# 3. Extraer palabras clave
|
|
|
|
| 218 |
try:
|
| 219 |
keywords = extract_visual_keywords_from_script(guion)
|
|
|
|
| 220 |
except Exception as e:
|
| 221 |
-
logger.error(f"Error extrayendo
|
| 222 |
-
keywords = ["naturaleza", "paisaje"]
|
| 223 |
-
|
| 224 |
-
logger.info(f"Palabras clave: {keywords}")
|
| 225 |
|
| 226 |
-
# 4. Buscar y descargar videos
|
|
|
|
| 227 |
videos_data = []
|
| 228 |
for keyword in keywords:
|
| 229 |
try:
|
|
@@ -232,17 +269,17 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
| 232 |
videos_data.extend(videos)
|
| 233 |
logger.info(f"Encontrados {len(videos)} videos para '{keyword}'")
|
| 234 |
except Exception as e:
|
| 235 |
-
logger.warning(f"Error buscando videos para '{keyword}': {e}")
|
| 236 |
|
| 237 |
-
# Si no encontramos videos, intentar con palabras clave genéricas
|
| 238 |
if not videos_data:
|
| 239 |
-
logger.warning("Usando palabras clave genéricas
|
| 240 |
for keyword in ["naturaleza", "ciudad", "paisaje"]:
|
| 241 |
videos = buscar_videos_pexels(keyword, PEXELS_API_KEY, per_page=3)
|
| 242 |
if videos:
|
| 243 |
videos_data.extend(videos)
|
| 244 |
|
| 245 |
if not videos_data:
|
|
|
|
| 246 |
raise ValueError("No se encontraron videos en Pexels para ninguna palabra clave")
|
| 247 |
|
| 248 |
video_paths = []
|
|
@@ -251,7 +288,6 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
| 251 |
continue
|
| 252 |
|
| 253 |
try:
|
| 254 |
-
# Seleccionar la mejor calidad de video
|
| 255 |
best_quality = max(
|
| 256 |
video['video_files'],
|
| 257 |
key=lambda x: x.get('width', 0) * x.get('height', 0)
|
|
@@ -264,12 +300,14 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
| 264 |
temp_files.append(path)
|
| 265 |
logger.info(f"Video descargado: {best_quality['link']}")
|
| 266 |
except Exception as e:
|
| 267 |
-
logger.warning(f"Error procesando video: {e}")
|
| 268 |
|
| 269 |
if not video_paths:
|
|
|
|
| 270 |
raise ValueError("No se pudo descargar ningún video")
|
| 271 |
|
| 272 |
# 5. Procesar videos
|
|
|
|
| 273 |
clips = []
|
| 274 |
current_duration = 0
|
| 275 |
|
|
@@ -281,54 +319,56 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
| 281 |
clip = VideoFileClip(path)
|
| 282 |
usable_duration = min(clip.duration, 10)
|
| 283 |
|
| 284 |
-
if usable_duration > 1:
|
| 285 |
clips.append(clip.subclip(0, usable_duration))
|
| 286 |
current_duration += usable_duration
|
| 287 |
-
logger.
|
| 288 |
except Exception as e:
|
| 289 |
-
logger.warning(f"Error procesando video {path}: {e}")
|
| 290 |
|
| 291 |
if not clips:
|
|
|
|
| 292 |
raise ValueError("No hay clips válidos para crear el video")
|
| 293 |
|
| 294 |
video_base = concatenate_videoclips(clips, method="compose")
|
|
|
|
| 295 |
|
| 296 |
-
# Ajustar duración del video si es necesario
|
| 297 |
if video_base.duration < audio_duration:
|
| 298 |
-
num_repeats =
|
| 299 |
-
|
| 300 |
-
video_base = concatenate_videoclips(
|
| 301 |
|
| 302 |
# 6. Manejar música de fondo
|
|
|
|
| 303 |
final_audio = audio_tts
|
| 304 |
|
| 305 |
if musica_file:
|
| 306 |
try:
|
| 307 |
-
# Copiar archivo de música
|
| 308 |
music_path = os.path.join(temp_dir, "musica.mp3")
|
| 309 |
shutil.copyfile(musica_file, music_path)
|
| 310 |
temp_files.append(music_path)
|
|
|
|
| 311 |
|
| 312 |
-
# Procesar música
|
| 313 |
musica_audio = AudioFileClip(music_path)
|
|
|
|
| 314 |
|
| 315 |
-
# Crear loop si es necesario
|
| 316 |
if musica_audio.duration < audio_duration:
|
| 317 |
musica_audio = loop_audio_to_length(musica_audio, audio_duration)
|
|
|
|
| 318 |
|
| 319 |
-
# Mezclar con el audio de voz
|
| 320 |
final_audio = CompositeAudioClip([
|
| 321 |
-
musica_audio.volumex(0.3),
|
| 322 |
-
audio_tts.volumex(1.0)
|
| 323 |
])
|
| 324 |
-
logger.info("
|
| 325 |
except Exception as e:
|
| 326 |
-
logger.warning(f"Error procesando música: {e}")
|
| 327 |
|
| 328 |
# 7. Crear video final
|
|
|
|
| 329 |
video_final = video_base.set_audio(final_audio)
|
| 330 |
-
|
| 331 |
output_path = os.path.join(temp_dir, "final_video.mp4")
|
|
|
|
| 332 |
video_final.write_videofile(
|
| 333 |
output_path,
|
| 334 |
fps=24,
|
|
@@ -339,39 +379,50 @@ def crear_video(prompt_type, input_text, musica_file=None):
|
|
| 339 |
logger=None
|
| 340 |
)
|
| 341 |
|
| 342 |
-
|
|
|
|
| 343 |
return output_path
|
| 344 |
|
| 345 |
except Exception as e:
|
| 346 |
-
logger.error(f"
|
| 347 |
raise
|
| 348 |
finally:
|
| 349 |
-
|
| 350 |
for path in temp_files:
|
| 351 |
try:
|
| 352 |
if os.path.isfile(path):
|
| 353 |
os.remove(path)
|
| 354 |
-
except:
|
| 355 |
-
|
| 356 |
-
|
|
|
|
| 357 |
shutil.rmtree(temp_dir, ignore_errors=True)
|
|
|
|
|
|
|
| 358 |
|
| 359 |
-
#
|
| 360 |
def run_app(prompt_type, prompt_ia, prompt_manual, musica_file):
|
|
|
|
|
|
|
| 361 |
input_text = prompt_ia if prompt_type == "Generar Guion con IA" else prompt_manual
|
| 362 |
|
| 363 |
if not input_text.strip():
|
|
|
|
| 364 |
return None, "Por favor ingresa texto"
|
| 365 |
|
| 366 |
try:
|
|
|
|
| 367 |
video_path = crear_video(prompt_type, input_text, musica_file)
|
|
|
|
| 368 |
return video_path, "✅ Video generado exitosamente"
|
| 369 |
except ValueError as ve:
|
|
|
|
| 370 |
return None, f"⚠️ {ve}"
|
| 371 |
except Exception as e:
|
|
|
|
| 372 |
return None, f"❌ Error: {str(e)}"
|
| 373 |
|
| 374 |
-
#
|
| 375 |
with gr.Blocks(title="Generador de Videos con IA", theme=gr.themes.Soft(), css="""
|
| 376 |
.gradio-container {max-width: 800px; margin: auto;}
|
| 377 |
h1 {text-align: center;}
|
|
@@ -424,7 +475,6 @@ with gr.Blocks(title="Generador de Videos con IA", theme=gr.themes.Soft(), css="
|
|
| 424 |
placeholder="Esperando acción..."
|
| 425 |
)
|
| 426 |
|
| 427 |
-
# Manejar visibilidad de columnas
|
| 428 |
prompt_type.change(
|
| 429 |
lambda x: (gr.update(visible=x == "Generar Guion con IA"),
|
| 430 |
gr.update(visible=x == "Usar Mi Guion")),
|
|
@@ -432,7 +482,6 @@ with gr.Blocks(title="Generador de Videos con IA", theme=gr.themes.Soft(), css="
|
|
| 432 |
outputs=[ia_guion_column, manual_guion_column]
|
| 433 |
)
|
| 434 |
|
| 435 |
-
# Lógica de generación
|
| 436 |
generate_btn.click(
|
| 437 |
lambda: (None, "⏳ Procesando... (esto puede tomar 2-5 minutos)"),
|
| 438 |
outputs=[video_output, status_output],
|
|
@@ -449,9 +498,14 @@ with gr.Blocks(title="Generador de Videos con IA", theme=gr.themes.Soft(), css="
|
|
| 449 |
- "Generar Guion con IA": Describe un tema
|
| 450 |
- "Usar Mi Guion": Escribe tu guion completo
|
| 451 |
2. **Sube música** (opcional): Selecciona un archivo de audio
|
| 452 |
-
3. **Haz clic en Generar Video
|
| 453 |
4. Espera a que se procese el video (puede tomar varios minutos)
|
| 454 |
""")
|
| 455 |
|
| 456 |
if __name__ == "__main__":
|
| 457 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
import shutil
|
| 18 |
import json
|
| 19 |
|
| 20 |
+
# Configuración de logging MEJORADA
|
| 21 |
+
logging.basicConfig(
|
| 22 |
+
level=logging.DEBUG,
|
| 23 |
+
format='%(asctime)s - %(levelname)s - %(message)s',
|
| 24 |
+
handlers=[
|
| 25 |
+
logging.StreamHandler(),
|
| 26 |
+
logging.FileHandler('video_generator_full.log', encoding='utf-8')
|
| 27 |
+
]
|
| 28 |
+
)
|
| 29 |
logger = logging.getLogger(__name__)
|
| 30 |
+
logger.info("="*80)
|
| 31 |
+
logger.info("INICIO DE EJECUCIÓN - GENERADOR DE VIDEOS")
|
| 32 |
+
logger.info("="*80)
|
| 33 |
|
| 34 |
+
# Clave API de Pexels (configuración segura)
|
| 35 |
PEXELS_API_KEY = os.environ.get("PEXELS_API_KEY")
|
| 36 |
+
if not PEXELS_API_KEY:
|
| 37 |
+
logger.critical("NO SE ENCONTRÓ PEXELS_API_KEY EN VARIABLES DE ENTORNO")
|
| 38 |
+
raise ValueError("API key de Pexels no configurada")
|
| 39 |
+
else:
|
| 40 |
+
logger.info("API key de Pexels configurada correctamente")
|
| 41 |
|
| 42 |
+
# Inicialización de modelos CON LOGS DETALLADOS
|
| 43 |
+
MODEL_NAME = "datificate/gpt2-small-spanish"
|
| 44 |
+
logger.info(f"Inicializando modelo GPT-2: {MODEL_NAME}")
|
| 45 |
+
try:
|
| 46 |
+
tokenizer = GPTizer.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"Modelo GPT-2 cargado | Vocabulario: {len(tokenizer)} tokens")
|
| 51 |
+
except Exception as e:
|
| 52 |
+
logger.error(f"FALLA CRÍTICA al cargar GPT-2: {str(e)}", exc_info=True)
|
| 53 |
+
tokenizer = model = None
|
| 54 |
+
|
| 55 |
+
logger.info("Cargando modelo KeyBERT...")
|
| 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 |
+
# [FUNCIÓN BUSCAR_VIDEOS_PEXELS ORIGINAL CON LOGS AÑADIDOS]
|
| 64 |
def buscar_videos_pexels(query, api_key, per_page=5):
|
| 65 |
+
logger.debug(f"Buscando en Pexels: '{query}' | Resultados: {per_page}")
|
| 66 |
headers = {"Authorization": api_key}
|
| 67 |
try:
|
| 68 |
params = {
|
|
|
|
| 72 |
"size": "medium"
|
| 73 |
}
|
| 74 |
|
| 75 |
+
logger.debug(f"Params: {params}")
|
| 76 |
response = requests.get(
|
| 77 |
"https://api.pexels.com/videos/search",
|
| 78 |
headers=headers,
|
|
|
|
| 81 |
)
|
| 82 |
response.raise_for_status()
|
| 83 |
|
|
|
|
| 84 |
try:
|
| 85 |
data = response.json()
|
| 86 |
+
logger.info(f"Pexels: {len(data.get('videos', []))} videos encontrados")
|
| 87 |
return data.get('videos', [])
|
| 88 |
except json.JSONDecodeError:
|
| 89 |
+
logger.error(f"Pexels: JSON inválido | Status: {response.status_code} | Respuesta: {response.text[:200]}...")
|
| 90 |
return []
|
| 91 |
|
| 92 |
except requests.exceptions.RequestException as e:
|
| 93 |
+
logger.error(f"Error de conexión Pexels: {str(e)}")
|
| 94 |
except Exception as e:
|
| 95 |
+
logger.error(f"Error inesperado Pexels: {str(e)}", exc_info=True)
|
| 96 |
|
| 97 |
return []
|
| 98 |
|
| 99 |
+
# [FUNCIÓN GENERATE_SCRIPT ORIGINAL CON LOGS]
|
| 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 no disponibles - Usando prompt original")
|
| 104 |
+
return prompt
|
| 105 |
|
| 106 |
try:
|
|
|
|
| 107 |
enhanced_prompt = f"Escribe un guion corto y coherente sobre: {prompt}"
|
|
|
|
| 108 |
inputs = tokenizer(enhanced_prompt, return_tensors="pt", truncation=True, max_length=512)
|
| 109 |
|
| 110 |
+
logger.debug("Generando texto con GPT-2...")
|
| 111 |
outputs = model.generate(
|
| 112 |
**inputs,
|
| 113 |
max_length=max_length,
|
|
|
|
| 119 |
pad_token_id=tokenizer.pad_token_id,
|
| 120 |
eos_token_id=tokenizer.eos_token_id
|
| 121 |
)
|
|
|
|
| 122 |
|
| 123 |
+
text = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 124 |
+
text = re.sub(r'<[^>]+>', '', text)
|
| 125 |
sentences = text.split('.')
|
| 126 |
+
|
| 127 |
if sentences:
|
| 128 |
+
final_text = sentences[0] + '.'
|
| 129 |
+
logger.info(f"Guion generado: '{final_text[:100]}...'")
|
| 130 |
+
return final_text
|
| 131 |
return text
|
| 132 |
except Exception as e:
|
| 133 |
+
logger.error(f"Error generando guion: {str(e)}", exc_info=True)
|
| 134 |
+
return prompt
|
| 135 |
|
| 136 |
+
# [FUNCIÓN TEXT_TO_SPEECH ORIGINAL CON LOGS]
|
| 137 |
async def text_to_speech(text, output_path, voice="es-ES-ElviraNeural"):
|
| 138 |
+
logger.info(f"Convirtiendo texto a voz | Caracteres: {len(text)} | Voz: {voice}")
|
| 139 |
try:
|
| 140 |
communicate = edge_tts.Communicate(text, voice)
|
| 141 |
await communicate.save(output_path)
|
| 142 |
+
logger.info(f"Audio guardado en: {output_path} | Tamaño: {os.path.getsize(output_path)} bytes")
|
| 143 |
return True
|
| 144 |
except Exception as e:
|
| 145 |
+
logger.error(f"Error en TTS: {str(e)}", exc_info=True)
|
| 146 |
return False
|
| 147 |
|
| 148 |
+
# [FUNCIÓN DOWNLOAD_VIDEO_FILE ORIGINAL CON LOGS]
|
| 149 |
+
def download_video_file(url, temp_dir):
|
| 150 |
if not url:
|
| 151 |
+
logger.warning("URL de video no proporcionada")
|
| 152 |
return None
|
| 153 |
|
| 154 |
try:
|
| 155 |
+
logger.info(f"Descargando video desde: {url[:50]}...")
|
|
|
|
|
|
|
| 156 |
file_name = f"video_{datetime.now().strftime('%H%M%S%f')}.mp4"
|
| 157 |
output_path = os.path.join(temp_dir, file_name)
|
| 158 |
|
| 159 |
+
with requests.get(url, stream=True, timeout=30) as r:
|
| 160 |
+
r.raise_for_status()
|
| 161 |
+
with open(output_path, 'wb') as f:
|
| 162 |
+
for chunk in r.iter_content(chunk_size=8192):
|
| 163 |
+
f.write(chunk)
|
| 164 |
+
|
| 165 |
+
logger.info(f"Video descargado: {output_path} | Tamaño: {os.path.getsize(output_path)} bytes")
|
| 166 |
return output_path
|
| 167 |
except Exception as e:
|
| 168 |
+
logger.error(f"Error descargando video: {str(e)}", exc_info=True)
|
| 169 |
return None
|
| 170 |
|
| 171 |
+
# [FUNCIÓN LOOP_AUDIO_TO_LENGTH ORIGINAL CON LOGS]
|
| 172 |
def loop_audio_to_length(audio_clip, target_duration):
|
| 173 |
+
logger.debug(f"Ajustando audio | Duración actual: {audio_clip.duration:.2f}s | Objetivo: {target_duration:.2f}s")
|
| 174 |
if audio_clip.duration >= target_duration:
|
| 175 |
return audio_clip.subclip(0, target_duration)
|
| 176 |
|
| 177 |
+
loops = math.ceil(target_duration / audio_clip.duration)
|
| 178 |
+
logger.debug(f"Creando {loops} loops de audio")
|
| 179 |
audios = [audio_clip] * loops
|
| 180 |
return concatenate_videoclips(audios).subclip(0, target_duration)
|
| 181 |
|
| 182 |
+
# [FUNCIÓN EXTRACT_VISUAL_KEYWORDS_FROM_SCRIPT ORIGINAL CON LOGS]
|
| 183 |
def extract_visual_keywords_from_script(script_text):
|
| 184 |
+
logger.info("Extrayendo palabras clave del guion")
|
| 185 |
clean_text = re.sub(r'[^\w\sáéíóúñ]', '', script_text.lower())
|
| 186 |
|
|
|
|
| 187 |
if kw_model:
|
| 188 |
try:
|
| 189 |
keywords = kw_model.extract_keywords(
|
|
|
|
| 193 |
top_n=3
|
| 194 |
)
|
| 195 |
if keywords:
|
| 196 |
+
logger.debug(f"KeyBERT keywords: {keywords}")
|
| 197 |
return [kw[0].replace(" ", "+") for kw in keywords]
|
| 198 |
except Exception as e:
|
| 199 |
+
logger.warning(f"KeyBERT falló: {str(e)}")
|
| 200 |
|
|
|
|
| 201 |
words = clean_text.split()
|
| 202 |
stop_words = {"el", "la", "los", "las", "de", "en", "y", "a", "que", "es", "un", "una", "con"}
|
| 203 |
keywords = [word for word in words if len(word) > 3 and word not in stop_words]
|
| 204 |
|
|
|
|
| 205 |
if not keywords:
|
| 206 |
logger.warning("Usando palabras clave predeterminadas")
|
| 207 |
return ["naturaleza", "ciudad", "paisaje"]
|
| 208 |
|
|
|
|
| 209 |
word_counts = Counter(keywords)
|
| 210 |
+
top_keywords = [word.replace(" ", "+") for word, _ in word_counts.most_common(3)]
|
| 211 |
+
logger.info(f"Palabras clave finales: {top_keywords}")
|
| 212 |
+
return top_keywords
|
| 213 |
|
| 214 |
+
# [FUNCIÓN CREAR_VIDEO ORIGINAL CON LOGS]
|
| 215 |
def crear_video(prompt_type, input_text, musica_file=None):
|
| 216 |
+
logger.info("="*80)
|
| 217 |
+
logger.info(f"INICIANDO CREACIÓN DE VIDEO | Tipo: {prompt_type}")
|
| 218 |
+
logger.debug(f"Input: '{input_text[:100]}...'")
|
| 219 |
|
| 220 |
# 1. Generar o usar guion
|
| 221 |
+
start_time = datetime.now()
|
| 222 |
if prompt_type == "Generar Guion con IA":
|
| 223 |
+
logger.info("Generando guion con IA...")
|
| 224 |
guion = generate_script(input_text)
|
| 225 |
else:
|
| 226 |
+
logger.info("Usando guion proporcionado")
|
| 227 |
guion = input_text
|
| 228 |
|
| 229 |
+
logger.info(f"Guion final ({len(guion)} caracteres): '{guion[:100]}...'")
|
| 230 |
|
|
|
|
| 231 |
if not guion.strip():
|
| 232 |
+
logger.error("El guion está vacío")
|
| 233 |
raise ValueError("El guion está vacío")
|
| 234 |
|
| 235 |
# Directorio temporal
|
| 236 |
temp_dir = tempfile.mkdtemp()
|
| 237 |
+
logger.info(f"Directorio temporal creado: {temp_dir}")
|
| 238 |
temp_files = []
|
| 239 |
|
| 240 |
try:
|
| 241 |
# 2. Generar audio de voz
|
| 242 |
+
logger.info("Generando audio de voz...")
|
| 243 |
voz_path = os.path.join(temp_dir, "voz.mp3")
|
| 244 |
if not asyncio.run(text_to_speech(guion, voz_path)):
|
| 245 |
+
logger.error("Fallo en generación de voz")
|
| 246 |
raise ValueError("Error generando voz")
|
| 247 |
temp_files.append(voz_path)
|
| 248 |
|
| 249 |
audio_tts = AudioFileClip(voz_path)
|
| 250 |
audio_duration = audio_tts.duration
|
| 251 |
+
logger.info(f"Duración audio voz: {audio_duration:.2f} segundos")
|
| 252 |
|
| 253 |
+
# 3. Extraer palabras clave
|
| 254 |
+
logger.info("Extrayendo palabras clave...")
|
| 255 |
try:
|
| 256 |
keywords = extract_visual_keywords_from_script(guion)
|
| 257 |
+
logger.info(f"Palabras clave identificadas: {keywords}")
|
| 258 |
except Exception as e:
|
| 259 |
+
logger.error(f"Error extrayendo keywords: {str(e)}")
|
| 260 |
+
keywords = ["naturaleza", "paisaje"]
|
|
|
|
|
|
|
| 261 |
|
| 262 |
+
# 4. Buscar y descargar videos
|
| 263 |
+
logger.info("Buscando videos en Pexels...")
|
| 264 |
videos_data = []
|
| 265 |
for keyword in keywords:
|
| 266 |
try:
|
|
|
|
| 269 |
videos_data.extend(videos)
|
| 270 |
logger.info(f"Encontrados {len(videos)} videos para '{keyword}'")
|
| 271 |
except Exception as e:
|
| 272 |
+
logger.warning(f"Error buscando videos para '{keyword}': {str(e)}")
|
| 273 |
|
|
|
|
| 274 |
if not videos_data:
|
| 275 |
+
logger.warning("No se encontraron videos - Usando palabras clave genéricas")
|
| 276 |
for keyword in ["naturaleza", "ciudad", "paisaje"]:
|
| 277 |
videos = buscar_videos_pexels(keyword, PEXELS_API_KEY, per_page=3)
|
| 278 |
if videos:
|
| 279 |
videos_data.extend(videos)
|
| 280 |
|
| 281 |
if not videos_data:
|
| 282 |
+
logger.error("No se encontraron videos para ninguna palabra clave")
|
| 283 |
raise ValueError("No se encontraron videos en Pexels para ninguna palabra clave")
|
| 284 |
|
| 285 |
video_paths = []
|
|
|
|
| 288 |
continue
|
| 289 |
|
| 290 |
try:
|
|
|
|
| 291 |
best_quality = max(
|
| 292 |
video['video_files'],
|
| 293 |
key=lambda x: x.get('width', 0) * x.get('height', 0)
|
|
|
|
| 300 |
temp_files.append(path)
|
| 301 |
logger.info(f"Video descargado: {best_quality['link']}")
|
| 302 |
except Exception as e:
|
| 303 |
+
logger.warning(f"Error procesando video: {str(e)}")
|
| 304 |
|
| 305 |
if not video_paths:
|
| 306 |
+
logger.error("No se pudo descargar ningún video")
|
| 307 |
raise ValueError("No se pudo descargar ningún video")
|
| 308 |
|
| 309 |
# 5. Procesar videos
|
| 310 |
+
logger.info("Procesando videos descargados...")
|
| 311 |
clips = []
|
| 312 |
current_duration = 0
|
| 313 |
|
|
|
|
| 319 |
clip = VideoFileClip(path)
|
| 320 |
usable_duration = min(clip.duration, 10)
|
| 321 |
|
| 322 |
+
if usable_duration > 1:
|
| 323 |
clips.append(clip.subclip(0, usable_duration))
|
| 324 |
current_duration += usable_duration
|
| 325 |
+
logger.debug(f"Clip añadido: {usable_duration:.1f}s (total: {current_duration:.1f}/{audio_duration:.1f}s)")
|
| 326 |
except Exception as e:
|
| 327 |
+
logger.warning(f"Error procesando video {path}: {str(e)}")
|
| 328 |
|
| 329 |
if not clips:
|
| 330 |
+
logger.error("No hay clips válidos para crear el video")
|
| 331 |
raise ValueError("No hay clips válidos para crear el video")
|
| 332 |
|
| 333 |
video_base = concatenate_videoclips(clips, method="compose")
|
| 334 |
+
logger.info(f"Duración base del video: {video_base.duration:.2f}s")
|
| 335 |
|
|
|
|
| 336 |
if video_base.duration < audio_duration:
|
| 337 |
+
num_repeats = math.ceil(audio_duration / video_base.duration)
|
| 338 |
+
logger.info(f"Repitiendo video {num_repeats} veces para ajustar duración")
|
| 339 |
+
video_base = concatenate_videoclips([video_base] * num_repeats).subclip(0, audio_duration)
|
| 340 |
|
| 341 |
# 6. Manejar música de fondo
|
| 342 |
+
logger.info("Procesando audio...")
|
| 343 |
final_audio = audio_tts
|
| 344 |
|
| 345 |
if musica_file:
|
| 346 |
try:
|
|
|
|
| 347 |
music_path = os.path.join(temp_dir, "musica.mp3")
|
| 348 |
shutil.copyfile(musica_file, music_path)
|
| 349 |
temp_files.append(music_path)
|
| 350 |
+
logger.info(f"Música copiada a: {music_path}")
|
| 351 |
|
|
|
|
| 352 |
musica_audio = AudioFileClip(music_path)
|
| 353 |
+
logger.debug(f"Duración música original: {musica_audio.duration:.2f}s")
|
| 354 |
|
|
|
|
| 355 |
if musica_audio.duration < audio_duration:
|
| 356 |
musica_audio = loop_audio_to_length(musica_audio, audio_duration)
|
| 357 |
+
logger.debug(f"Música looped: {musica_audio.duration:.2f}s")
|
| 358 |
|
|
|
|
| 359 |
final_audio = CompositeAudioClip([
|
| 360 |
+
musica_audio.volumex(0.3),
|
| 361 |
+
audio_tts.volumex(1.0)
|
| 362 |
])
|
| 363 |
+
logger.info("Mezcla de audio completada")
|
| 364 |
except Exception as e:
|
| 365 |
+
logger.warning(f"Error procesando música: {str(e)}")
|
| 366 |
|
| 367 |
# 7. Crear video final
|
| 368 |
+
logger.info("Renderizando video final...")
|
| 369 |
video_final = video_base.set_audio(final_audio)
|
|
|
|
| 370 |
output_path = os.path.join(temp_dir, "final_video.mp4")
|
| 371 |
+
|
| 372 |
video_final.write_videofile(
|
| 373 |
output_path,
|
| 374 |
fps=24,
|
|
|
|
| 379 |
logger=None
|
| 380 |
)
|
| 381 |
|
| 382 |
+
total_time = (datetime.now() - start_time).total_seconds()
|
| 383 |
+
logger.info(f"VIDEO FINALIZADO: {output_path} | Tiempo total: {total_time:.2f}s")
|
| 384 |
return output_path
|
| 385 |
|
| 386 |
except Exception as e:
|
| 387 |
+
logger.error(f"ERROR EN CREAR_VIDEO: {str(e)}", exc_info=True)
|
| 388 |
raise
|
| 389 |
finally:
|
| 390 |
+
logger.info("Limpiando archivos temporales...")
|
| 391 |
for path in temp_files:
|
| 392 |
try:
|
| 393 |
if os.path.isfile(path):
|
| 394 |
os.remove(path)
|
| 395 |
+
except Exception as e:
|
| 396 |
+
logger.warning(f"No se pudo eliminar {path}: {str(e)}")
|
| 397 |
+
|
| 398 |
+
try:
|
| 399 |
shutil.rmtree(temp_dir, ignore_errors=True)
|
| 400 |
+
except Exception as e:
|
| 401 |
+
logger.warning(f"No se pudo eliminar {temp_dir}: {str(e)}")
|
| 402 |
|
| 403 |
+
# [FUNCIÓN RUN_APP ORIGINAL CON LOGS]
|
| 404 |
def run_app(prompt_type, prompt_ia, prompt_manual, musica_file):
|
| 405 |
+
logger.info("="*80)
|
| 406 |
+
logger.info("SOLICITUD RECIBIDA EN INTERFAZ")
|
| 407 |
input_text = prompt_ia if prompt_type == "Generar Guion con IA" else prompt_manual
|
| 408 |
|
| 409 |
if not input_text.strip():
|
| 410 |
+
logger.warning("Texto de entrada vacío")
|
| 411 |
return None, "Por favor ingresa texto"
|
| 412 |
|
| 413 |
try:
|
| 414 |
+
logger.info("Iniciando creación de video...")
|
| 415 |
video_path = crear_video(prompt_type, input_text, musica_file)
|
| 416 |
+
logger.info("Video creado exitosamente")
|
| 417 |
return video_path, "✅ Video generado exitosamente"
|
| 418 |
except ValueError as ve:
|
| 419 |
+
logger.warning(f"Error de validación: {str(ve)}")
|
| 420 |
return None, f"⚠️ {ve}"
|
| 421 |
except Exception as e:
|
| 422 |
+
logger.error(f"Error crítico: {str(e)}", exc_info=True)
|
| 423 |
return None, f"❌ Error: {str(e)}"
|
| 424 |
|
| 425 |
+
# [INTERFAZ DE GRADIO ORIGINAL COMPLETA]
|
| 426 |
with gr.Blocks(title="Generador de Videos con IA", theme=gr.themes.Soft(), css="""
|
| 427 |
.gradio-container {max-width: 800px; margin: auto;}
|
| 428 |
h1 {text-align: center;}
|
|
|
|
| 475 |
placeholder="Esperando acción..."
|
| 476 |
)
|
| 477 |
|
|
|
|
| 478 |
prompt_type.change(
|
| 479 |
lambda x: (gr.update(visible=x == "Generar Guion con IA"),
|
| 480 |
gr.update(visible=x == "Usar Mi Guion")),
|
|
|
|
| 482 |
outputs=[ia_guion_column, manual_guion_column]
|
| 483 |
)
|
| 484 |
|
|
|
|
| 485 |
generate_btn.click(
|
| 486 |
lambda: (None, "⏳ Procesando... (esto puede tomar 2-5 minutos)"),
|
| 487 |
outputs=[video_output, status_output],
|
|
|
|
| 498 |
- "Generar Guion con IA": Describe un tema
|
| 499 |
- "Usar Mi Guion": Escribe tu guion completo
|
| 500 |
2. **Sube música** (opcional): Selecciona un archivo de audio
|
| 501 |
+
3. **Haz clic en Generar Video"
|
| 502 |
4. Espera a que se procese el video (puede tomar varios minutos)
|
| 503 |
""")
|
| 504 |
|
| 505 |
if __name__ == "__main__":
|
| 506 |
+
logger.info("Iniciando aplicación Gradio...")
|
| 507 |
+
try:
|
| 508 |
+
app.launch(server_name="0.0.0.0", server_port=7860)
|
| 509 |
+
except Exception as e:
|
| 510 |
+
logger.critical(f"No se pudo iniciar la app: {str(e)}", exc_info=True)
|
| 511 |
+
raise
|