Update app.py

app.py

@@ -154,24 +154,10 @@ def extract_keywords(text: str) -> list[str]:
         clean_text = re.sub(r"[^\w\sáéíóúñÁÉÍÓÚÑ]", "", text.lower())
         kws = local_kw_model.extract_keywords(clean_text, stop_words="spanish", top_n=5)
         keywords = [k.replace(" ", "+") for k, _ in kws if k]
-        return keywords if keywords else ["mystery", "conspiracy", "alien", "UFO", "secret", "cover-up", "illusion", "paranoia", 
-                "secret society", "lie", "simulation", "matrix", "terror", "darkness", "shadow", "enigma", 
-                "urban legend", "unknown", "hidden", "mistrust", "experiment", "government", "control", 
-                "surveillance", "propaganda", "deception", "whistleblower", "anomaly", "extraterrestrial", 
-                "shadow government", "cabal", "deep state", "new world order", "mind control", "brainwashing", 
-                "disinformation", "false flag", "assassin", "black ops", "anomaly", "men in black", "abduction", 
-                "hybrid", "ancient aliens", "hollow earth", "simulation theory", "alternate reality", "predictive programming", 
-                "symbolism", "occult", "eerie", "haunting", "unexplained", "forbidden knowledge", "redacted", "conspiracy theorist"]
+        return keywords if keywords else ["mystery", "conspiracy", "alien", "UFO", "secret", "cover-up", "illusion", "paranoia"]
     except Exception as e:
         logger.error(f"Error extrayendo keywords: {e}")
-        return ["mystery", "conspiracy", "alien", "UFO", "secret", "cover-up", "illusion", "paranoia", 
-                "secret society", "lie", "simulation", "matrix", "terror", "darkness", "shadow", "enigma", 
-                "urban legend", "unknown", "hidden", "mistrust", "experiment", "government", "control", 
-                "surveillance", "propaganda", "deception", "whistleblower", "anomaly", "extraterrestrial", 
-                "shadow government", "cabal", "deep state", "new world order", "mind control", "brainwashing", 
-                "disinformation", "false flag", "assassin", "black ops", "anomaly", "men in black", "abduction", 
-                "hybrid", "ancient aliens", "hollow earth", "simulation theory", "alternate reality", "predictive programming", 
-                "symbolism", "occult", "eerie", "haunting", "unexplained", "forbidden knowledge", "redacted", "conspiracy theorist"]
+        return ["mystery", "conspiracy", "alien", "UFO", "secret", "cover-up", "illusion", "paranoia"]
 
 def search_pexels_videos(query: str, count: int = 3) -> list[dict]:
     if not PEXELS_API_KEY:
@@ -211,54 +197,6 @@ def download_video(url: str, folder: str) -> str | None:
         logger.error(f"Error descargando video {url}: {e}")
         return None
 
-def create_subtitle_clips(script: str, video_width: int, video_height: int, duration: float):
-    try:
-        sentences = [s.strip() for s in re.split(r"[.!?¿¡]", script) if s.strip()]
-        if not sentences:
-            return []
-        
-        total_words = sum(len(s.split()) for s in sentences) or 1
-        time_per_word = duration / total_words
-        
-        clips = []
-        current_time = 0.0
-        
-        for sentence in sentences:
-            num_words = len(sentence.split())
-            sentence_duration = num_words * time_per_word
-            
-            if sentence_duration < 0.5:
-                continue
-                
-            try:
-                txt_clip = (
-                    TextClip(
-                        sentence,
-                        fontsize=max(20, int(video_height * 0.05)),
-                        color="white",
-                        stroke_color="black",
-                        stroke_width=2,
-                        method="caption",
-                        size=(int(video_width * 0.9), None),
-                        font="Arial-Bold"
-                    )
-                    .set_start(current_time)
-                    .set_duration(sentence_duration)
-                    .set_position(("center", "bottom"))
-                )
-                if txt_clip is not None:
-                    clips.append(txt_clip)
-            except Exception as e:
-                logger.error(f"Error creando subtítulo para '{sentence}': {e}")
-                continue
-            
-            current_time += sentence_duration
-            
-        return clips
-    except Exception as e:
-        logger.error(f"Error creando subtítulos: {e}")
-        return []
-
 def loop_audio_to_duration(audio_clip: AudioFileClip, target_duration: float) -> AudioFileClip:
     if audio_clip is None:
         return None
@@ -277,6 +215,7 @@ def create_video(script_text: str, generate_script: bool, music_path: str | None
     temp_dir = tempfile.mkdtemp()
     TARGET_FPS = 24
     TARGET_RESOLUTION = (1280, 720)
+    MAX_CLIP_DURATION = 8  # Máximo de segundos por clip
     
     def normalize_clip(clip):
         if clip is None:
@@ -318,19 +257,22 @@ def create_video(script_text: str, generate_script: bool, music_path: str | None
         if video_duration < 1:
             raise ValueError("El audio generado es demasiado corto")
         
-        # Paso 3: Buscar y descargar videos
+        # Paso 3: Buscar y descargar videos (adaptado a la duración del audio)
         update_task_progress(task_id, "Paso 3/7: Buscando videos en Pexels...")
         video_paths = []
         keywords = extract_keywords(script)
         
-        for i, keyword in enumerate(keywords[:3]):
-            update_task_progress(task_id, f"Paso 3/7: Buscando videos para '{keyword}' ({i+1}/{len(keywords[:3])})")
+        # Calcular cuántos clips necesitamos aproximadamente
+        estimated_clips_needed = max(1, math.ceil(video_duration / MAX_CLIP_DURATION))
+        
+        for i, keyword in enumerate(keywords):
+            if len(video_paths) >= estimated_clips_needed * 2:  # Buscar el doble para tener opciones
+                break
+                
+            update_task_progress(task_id, f"Paso 3/7: Buscando videos para '{keyword}' ({i+1}/{len(keywords)})")
             
-            videos = search_pexels_videos(keyword, 2)
+            videos = search_pexels_videos(keyword, 3)  # Buscar 3 videos por keyword
             for video_data in videos:
-                if len(video_paths) >= 6:
-                    break
-                
                 video_files = video_data.get("video_files", [])
                 if video_files:
                     # Encontrar el video con la mejor calidad que sea MP4
@@ -349,15 +291,21 @@ def create_video(script_text: str, generate_script: bool, music_path: str | None
                         downloaded_path = download_video(video_url, temp_dir)
                         if downloaded_path:
                             video_paths.append(downloaded_path)
+                            if len(video_paths) >= estimated_clips_needed * 2:
+                                break
         
         if not video_paths:
             raise RuntimeError("No se pudieron descargar videos de Pexels")
         
         # Paso 4: Procesar videos - MANEJO CORRECTO DE ERRORES
-        update_task_progress(task_id, f"Paso 4/7: Procesando {len(video_paths)} videos...")
+        update_task_progress(task_id, f"Paso 4/7: Procesando videos...")
         video_clips = []
+        total_duration = 0
         
         for path in video_paths:
+            if total_duration >= video_duration:
+                break
+                
             clip = None
             try:
                 # Cargar el video con verificación adicional
@@ -374,10 +322,15 @@ def create_video(script_text: str, generate_script: bool, music_path: str | None
                     clip.close()
                     continue
                 
-                # Tomar máximo 8 segundos de cada clip
+                # Tomar máximo 8 segundos de cada clip o lo que necesitemos
                 try:
-                    duration = min(8, clip.duration)
-                    processed_clip = clip.subclip(0, duration)
+                    clip_duration = min(MAX_CLIP_DURATION, clip.duration)
+                    # Si ya tenemos suficiente duración, tomar solo lo necesario
+                    remaining_duration = video_duration - total_duration
+                    if remaining_duration < clip_duration:
+                        clip_duration = remaining_duration
+                    
+                    processed_clip = clip.subclip(0, clip_duration)
                 except Exception as e:
                     logger.error(f"Error al recortar video {path}: {e}")
                     clip.close()
@@ -388,6 +341,7 @@ def create_video(script_text: str, generate_script: bool, music_path: str | None
                     processed_clip = normalize_clip(processed_clip)
                     if processed_clip is not None:
                         video_clips.append(processed_clip)
+                        total_duration += processed_clip.duration
                     else:
                         if 'processed_clip' in locals():
                             processed_clip.close()
@@ -411,11 +365,6 @@ def create_video(script_text: str, generate_script: bool, music_path: str | None
         # Concatenar videos
         base_video = concatenate_videoclips(video_clips, method="chain")
         
-        # Extender video si es más corto que el audio
-        if base_video.duration < video_duration:
-            loops_needed = math.ceil(video_duration / base_video.duration)
-            base_video = concatenate_videoclips([base_video] * loops_needed)
-        
         # Asegurar duración exacta
         base_video = base_video.subclip(0, video_duration)
         
@@ -432,14 +381,8 @@ def create_video(script_text: str, generate_script: bool, music_path: str | None
         else:
             final_audio = voice_clip
         
-        # Paso 6: Agregar subtítulos
-        update_task_progress(task_id, "Paso 6/7: Agregando subtítulos...")
-        subtitle_clips = create_subtitle_clips(script, base_video.w, base_video.h, video_duration)
-        if subtitle_clips:
-            base_video = CompositeVideoClip([base_video] + subtitle_clips)
-        
-        # Paso 7: Renderizar video final
-        update_task_progress(task_id, "Paso 7/7: Renderizando video final...")
+        # Paso 6: Renderizar video final
+        update_task_progress(task_id, "Paso 6/7: Renderizando video final...")
         final_video = base_video.set_audio(final_audio)
         
         output_path = os.path.join(RESULTS_DIR, f"video_{task_id}.mp4")
@@ -455,6 +398,9 @@ def create_video(script_text: str, generate_script: bool, music_path: str | None
             verbose=False
         )
         
+        # Paso 7: Limpiar recursos
+        update_task_progress(task_id, "Paso 7/7: Finalizando...")
+        
         # Limpiar clips
         voice_clip.close()
         if 'music_clip' in locals():
@@ -564,7 +510,6 @@ with gr.Blocks(title="🎬 Generador de Videos IA", theme=gr.themes.Soft()) as d
     - **Edge TTS** para voz en español
     - **GPT-2** para generación de guiones
     - **Pexels API** para videos de stock
-    - **Subtítulos automáticos** y efectos visuales
     
     El progreso se mostrará en tiempo real.
     """)