core/visual_engine.py

# core/visual_engine.py
from PIL import Image, ImageDraw, ImageFont, ImageOps
# --- MONKEY PATCH FOR Image.ANTIALIAS ---
try:
    if hasattr(Image, 'Resampling') and hasattr(Image.Resampling, 'LANCZOS'): # Pillow 9+
        if not hasattr(Image, 'ANTIALIAS'): Image.ANTIALIAS = Image.Resampling.LANCZOS
    elif hasattr(Image, 'LANCZOS'): # Pillow 8
         if not hasattr(Image, 'ANTIALIAS'): Image.ANTIALIAS = Image.LANCZOS
    elif not hasattr(Image, 'ANTIALIAS'):
             print("WARNING: Pillow version lacks common Resampling attributes or ANTIALIAS. Video effects might fail.")
except Exception as e_mp: print(f"WARNING: ANTIALIAS monkey-patch error: {e_mp}")
# --- END MONKEY PATCH ---

from moviepy.editor import (ImageClip, VideoFileClip, concatenate_videoclips, TextClip,
                            CompositeVideoClip, AudioFileClip)
import moviepy.video.fx.all as vfx
import numpy as np
import os
import openai
import requests
import io
import time
import random
import logging

logger = logging.getLogger(__name__)
logger.setLevel(logging.INFO) # Set default logging level for this module

# --- ElevenLabs Client Import ---
ELEVENLABS_CLIENT_IMPORTED = False
ElevenLabsAPIClient = None
Voice = None
VoiceSettings = None
try:
    from elevenlabs.client import ElevenLabs as ImportedElevenLabsClient
    from elevenlabs import Voice as ImportedVoice, VoiceSettings as ImportedVoiceSettings
    ElevenLabsAPIClient = ImportedElevenLabsClient
    Voice = ImportedVoice
    VoiceSettings = ImportedVoiceSettings
    ELEVENLABS_CLIENT_IMPORTED = True
    logger.info("ElevenLabs client components imported successfully.")
except Exception as e_eleven:
    logger.warning(f"ElevenLabs client import failed: {e_eleven}. Audio generation will be disabled.")

# --- RunwayML Client Import (Placeholder) ---
RUNWAYML_SDK_IMPORTED = False
RunwayMLClient = None # Placeholder for the actual RunwayML client class
try:
    # Example: from runwayml import RunwayClient as ImportedRunwayMLClient
    # RunwayMLClient = ImportedRunwayMLClient
    # RUNWAYML_SDK_IMPORTED = True
    logger.info("RunwayML SDK import is a placeholder. Actual SDK needed for Runway features.")
except ImportError:
    logger.warning("RunwayML SDK (placeholder) not found. RunwayML video generation will be disabled.")
except Exception as e_runway_sdk:
    logger.warning(f"Error importing RunwayML SDK (placeholder): {e_runway_sdk}. RunwayML features disabled.")


class VisualEngine:
    def __init__(self, output_dir="temp_cinegen_media", default_elevenlabs_voice_id="Rachel"):
        self.output_dir = output_dir
        os.makedirs(self.output_dir, exist_ok=True)

        self.font_filename = "arial.ttf"
        font_paths_to_try = [
            self.font_filename,
            f"/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf", # Common on Linux
            f"/usr/share/fonts/truetype/liberation/LiberationSans-Bold.ttf", # Common on Linux
            f"/System/Library/Fonts/Supplemental/Arial.ttf", # macOS
            f"C:/Windows/Fonts/arial.ttf", # Windows
            f"/usr/local/share/fonts/truetype/mycustomfonts/{self.font_filename}" # Custom container path
        ]
        self.font_path_pil = next((p for p in font_paths_to_try if os.path.exists(p)), None)
        self.font_size_pil = 20
        self.video_overlay_font_size = 30
        self.video_overlay_font_color = 'white'
        self.video_overlay_font = 'Liberation-Sans-Bold' # For MoviePy TextClip (ImageMagick name)

        try:
            if self.font_path_pil:
                self.font = ImageFont.truetype(self.font_path_pil, self.font_size_pil)
                logger.info(f"Pillow font loaded: {self.font_path_pil}.")
            else: # Fallback to default if no path found
                self.font = ImageFont.load_default()
                logger.warning("Custom Pillow font not found from paths. Using default. Text rendering might be basic.")
                self.font_size_pil = 10 # Default font is smaller
        except IOError as e_font: # Catch specific IOError for font loading
            logger.error(f"Pillow font loading IOError for '{self.font_path_pil if self.font_path_pil else 'default'}': {e_font}. Using default.")
            self.font = ImageFont.load_default()
            self.font_size_pil = 10

        self.openai_api_key = None; self.USE_AI_IMAGE_GENERATION = False
        self.dalle_model = "dall-e-3"; self.image_size_dalle3 = "1792x1024"
        self.video_frame_size = (1280, 720) # Standard HD 16:9

        self.elevenlabs_api_key = None; self.USE_ELEVENLABS = False
        self.elevenlabs_client = None
        self.elevenlabs_voice_id = default_elevenlabs_voice_id
        if VoiceSettings and ELEVENLABS_CLIENT_IMPORTED:
            self.elevenlabs_voice_settings = VoiceSettings(stability=0.60, similarity_boost=0.80, style=0.15, use_speaker_boost=True)
        else: self.elevenlabs_voice_settings = None

        self.pexels_api_key = None; self.USE_PEXELS = False
        self.runway_api_key = None; self.USE_RUNWAYML = False
        self.runway_client = None

        logger.info("VisualEngine initialized.")

    def set_openai_api_key(self,k):
        self.openai_api_key=k; self.USE_AI_IMAGE_GENERATION=bool(k)
        logger.info(f"DALL-E ({self.dalle_model}) {'Ready.' if k else 'Disabled (no API key).'}")

    def set_elevenlabs_api_key(self,api_key, voice_id_from_secret=None):
        self.elevenlabs_api_key=api_key
        if voice_id_from_secret: self.elevenlabs_voice_id = voice_id_from_secret
        if api_key and ELEVENLABS_CLIENT_IMPORTED and ElevenLabsAPIClient:
            try:
                self.elevenlabs_client = ElevenLabsAPIClient(api_key=api_key)
                self.USE_ELEVENLABS=bool(self.elevenlabs_client)
                logger.info(f"ElevenLabs Client {'Ready' if self.USE_ELEVENLABS else 'Failed Init'} (Voice ID: {self.elevenlabs_voice_id}).")
            except Exception as e: logger.error(f"ElevenLabs client init error: {e}. Disabled.", exc_info=True); self.USE_ELEVENLABS=False
        else: self.USE_ELEVENLABS=False; logger.info("ElevenLabs Disabled (no API key or SDK issue).")

    def set_pexels_api_key(self,k):
        self.pexels_api_key=k; self.USE_PEXELS=bool(k)
        logger.info(f"Pexels Search {'Ready.' if k else 'Disabled (no API key).'}")

    def set_runway_api_key(self, k):
        self.runway_api_key = k
        if k and RUNWAYML_SDK_IMPORTED and RunwayMLClient:
            try:
                # self.runway_client = RunwayMLClient(api_key=k) # Actual initialization
                self.USE_RUNWAYML = True # Assume success for placeholder with hypothetical SDK
                logger.info(f"RunwayML Client (Placeholder with SDK) {'Ready.' if self.USE_RUNWAYML else 'Failed Init.'}")
            except Exception as e: logger.error(f"RunwayML client (Placeholder with SDK) init error: {e}. Disabled.", exc_info=True); self.USE_RUNWAYML = False
        elif k: # API key provided, but SDK might not be used/imported (e.g., direct HTTP)
            self.USE_RUNWAYML = True
            logger.info("RunwayML API Key set. Using direct API calls or placeholder (SDK not fully integrated/imported).")
        else: self.USE_RUNWAYML = False; logger.info("RunwayML Disabled (no API key).")

    def _get_text_dimensions(self,text_content,font_obj):
        if not text_content: return 0,self.font_size_pil
        try:
            if hasattr(font_obj,'getbbox'):
                bbox=font_obj.getbbox(text_content);w=bbox[2]-bbox[0];h=bbox[3]-bbox[1]
                return w, h if h > 0 else self.font_size_pil
            elif hasattr(font_obj,'getsize'):
                w,h=font_obj.getsize(text_content)
                return w, h if h > 0 else self.font_size_pil
            else: return int(len(text_content)*self.font_size_pil*0.6),int(self.font_size_pil*1.2 if self.font_size_pil*1.2>0 else self.font_size_pil)
        except Exception as e: logger.warning(f"Error in _get_text_dimensions for '{text_content[:20]}...': {e}"); return int(len(text_content)*self.font_size_pil*0.6),int(self.font_size_pil*1.2)

    def _create_placeholder_image_content(self,text_description,filename,size=None):
        # (No significant changes from your previous correct version)
        if size is None: size = self.video_frame_size
        img=Image.new('RGB',size,color=(20,20,40));d=ImageDraw.Draw(img);padding=25;max_w=size[0]-(2*padding);lines=[];
        if not text_description: text_description="(Placeholder: No prompt text)"
        words=text_description.split();current_line=""
        for word in words:
            test_line=current_line+word+" ";
            if self._get_text_dimensions(test_line,self.font)[0] <= max_w: current_line=test_line
            else:
                if current_line: lines.append(current_line.strip());
                current_line=word+" "
        if current_line.strip(): lines.append(current_line.strip())
        if not lines and text_description: lines.append(text_description[:int(max_w//(self.font_size_pil*0.6 +1))]+"..." if text_description else "(Text too long)")
        elif not lines: lines.append("(Placeholder Text Error)")
        _,single_line_h=self._get_text_dimensions("Ay",self.font); single_line_h = single_line_h if single_line_h > 0 else self.font_size_pil + 2
        max_lines_to_display=min(len(lines),(size[1]-(2*padding))//(single_line_h+2)) if single_line_h > 0 else 1
        if max_lines_to_display <=0: max_lines_to_display = 1
        y_text_start = padding + (size[1]-(2*padding) - max_lines_to_display*(single_line_h+2))/2.0
        y_text = y_text_start
        for i in range(max_lines_to_display):
            line_content=lines[i];line_w,_=self._get_text_dimensions(line_content,self.font);x_text=(size[0]-line_w)/2.0
            d.text((x_text,y_text),line_content,font=self.font,fill=(200,200,180));y_text+=single_line_h+2
            if i==6 and max_lines_to_display > 7: d.text((x_text,y_text),"...",font=self.font,fill=(200,200,180));break
        filepath=os.path.join(self.output_dir,filename);
        try:img.save(filepath);return filepath
        except Exception as e:logger.error(f"Saving placeholder image {filepath}: {e}", exc_info=True);return None

    def _search_pexels_image(self, query, output_filename_base):
        # (No significant changes from your previous correct version)
        if not self.USE_PEXELS or not self.pexels_api_key: return None
        headers = {"Authorization": self.pexels_api_key}; params = {"query": query, "per_page": 1, "orientation": "landscape", "size": "large"}
        pexels_filename = output_filename_base.replace(".png", f"_pexels_{random.randint(1000,9999)}.jpg").replace(".mp4", f"_pexels_{random.randint(1000,9999)}.jpg")
        filepath = os.path.join(self.output_dir, pexels_filename)
        try:
            logger.info(f"Searching Pexels for: '{query}'"); effective_query = " ".join(query.split()[:5]); params["query"] = effective_query
            response = requests.get("https://api.pexels.com/v1/search", headers=headers, params=params, timeout=20)
            response.raise_for_status(); data = response.json()
            if data.get("photos") and len(data["photos"]) > 0:
                photo_url = data["photos"][0]["src"]["large2x"]
                image_response = requests.get(photo_url, timeout=60); image_response.raise_for_status()
                img_data = Image.open(io.BytesIO(image_response.content))
                if img_data.mode != 'RGB': img_data = img_data.convert('RGB')
                img_data.save(filepath); logger.info(f"Pexels image saved: {filepath}"); return filepath
            else: logger.info(f"No photos found on Pexels for query: '{effective_query}'")
        except Exception as e: logger.error(f"Pexels search/download for query '{query}': {e}", exc_info=True)
        return None

    def _generate_video_clip_with_runwayml(self, prompt_text, scene_identifier_filename_base, target_duration_seconds=4, input_image_path=None):
        if not self.USE_RUNWAYML or not self.runway_api_key:
            logger.warning("RunwayML not enabled or API key missing. Cannot generate video clip.")
            return None
        output_video_filename = scene_identifier_filename_base.replace(".png", ".mp4") # Ensure .mp4 extension
        output_video_filepath = os.path.join(self.output_dir, output_video_filename)
        logger.info(f"Attempting RunwayML video generation for: {prompt_text[:100]}... (Target duration: {target_duration_seconds}s)")
        # --- START ACTUAL RUNWAYML API INTERACTION (HYPOTHETICAL - NEEDS IMPLEMENTATION) ---
        # ... (Your actual RunwayML API call logic would go here) ...
        # --- END ACTUAL RUNWAYML API INTERACTION (HYPOTHETICAL) ---
        logger.warning("Using PLACEHOLDER video generation for RunwayML as actual API calls are not implemented.")
        return self._create_placeholder_video_content(f"[RunwayML Placeholder] {prompt_text}", output_video_filename, duration=target_duration_seconds)

    def _create_placeholder_video_content(self, text_description, filename, duration=4, size=None):
        if size is None: size = self.video_frame_size
        filepath = os.path.join(self.output_dir, filename)
        txt_clip = TextClip(text_description, fontsize=50, color='white', font=self.video_overlay_font,
                            bg_color='black', size=size, method='caption').set_duration(duration)
        try:
            txt_clip.write_videofile(filepath, fps=24, codec='libx264', preset='ultrafast', logger=None)
            logger.info(f"Placeholder video saved: {filepath}")
            return filepath
        except Exception as e: logger.error(f"Failed to create placeholder video {filepath}: {e}", exc_info=True); return None
        finally:
            if hasattr(txt_clip, 'close'): txt_clip.close()

    def generate_scene_asset(self, image_prompt_text, scene_data, scene_identifier_filename_base,
                             generate_as_video_clip=False, runway_target_duration=4, input_image_for_runway=None):
        base_name, _ = os.path.splitext(scene_identifier_filename_base)
        asset_info = {'path': None, 'type': 'none', 'error': True, 'prompt_used': image_prompt_text, 'error_message': 'Generation not attempted'}

        if generate_as_video_clip and self.USE_RUNWAYML:
            logger.info(f"Attempting RunwayML video clip generation for {base_name}")
            video_path = self._generate_video_clip_with_runwayml(
                image_prompt_text, base_name,
                target_duration_seconds=runway_target_duration,
                input_image_path=input_image_for_runway
            )
            if video_path and os.path.exists(video_path):
                asset_info = {'path': video_path, 'type': 'video', 'error': False, 'prompt_used': image_prompt_text}
                return asset_info # Successfully generated video
            else:
                logger.warning(f"RunwayML video clip generation failed for {base_name}. Falling back to image.")
                asset_info['error_message'] = "RunwayML video generation failed."
                # Fall through to image generation

        # Image Generation (DALL-E, Pexels, Placeholder)
        image_filename_with_ext = base_name + ".png" # Ensure .png for image
        filepath = os.path.join(self.output_dir, image_filename_with_ext)
        asset_info['type'] = 'image' # Tentatively set type to image for this path

        if self.USE_AI_IMAGE_GENERATION and self.openai_api_key:
            max_retries = 2
            for attempt in range(max_retries):
                try:
                    logger.info(f"Attempt {attempt+1}: DALL-E ({self.dalle_model}) for: {image_prompt_text[:100]}...")
                    client = openai.OpenAI(api_key=self.openai_api_key, timeout=90.0)
                    response = client.images.generate(model=self.dalle_model, prompt=image_prompt_text, n=1, size=self.image_size_dalle3, quality="hd", response_format="url", style="vivid")
                    image_url = response.data[0].url; revised_prompt = getattr(response.data[0], 'revised_prompt', None)
                    if revised_prompt: logger.info(f"DALL-E 3 revised_prompt: {revised_prompt[:100]}...")
                    image_response = requests.get(image_url, timeout=120); image_response.raise_for_status()
                    img_data = Image.open(io.BytesIO(image_response.content));
                    if img_data.mode != 'RGB': img_data = img_data.convert('RGB')
                    img_data.save(filepath); logger.info(f"AI Image (DALL-E) saved: {filepath}");
                    asset_info = {'path': filepath, 'type': 'image', 'error': False, 'prompt_used': image_prompt_text, 'revised_prompt': revised_prompt}
                    return asset_info
                except openai.RateLimitError as e_rate: logger.warning(f"OpenAI Rate Limit: {e_rate}. Retrying..."); time.sleep(5 * (attempt + 1)); asset_info['error_message'] = str(e_rate)
                except openai.APIError as e_api: logger.error(f"OpenAI API Error: {e_api}"); asset_info['error_message'] = str(e_api); break
                except requests.exceptions.RequestException as e_req: logger.error(f"Requests Error (DALL-E download): {e_req}"); asset_info['error_message'] = str(e_req); break
                except Exception as e_gen: logger.error(f"Generic error (DALL-E gen): {e_gen}", exc_info=True); asset_info['error_message'] = str(e_gen); break
                if attempt == max_retries - 1: logger.error("Max retries for DALL-E RateLimitError."); break
            if asset_info['error']: logger.warning("DALL-E generation failed. Trying Pexels fallback...")
        
        if self.USE_PEXELS and (asset_info['error'] or not (self.USE_AI_IMAGE_GENERATION and self.openai_api_key)): # Try Pexels if DALL-E failed or disabled
            pexels_query_text = scene_data.get('pexels_search_query_감독', f"{scene_data.get('emotional_beat','')} {scene_data.get('setting_description','')}")
            pexels_path = self._search_pexels_image(pexels_query_text, image_filename_with_ext)
            if pexels_path:
                asset_info = {'path': pexels_path, 'type': 'image', 'error': False, 'prompt_used': f"Pexels: {pexels_query_text}"}
                return asset_info
            asset_info['error_message'] = (asset_info.get('error_message', "") + " Pexels search also failed or disabled.").strip()
            if not asset_info['error']: logger.warning("Pexels search failed or disabled.") # If DALL-E wasn't even tried

        # Fallback to placeholder if all else fails
        if asset_info['error']: # Only create placeholder if previous steps failed
            logger.warning("All generation methods failed. Using placeholder image.")
            placeholder_prompt_text = asset_info.get('prompt_used', image_prompt_text) # Use the prompt that was attempted
            placeholder_path = self._create_placeholder_image_content(f"[Fallback Placeholder] {placeholder_prompt_text[:100]}...", image_filename_with_ext)
            if placeholder_path:
                asset_info = {'path': placeholder_path, 'type': 'image', 'error': False, 'prompt_used': placeholder_prompt_text}
                return asset_info
            else: # Final failure
                asset_info['error_message'] = (asset_info.get('error_message', "") + " Placeholder creation also failed.").strip()
        return asset_info # Return whatever state asset_info is in (could be error=True)

    def generate_narration_audio(self, text_to_narrate, output_filename="narration_overall.mp3"):
        # (No significant changes from your previous correct version, ensure error handling is robust)
        if not self.USE_ELEVENLABS or not self.elevenlabs_client or not text_to_narrate:
            logger.info("ElevenLabs conditions not met (API key, client init, or text). Skipping audio.")
            return None
        audio_filepath = os.path.join(self.output_dir, output_filename)
        try:
            logger.info(f"Generating ElevenLabs audio (Voice ID: {self.elevenlabs_voice_id}) for: {text_to_narrate[:70]}...")
            audio_stream_method = None
            if hasattr(self.elevenlabs_client, 'text_to_speech') and hasattr(self.elevenlabs_client.text_to_speech, 'stream'):
                audio_stream_method = self.elevenlabs_client.text_to_speech.stream; logger.info("Using elevenlabs_client.text_to_speech.stream()")
            elif hasattr(self.elevenlabs_client, 'generate_stream') : audio_stream_method = self.elevenlabs_client.generate_stream; logger.info("Using elevenlabs_client.generate_stream()")
            elif hasattr(self.elevenlabs_client, 'generate'):
                logger.info("Using elevenlabs_client.generate() (non-streaming).")
                voice_param = Voice(voice_id=str(self.elevenlabs_voice_id), settings=self.elevenlabs_voice_settings) if Voice and self.elevenlabs_voice_settings else str(self.elevenlabs_voice_id)
                audio_bytes = self.elevenlabs_client.generate(text=text_to_narrate, voice=voice_param, model="eleven_multilingual_v2")
                with open(audio_filepath, "wb") as f: f.write(audio_bytes)
                logger.info(f"ElevenLabs audio (non-streamed) saved: {audio_filepath}"); return audio_filepath
            else: logger.error("No recognized audio generation method found on ElevenLabs client."); return None

            if audio_stream_method:
                voice_param_for_stream = {"voice_id": str(self.elevenlabs_voice_id)}
                if self.elevenlabs_voice_settings and hasattr(self.elevenlabs_voice_settings, 'model_dump'): # Pydantic v2 for elevenlabs sdk >=1.0
                     voice_param_for_stream["voice_settings"] = self.elevenlabs_voice_settings.model_dump()
                elif self.elevenlabs_voice_settings and hasattr(self.elevenlabs_voice_settings, 'dict'): # Pydantic v1 for elevenlabs sdk <1.0
                     voice_param_for_stream["voice_settings"] = self.elevenlabs_voice_settings.dict()
                elif self.elevenlabs_voice_settings : voice_param_for_stream["voice_settings"] = self.elevenlabs_voice_settings
                
                audio_data_iterator = audio_stream_method(text=text_to_narrate, model_id="eleven_multilingual_v2", **voice_param_for_stream)
                with open(audio_filepath, "wb") as f:
                    for chunk in audio_data_iterator:
                        if chunk: f.write(chunk)
                logger.info(f"ElevenLabs audio (streamed) saved: {audio_filepath}"); return audio_filepath
        except AttributeError as ae: logger.error(f"AttributeError with ElevenLabs client: {ae}. SDK method/params might be different.", exc_info=True)
        except Exception as e: logger.error(f"Error generating ElevenLabs audio: {e}", exc_info=True)
        return None

    def assemble_animatic_from_assets(self, asset_data_list, overall_narration_path=None, output_filename="final_video.mp4", fps=24):
        if not asset_data_list:
            logger.warning("No asset data provided for animatic assembly.")
            return None

        processed_moviepy_clips = []
        narration_audio_clip = None
        final_composite_clip = None # Renamed to avoid conflict in finally block
        total_video_duration_from_assets = sum(item.get('duration', 4.5) for item in asset_data_list)
        logger.info(f"Assembling animatic from {len(asset_data_list)} assets. Target frame: {self.video_frame_size}. Approx total duration: {total_video_duration_from_assets:.2f}s.")

        for i, asset_info in enumerate(asset_data_list):
            asset_path = asset_info.get('path')
            asset_type = asset_info.get('type')
            target_scene_duration = asset_info.get('duration', 4.5)
            scene_num = asset_info.get('scene_num', i + 1)
            key_action = asset_info.get('key_action', '')

            logger.info(f"Processing Scene {scene_num}: Path='{asset_path}', Type='{asset_type}', Target Duration='{target_scene_duration}'s")

            if not (asset_path and os.path.exists(asset_path)):
                logger.warning(f"Asset not found for Scene {scene_num}: {asset_path}. Skipping.")
                continue
            if target_scene_duration <= 0:
                logger.warning(f"Scene {scene_num} has invalid duration ({target_scene_duration}s). Skipping.")
                continue

            current_clip_for_scene = None
            try:
                if asset_type == 'image':
                    logger.debug(f"S{scene_num}: Loading image asset from {asset_path}")
                    pil_img = Image.open(asset_path)
                    logger.debug(f"S{scene_num}: Image loaded. Mode: {pil_img.mode}, Size: {pil_img.size}")

                    # Ensure image is RGBA for consistent pasting, then convert to RGB for MoviePy
                    if pil_img.mode != 'RGBA':
                        pil_img = pil_img.convert('RGBA') # Convert to RGBA to handle transparency uniformly

                    img_copy = pil_img.copy()
                    resample_filter = Image.Resampling.LANCZOS if hasattr(Image.Resampling, 'LANCZOS') else (Image.ANTIALIAS if hasattr(Image, 'ANTIALIAS') else Image.BILINEAR)
                    img_copy.thumbnail(self.video_frame_size, resample_filter)
                    logger.debug(f"S{scene_num}: Image thumbnailed to: {img_copy.size}")

                    # Create an RGBA canvas, paste the (potentially RGBA) image onto it
                    canvas_rgba = Image.new('RGBA', self.video_frame_size, (0, 0, 0, 0)) # Fully transparent
                    xo, yo = (self.video_frame_size[0] - img_copy.width) // 2, (self.video_frame_size[1] - img_copy.height) // 2
                    canvas_rgba.paste(img_copy, (xo, yo), img_copy) # Paste using image's own alpha
                    logger.debug(f"S{scene_num}: Image pasted onto RGBA canvas.")

                    # Now create a final RGB canvas and paste the RGBA canvas onto it, effectively blending alpha
                    final_rgb_canvas = Image.new("RGB", self.video_frame_size, (random.randint(0,5), random.randint(0,5), random.randint(0,5))) # Dark background
                    final_rgb_canvas.paste(canvas_rgba, mask=canvas_rgba.split()[3]) # Use alpha channel of canvas_rgba as mask
                    
                    debug_canvas_path = os.path.join(self.output_dir, f"debug_final_rgb_canvas_scene_{scene_num}.png")
                    try: final_rgb_canvas.save(debug_canvas_path); logger.info(f"DEBUG: Saved final RGB canvas for scene {scene_num} to {debug_canvas_path}")
                    except Exception as e_save_canvas: logger.error(f"DEBUG: Failed to save final RGB canvas for scene {scene_num}: {e_save_canvas}")

                    frame_np = np.array(final_rgb_canvas)
                    logger.debug(f"S{scene_num}: Final RGB canvas to NumPy. Shape: {frame_np.shape}, Dtype: {frame_np.dtype}")
                    if frame_np.size == 0: logger.error(f"S{scene_num}: NumPy array for ImageClip is empty! Skipping."); continue
                    
                    current_clip_base = ImageClip(frame_np, transparent=False, ismask=False).set_duration(target_scene_duration)
                    logger.debug(f"S{scene_num}: Base ImageClip created.")

                    current_clip_for_scene = current_clip_base
                    try: # Ken Burns
                        end_scale = random.uniform(1.03, 1.08)
                        current_clip_for_scene = current_clip_base.fx(vfx.resize, lambda t: 1 + (end_scale - 1) * (t / target_scene_duration)).set_position('center')
                        logger.debug(f"S{scene_num}: Ken Burns effect applied.")
                    except Exception as e_fx: logger.error(f"S{scene_num}: Ken Burns error: {e_fx}. Using static.", exc_info=False); current_clip_for_scene = current_clip_base

                elif asset_type == 'video':
                    logger.debug(f"S{scene_num}: Loading video asset from {asset_path}")
                    # Ensure target_resolution is (height, width) for VideoFileClip resizing parameter
                    source_video_clip = VideoFileClip(asset_path, target_resolution=(self.video_frame_size[1], self.video_frame_size[0]) if self.video_frame_size else None)
                    
                    temp_clip = source_video_clip # Work with a temporary variable
                    if source_video_clip.duration > target_scene_duration:
                        temp_clip = source_video_clip.subclip(0, target_scene_duration)
                    elif source_video_clip.duration < target_scene_duration:
                        if target_scene_duration / source_video_clip.duration > 1.5 and source_video_clip.duration > 0.1:
                             temp_clip = source_video_clip.loop(duration=target_scene_duration)
                        else: # Play once, MoviePy will pad if needed during concatenation if durations differ
                             temp_clip = source_video_clip.set_duration(source_video_clip.duration) # Keep its own duration
                             logger.info(f"Video clip for S{scene_num} ({source_video_clip.duration:.2f}s) is shorter than target animatic duration ({target_scene_duration:.2f}s). It will play once at its native length.")
                    
                    # Crucially, ensure the clip used in concatenation has the target_scene_duration
                    current_clip_for_scene = temp_clip.set_duration(target_scene_duration)

                    if current_clip_for_scene.size != list(self.video_frame_size):
                        logger.debug(f"S{scene_num}: Resizing video clip from {current_clip_for_scene.size} to {self.video_frame_size}")
                        current_clip_for_scene = current_clip_for_scene.resize(self.video_frame_size)
                    
                    # Only close source_video_clip if it's different from what we are keeping (e.g., after subclip)
                    # And if it's not the same object as current_clip_for_scene
                    if source_video_clip is not current_clip_for_scene and hasattr(source_video_clip, 'close'):
                        source_video_clip.close()
                    logger.debug(f"S{scene_num}: Video asset processed. Final duration for scene: {current_clip_for_scene.duration:.2f}s")

                else: logger.warning(f"S{scene_num}: Unknown asset type '{asset_type}'. Skipping."); continue

                if current_clip_for_scene and key_action: # Add text overlay
                    logger.debug(f"S{scene_num}: Adding text overlay: '{key_action}'")
                    text_overlay_duration = min(target_scene_duration - 0.5, target_scene_duration * 0.8) if target_scene_duration > 0.5 else target_scene_duration
                    text_overlay_start = (target_scene_duration - text_overlay_duration) / 2.0
                    if text_overlay_duration > 0:
                        txt_clip = TextClip(f"Scene {scene_num}\n{key_action}",
                                            fontsize=self.video_overlay_font_size, color=self.video_overlay_font_color,
                                            font=self.video_overlay_font, bg_color='rgba(10,10,20,0.7)',
                                            method='caption', align='West', size=(self.video_frame_size[0] * 0.9, None),
                                            kerning=-1, stroke_color='black', stroke_width=1.5
                                           ).set_duration(text_overlay_duration).set_start(text_overlay_start).set_position(('center', 0.92), relative=True)
                        current_clip_for_scene = CompositeVideoClip([current_clip_for_scene, txt_clip], size=self.video_frame_size, use_bgclip=True)
                        logger.debug(f"S{scene_num}: Text overlay composited.")
                
                if current_clip_for_scene: processed_moviepy_clips.append(current_clip_for_scene); logger.info(f"S{scene_num}: Asset successfully processed and added to final list.")
            except Exception as e: logger.error(f"Error processing asset for Scene {scene_num} ({asset_path}): {e}", exc_info=True)
            finally: # Ensure individual clips are closed if they were opened and an error occurred mid-processing
                if current_clip_for_scene and asset_type == 'video' and hasattr(current_clip_for_scene, 'reader') and current_clip_for_scene.reader:
                    if hasattr(current_clip_for_scene, 'close'): current_clip_for_scene.close()


        if not processed_moviepy_clips: logger.warning("No MoviePy clips processed. Aborting animatic assembly."); return None
        
        transition_duration = 0.75
        try:
            if len(processed_moviepy_clips) > 1: final_composite_clip = concatenate_videoclips(processed_moviepy_clips, padding=-transition_duration, method="compose")
            elif processed_moviepy_clips: final_composite_clip = processed_moviepy_clips[0]
            else: logger.error("No clips for final concatenation."); return None

            if final_composite_clip.duration > transition_duration * 2: final_composite_clip = final_composite_clip.fx(vfx.fadein, transition_duration).fx(vfx.fadeout, transition_duration)
            elif final_composite_clip.duration > 0: final_composite_clip = final_composite_clip.fx(vfx.fadein, min(transition_duration, final_composite_clip.duration/2.0))

            if overall_narration_path and os.path.exists(overall_narration_path) and final_composite_clip.duration > 0:
                try:
                    narration_audio_clip = AudioFileClip(overall_narration_path)
                    if narration_audio_clip.duration < final_composite_clip.duration:
                        logger.info(f"Narration ({narration_audio_clip.duration:.2f}s) shorter than visuals ({final_composite_clip.duration:.2f}s). Trimming video.")
                        final_composite_clip = final_composite_clip.subclip(0, narration_audio_clip.duration)
                    final_composite_clip = final_composite_clip.set_audio(narration_audio_clip); logger.info("Overall narration added.")
                except Exception as e: logger.error(f"Adding narration error: {e}", exc_info=True)
            elif final_composite_clip.duration <= 0 : logger.warning("Video has no duration. Audio not added.")
            
            if final_composite_clip and final_composite_clip.duration > 0:
                output_path = os.path.join(self.output_dir, output_filename)
                logger.info(f"Writing final animatic: {output_path} (Duration: {final_composite_clip.duration:.2f}s)")
                final_composite_clip.write_videofile(output_path, fps=fps, codec='libx264', preset='medium', audio_codec='aac',
                                            temp_audiofile=os.path.join(self.output_dir, f'temp-audio-{os.urandom(4).hex()}.m4a'),
                                            remove_temp=True, threads=os.cpu_count() or 2, logger='bar', bitrate="5000k")
                logger.info(f"Animatic created: {output_path}"); return output_path
            else: logger.error("Final animatic clip invalid. Not writing file."); return None
        except Exception as e: logger.error(f"Animatic writing error: {e}", exc_info=True); return None
        finally:
            for clip_obj in processed_moviepy_clips:
                if hasattr(clip_obj, 'close'): clip_obj.close()
            if narration_audio_clip and hasattr(narration_audio_clip, 'close'): narration_audio_clip.close()
            if final_composite_clip and hasattr(final_composite_clip, 'close'): final_composite_clip.close()