# core/visual_engine.py from PIL import Image, ImageDraw, ImageFont, ImageOps import base64 # For Data URI conversion # --- MONKEY PATCH --- try: if hasattr(Image, 'Resampling') and hasattr(Image.Resampling, 'LANCZOS'): if not hasattr(Image, 'ANTIALIAS'): Image.ANTIALIAS = Image.Resampling.LANCZOS elif hasattr(Image, 'LANCZOS'): if not hasattr(Image, 'ANTIALIAS'): Image.ANTIALIAS = Image.LANCZOS elif not hasattr(Image, 'ANTIALIAS'): print("WARNING: Pillow ANTIALIAS/Resampling issue.") except Exception as e_mp: print(f"WARNING: ANTIALIAS patch error: {e_mp}") 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 import mimetypes # For Data URI logger = logging.getLogger(__name__) logger.setLevel(logging.INFO) # --- SERVICE CLIENT IMPORTS --- 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.") except Exception as e_eleven: logger.warning(f"ElevenLabs client import failed: {e_eleven}. Audio disabled.") RUNWAYML_SDK_IMPORTED = False; RunwayMLAPIClient = None try: from runwayml import RunwayML as ImportedRunwayMLClient RunwayMLAPIClient = ImportedRunwayMLClient RUNWAYML_SDK_IMPORTED = True logger.info("RunwayML SDK imported successfully.") except ImportError: logger.warning("RunwayML SDK not found (pip install runwayml). RunwayML video generation will be disabled.") except Exception as e_runway_sdk: logger.warning(f"Error importing RunwayML SDK: {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 = "DejaVuSans-Bold.ttf" font_paths_to_try = [ self.font_filename, "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf", "/usr/share/fonts/truetype/liberation/LiberationSans-Bold.ttf", "/System/Library/Fonts/Supplemental/Arial.ttf", "C:/Windows/Fonts/arial.ttf", f"/usr/local/share/fonts/truetype/mycustomfonts/arial.ttf"] 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 = 'DejaVu-Sans-Bold' try: self.font = ImageFont.truetype(self.font_path_pil, self.font_size_pil) if self.font_path_pil else ImageFont.load_default() if self.font_path_pil: logger.info(f"Pillow font: {self.font_path_pil}.") else: logger.warning("Default Pillow font."); self.font_size_pil = 10 except IOError as e_font: logger.error(f"Pillow font IOError: {e_font}. 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) 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 if RUNWAYML_SDK_IMPORTED and RunwayMLAPIClient: try: if os.getenv("RUNWAYML_API_SECRET"): self.runway_client = RunwayMLAPIClient() logger.info("RunwayML Client initialized using RUNWAYML_API_SECRET env var.") # else: # No explicit else, will be handled by set_runway_api_key if key provided later except Exception as e_runway_init: logger.error(f"Failed to initialize RunwayML client during __init__: {e_runway_init}", exc_info=True) 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.'}") 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 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.'}") def set_runway_api_key(self, k): self.runway_api_key = k if k: if RUNWAYML_SDK_IMPORTED and RunwayMLAPIClient: if not self.runway_client: try: if not os.getenv("RUNWAYML_API_SECRET"): logger.info("Setting RUNWAYML_API_SECRET environment variable from provided key for SDK.") os.environ["RUNWAYML_API_SECRET"] = k self.runway_client = RunwayMLAPIClient() self.USE_RUNWAYML = True logger.info("RunwayML Client initialized successfully via set_runway_api_key.") except Exception as e_client_init: logger.error(f"RunwayML Client initialization failed in set_runway_api_key: {e_client_init}", exc_info=True) self.USE_RUNWAYML = False else: # Client already initialized self.USE_RUNWAYML = True; logger.info("RunwayML Client was already initialized.") else: logger.warning("RunwayML SDK not imported. API key set, but integration requires SDK."); self.USE_RUNWAYML = False else: self.USE_RUNWAYML = False; logger.info("RunwayML Disabled (no API key).") def _image_to_data_uri(self, image_path): try: mime_type, _ = mimetypes.guess_type(image_path) if not mime_type: ext = os.path.splitext(image_path)[1].lower() if ext == ".png": mime_type = "image/png" elif ext in [".jpg", ".jpeg"]: mime_type = "image/jpeg" else: mime_type = "application/octet-stream"; logger.warning(f"Unknown MIME for {image_path}, using {mime_type}.") with open(image_path, "rb") as image_file: encoded_string = base64.b64encode(image_file.read()).decode('utf-8') data_uri = f"data:{mime_type};base64,{encoded_string}" logger.debug(f"Data URI for {image_path} (first 100): {data_uri[:100]}"); return data_uri except Exception as e: logger.error(f"Error converting {image_path} to data URI: {e}", exc_info=True); return None def _map_resolution_to_runway_ratio(self, width, height): # Based on Gen-4 supported ratios: "1280:720", "720:1280", "1104:832", "832:1104", "960:960", "1584:672" ratio_str = f"{width}:{height}" supported_ratios = ["1280:720", "720:1280", "1104:832", "832:1104", "960:960", "1584:672"] if ratio_str in supported_ratios: return ratio_str logger.warning(f"Resolution {ratio_str} not directly supported by Gen-4. Defaulting to 1280:720.") return "1280:720" def _get_text_dimensions(self,text_content,font_obj): # (Corrected version from previous, assuming font_obj.size exists or font_size_pil is fallback) default_char_height = getattr(font_obj, 'size', self.font_size_pil) if not text_content: return 0, default_char_height try: if hasattr(font_obj,'getbbox'): # Pillow 8.0.0+ bbox=font_obj.getbbox(text_content);w=bbox[2]-bbox[0];h=bbox[3]-bbox[1] return w, h if h > 0 else default_char_height elif hasattr(font_obj,'getsize'): # Older Pillow w,h=font_obj.getsize(text_content) return w, h if h > 0 else default_char_height else: # Fallback if no standard method (should not happen for ImageFont) return int(len(text_content)*default_char_height*0.6),int(default_char_height*1.2) 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) # Fallback to global default def _create_placeholder_image_content(self,text_description,filename,size=None): # <<< THIS IS THE CORRECTED METHOD >>> 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 Image)" words=text_description.split();current_line="" for word_idx, word in enumerate(words): # Add space correctly, not after the very last word of the text prospective_line_addition = word + (" " if word_idx < len(words) - 1 else "") test_line = current_line + prospective_line_addition current_line_width, _ = self._get_text_dimensions(test_line, self.font) if current_line_width == 0 and test_line.strip(): # Estimate if Pillow returns 0 current_line_width = len(test_line) * (self.font_size_pil * 0.6) if current_line_width <= max_w: current_line = test_line else: # Word doesn't fit if current_line.strip(): # Add previous line if it had content lines.append(current_line.strip()) current_line = prospective_line_addition # Start new line with current word (plus its space if not last) # If the word itself is too long for a line, it will just be one long line. # Pillow's d.text will handle overflow if text anchor isn't 'lt' (left-top). # For centered text, it might go off-canvas; more complex word splitting needed for that. if current_line.strip(): # Add any remaining part lines.append(current_line.strip()) if not lines and text_description: avg_char_width, _ = self._get_text_dimensions("W", self.font) if avg_char_width == 0: avg_char_width = self.font_size_pil * 0.6 # Estimate chars_per_line = int(max_w / avg_char_width) if avg_char_width > 0 else 20 lines.append(text_description[:chars_per_line] + ("..." if len(text_description) > chars_per_line else "")) elif not lines: lines.append("(Placeholder 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) if line_w == 0 and line_content.strip(): line_w = len(line_content) * (self.font_size_pil * 0.6) x_text=(size[0]-line_w)/2.0 try: d.text((x_text,y_text),line_content,font=self.font,fill=(200,200,180)) except Exception as e_draw: logger.error(f"Pillow d.text error: {e_draw} for line '{line_content}'") y_text+=single_line_h+2 if i==6 and max_lines_to_display > 7: try: d.text((x_text,y_text),"...",font=self.font,fill=(200,200,180)) except Exception as e_ellipsis: logger.error(f"Pillow d.text ellipsis error: {e_ellipsis}") 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, q, ofnb): # (Keep as before) if not self.USE_PEXELS or not self.pexels_api_key: return None; h={"Authorization":self.pexels_api_key};p={"query":q,"per_page":1,"orientation":"landscape","size":"large2x"} pfn=ofnb.replace(".png",f"_pexels_{random.randint(1000,9999)}.jpg").replace(".mp4",f"_pexels_{random.randint(1000,9999)}.jpg");fp=os.path.join(self.output_dir,pfn) try: logger.info(f"Pexels search: '{q}'");eq=" ".join(q.split()[:5]);p["query"]=eq;r=requests.get("https://api.pexels.com/v1/search",headers=h,params=p,timeout=20) r.raise_for_status();d=r.json() if d.get("photos") and len(d["photos"])>0:pu=d["photos"][0]["src"]["large2x"];ir=requests.get(pu,timeout=60);ir.raise_for_status();id_img=Image.open(io.BytesIO(ir.content)) if id_img.mode!='RGB':id_img=id_img.convert('RGB');id_img.save(fp);logger.info(f"Pexels saved: {fp}");return fp # Fixed id to id_img else: id_img.save(fp);logger.info(f"Pexels saved (was RGB): {fp}");return fp # Save even if already RGB else: logger.info(f"No Pexels for: '{eq}'") # This else was misplaced except Exception as e:logger.error(f"Pexels error ('{q}'): {e}",exc_info=True);return None # Fixed indent def _generate_video_clip_with_runwayml(self, text_prompt_for_motion, input_image_path, scene_identifier_filename_base, target_duration_seconds=5): # (Updated RunwayML integration) if not self.USE_RUNWAYML or not self.runway_client: logger.warning("RunwayML not enabled/client not init. Skip video."); return None if not input_image_path or not os.path.exists(input_image_path): logger.error(f"Runway Gen-4 needs input image. Path invalid: {input_image_path}"); return None image_data_uri = self._image_to_data_uri(input_image_path) if not image_data_uri: return None runway_duration = 10 if target_duration_seconds > 7 else 5 runway_ratio_str = self._map_resolution_to_runway_ratio(self.video_frame_size[0], self.video_frame_size[1]) output_video_filename = scene_identifier_filename_base.replace(".png", f"_runway_gen4_d{runway_duration}s.mp4") output_video_filepath = os.path.join(self.output_dir, output_video_filename) logger.info(f"Runway Gen-4 task: motion='{text_prompt_for_motion[:100]}...', img='{os.path.basename(input_image_path)}', dur={runway_duration}s, ratio='{runway_ratio_str}'") try: task = self.runway_client.image_to_video.create(model='gen4_turbo', prompt_image=image_data_uri, prompt_text=text_prompt_for_motion, duration=runway_duration, ratio=runway_ratio_str) logger.info(f"Runway Gen-4 task ID: {task.id}. Polling...") poll_interval=10; max_polls=36 # Max 6 mins for _ in range(max_polls): time.sleep(poll_interval); task_details = self.runway_client.tasks.retrieve(id=task.id) logger.info(f"Runway task {task.id} status: {task_details.status}") if task_details.status == 'SUCCEEDED': output_url = getattr(getattr(task_details, 'output', None), 'url', None) or \ (getattr(task_details, 'artifacts', None) and task_details.artifacts[0].url if task_details.artifacts and hasattr(task_details.artifacts[0], 'url') else None) or \ (getattr(task_details, 'artifacts', None) and task_details.artifacts[0].download_url if task_details.artifacts and hasattr(task_details.artifacts[0], 'download_url') else None) if not output_url: logger.error(f"Runway task {task.id} SUCCEEDED, but no output URL in details: {vars(task_details) if hasattr(task_details, '__dict__') else task_details}"); return None logger.info(f"Runway task {task.id} SUCCEEDED. Downloading from: {output_url}") video_response = requests.get(output_url, stream=True, timeout=300); video_response.raise_for_status() with open(output_video_filepath, 'wb') as f: for chunk in video_response.iter_content(chunk_size=8192): f.write(chunk) logger.info(f"Runway Gen-4 video saved: {output_video_filepath}"); return output_video_filepath elif task_details.status in ['FAILED', 'ABORTED']: em = getattr(task_details,'error_message',None) or getattr(getattr(task_details,'output',None),'error', "Unknown error") logger.error(f"Runway task {task.id} status: {task_details.status}. Error: {em}"); return None logger.warning(f"Runway task {task.id} timed out."); return None except AttributeError as ae: logger.error(f"RunwayML SDK AttributeError: {ae}. SDK/methods might differ.", exc_info=True); return None except Exception as e: logger.error(f"Runway Gen-4 API error: {e}", exc_info=True); return None def _create_placeholder_video_content(self, td, fn, dur=4, sz=None): # Generic placeholder if input_image not available if sz is None: sz = self.video_frame_size; fp = os.path.join(self.output_dir, fn); tc = None try: tc = TextClip(td, fontsize=50, color='white', font=self.video_overlay_font, bg_color='black', size=sz, method='caption').set_duration(dur); tc.write_videofile(fp, fps=24, codec='libx264', preset='ultrafast', logger=None, threads=2); logger.info(f"Generic placeholder video: {fp}"); return fp except Exception as e: logger.error(f"Generic placeholder video error {fp}: {e}", exc_info=True); return None finally: if tc and hasattr(tc, 'close'): tc.close() def generate_scene_asset(self, image_generation_prompt_text, motion_prompt_text_for_video, scene_data, scene_identifier_filename_base, generate_as_video_clip=False, runway_target_duration=5): # (Logic mostly as before, ensuring base image is robustly generated first) base_name, _ = os.path.splitext(scene_identifier_filename_base) asset_info = {'path': None, 'type': 'none', 'error': True, 'prompt_used': image_generation_prompt_text, 'error_message': 'Asset generation init failed'} input_image_for_runway_path = None base_image_filename = base_name + ("_base_for_video.png" if generate_as_video_clip else ".png") base_image_filepath = os.path.join(self.output_dir, base_image_filename) # Attempt base image generation if self.USE_AI_IMAGE_GENERATION and self.openai_api_key: # DALL-E max_r, att_n = 2,0; for att_n in range(max_r): try:logger.info(f"Att {att_n+1} DALL-E (base img): {image_generation_prompt_text[:70]}...");cl=openai.OpenAI(api_key=self.openai_api_key,timeout=90.0);r=cl.images.generate(model=self.dalle_model,prompt=image_generation_prompt_text,n=1,size=self.image_size_dalle3,quality="hd",response_format="url",style="vivid");iu=r.data[0].url;rp=getattr(r.data[0],'revised_prompt',None); if rp:logger.info(f"DALL-E revised: {rp[:70]}...");ir=requests.get(iu,timeout=120);ir.raise_for_status();id_img=Image.open(io.BytesIO(ir.content)); if id_img.mode!='RGB':id_img=id_img.convert('RGB');id_img.save(base_image_filepath);logger.info(f"DALL-E base img saved: {base_image_filepath}");input_image_for_runway_path=base_image_filepath;asset_info={'path':base_image_filepath,'type':'image','error':False,'prompt_used':image_generation_prompt_text,'revised_prompt':rp};break except openai.RateLimitError as e:logger.warning(f"OpenAI RateLimit {att_n+1}:{e}.Retry...");time.sleep(5*(att_n+1));asset_info['error_message']=str(e) except Exception as e:logger.error(f"DALL-E base img error:{e}",exc_info=True);asset_info['error_message']=str(e);break if asset_info['error']:logger.warning(f"DALL-E failed after {att_n+1} attempts for base img.") if asset_info['error'] and self.USE_PEXELS: # Pexels Fallback logger.info("Trying Pexels for base img.");pqt=scene_data.get('pexels_search_query_감독',f"{scene_data.get('emotional_beat','')} {scene_data.get('setting_description','')}");pp=self._search_pexels_image(pqt,base_image_filename); if pp:input_image_for_runway_path=pp;asset_info={'path':pp,'type':'image','error':False,'prompt_used':f"Pexels:{pqt}"} else:current_em=asset_info.get('error_message',"");asset_info['error_message']=(current_em+" Pexels failed for base.").strip() if asset_info['error']: # Placeholder Fallback logger.warning("Base img (DALL-E/Pexels) failed. Using placeholder.");ppt=asset_info.get('prompt_used',image_generation_prompt_text);php=self._create_placeholder_image_content(f"[Base Placeholder]{ppt[:70]}...",base_image_filename); if php:input_image_for_runway_path=php;asset_info={'path':php,'type':'image','error':False,'prompt_used':ppt} else:current_em=asset_info.get('error_message',"");asset_info['error_message']=(current_em+" Base placeholder failed.").strip() if generate_as_video_clip: # Now attempt RunwayML if requested if not input_image_for_runway_path:logger.error("RunwayML video: base img failed.");asset_info['error']=True;asset_info['error_message']=(asset_info.get('error_message',"")+" Base img miss, Runway abort.").strip();asset_info['type']='none';return asset_info if self.USE_RUNWAYML: logger.info(f"Runway Gen-4 video for {base_name} using base: {input_image_for_runway_path}") video_path=self._generate_video_clip_with_runwayml(motion_prompt_text_for_video,input_image_for_runway_path,base_name,runway_target_duration) if video_path and os.path.exists(video_path):asset_info={'path':video_path,'type':'video','error':False,'prompt_used':motion_prompt_text_for_video,'base_image_path':input_image_for_runway_path} else:logger.warning(f"RunwayML video failed for {base_name}. Fallback to base img.");asset_info['error']=True;asset_info['error_message']=(asset_info.get('error_message',"Base img ok.")+" RunwayML video fail; use base img.").strip();asset_info['path']=input_image_for_runway_path;asset_info['type']='image';asset_info['prompt_used']=image_generation_prompt_text else:logger.warning("RunwayML selected but disabled. Use base img.");asset_info['error']=True;asset_info['error_message']=(asset_info.get('error_message',"Base img ok.")+" RunwayML disabled; use base img.").strip();asset_info['path']=input_image_for_runway_path;asset_info['type']='image';asset_info['prompt_used']=image_generation_prompt_text return asset_info # Return image info if not video, or video result def generate_narration_audio(self, text_to_narrate, output_filename="narration_overall.mp3"): # (Keep as before) if not self.USE_ELEVENLABS or not self.elevenlabs_client or not text_to_narrate: logger.info("11L skip."); return None; afp=os.path.join(self.output_dir,output_filename) try: logger.info(f"11L audio (Voice:{self.elevenlabs_voice_id}): {text_to_narrate[:70]}..."); asm=None if hasattr(self.elevenlabs_client,'text_to_speech')and hasattr(self.elevenlabs_client.text_to_speech,'stream'):asm=self.elevenlabs_client.text_to_speech.stream;logger.info("Using 11L .text_to_speech.stream()") elif hasattr(self.elevenlabs_client,'generate_stream'):asm=self.elevenlabs_client.generate_stream;logger.info("Using 11L .generate_stream()") elif hasattr(self.elevenlabs_client,'generate'):logger.info("Using 11L .generate()");vp=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);ab=self.elevenlabs_client.generate(text=text_to_narrate,voice=vp,model="eleven_multilingual_v2"); with open(afp,"wb")as f:f.write(ab);logger.info(f"11L audio (non-stream): {afp}");return afp else:logger.error("No 11L audio method.");return None if asm:vps={"voice_id":str(self.elevenlabs_voice_id)} if self.elevenlabs_voice_settings: if hasattr(self.elevenlabs_voice_settings,'model_dump'):vps["voice_settings"]=self.elevenlabs_voice_settings.model_dump() elif hasattr(self.elevenlabs_voice_settings,'dict'):vps["voice_settings"]=self.elevenlabs_voice_settings.dict() else:vps["voice_settings"]=self.elevenlabs_voice_settings adi=asm(text=text_to_narrate,model_id="eleven_multilingual_v2",**vps) with open(afp,"wb")as f: for c in adi: if c:f.write(c) logger.info(f"11L audio (stream): {afp}");return afp except Exception as e:logger.error(f"11L audio error: {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): # (Keep as in the version that has the robust image processing, C-contiguous array, and debug image saves) if not asset_data_list: logger.warning("No assets for animatic."); return None processed_clips = []; narration_clip = None; final_clip = None logger.info(f"Assembling from {len(asset_data_list)} assets. Frame: {self.video_frame_size}.") for i, asset_info in enumerate(asset_data_list): asset_path, asset_type, scene_dur = asset_info.get('path'), asset_info.get('type'), asset_info.get('duration', 4.5) scene_num, key_action = asset_info.get('scene_num', i + 1), asset_info.get('key_action', '') logger.info(f"S{scene_num}: Path='{asset_path}', Type='{asset_type}', Dur='{scene_dur}'s") if not (asset_path and os.path.exists(asset_path)): logger.warning(f"S{scene_num}: Not found '{asset_path}'. Skip."); continue if scene_dur <= 0: logger.warning(f"S{scene_num}: Invalid duration ({scene_dur}s). Skip."); continue current_scene_mvpy_clip = None try: if asset_type == 'image': pil_img = Image.open(asset_path); logger.debug(f"S{scene_num}: Loaded img. Mode:{pil_img.mode}, Size:{pil_img.size}") img_rgba = pil_img.convert('RGBA') if pil_img.mode != 'RGBA' else pil_img.copy() thumb = img_rgba.copy(); rf = Image.Resampling.LANCZOS if hasattr(Image.Resampling,'LANCZOS') else Image.BILINEAR; thumb.thumbnail(self.video_frame_size,rf) cv_rgba = Image.new('RGBA',self.video_frame_size,(0,0,0,0)); xo,yo=(self.video_frame_size[0]-thumb.width)//2,(self.video_frame_size[1]-thumb.height)//2 cv_rgba.paste(thumb,(xo,yo),thumb) final_rgb_pil = Image.new("RGB",self.video_frame_size,(0,0,0)); final_rgb_pil.paste(cv_rgba,mask=cv_rgba.split()[3]) dbg_path = os.path.join(self.output_dir,f"debug_PRE_NUMPY_S{scene_num}.png"); final_rgb_pil.save(dbg_path); logger.info(f"DEBUG: Saved PRE_NUMPY_S{scene_num} to {dbg_path}") frame_np = np.array(final_rgb_pil,dtype=np.uint8); if not frame_np.flags['C_CONTIGUOUS']: frame_np=np.ascontiguousarray(frame_np,dtype=np.uint8) logger.debug(f"S{scene_num}: NumPy for MoviePy. Shape:{frame_np.shape}, DType:{frame_np.dtype}, C-Contig:{frame_np.flags['C_CONTIGUOUS']}") if frame_np.size==0 or frame_np.ndim!=3 or frame_np.shape[2]!=3: logger.error(f"S{scene_num}: Invalid NumPy. Skip."); continue clip_base = ImageClip(frame_np,transparent=False).set_duration(scene_dur) mvpy_dbg_path=os.path.join(self.output_dir,f"debug_MOVIEPY_FRAME_S{scene_num}.png"); clip_base.save_frame(mvpy_dbg_path,t=0.1); logger.info(f"DEBUG: Saved MOVIEPY_FRAME_S{scene_num} to {mvpy_dbg_path}") clip_fx = clip_base try: es=random.uniform(1.03,1.08); clip_fx=clip_base.fx(vfx.resize,lambda t:1+(es-1)*(t/scene_dur) if scene_dur>0 else 1).set_position('center') except Exception as e: logger.error(f"S{scene_num} Ken Burns error: {e}",exc_info=False) current_scene_mvpy_clip = clip_fx elif asset_type == 'video': src_clip=None try: src_clip=VideoFileClip(asset_path,target_resolution=(self.video_frame_size[1],self.video_frame_size[0])if self.video_frame_size else None, audio=False) tmp_clip=src_clip if src_clip.duration!=scene_dur: if src_clip.duration>scene_dur:tmp_clip=src_clip.subclip(0,scene_dur) else: if scene_dur/src_clip.duration > 1.5 and src_clip.duration>0.1:tmp_clip=src_clip.loop(duration=scene_dur) else:tmp_clip=src_clip.set_duration(src_clip.duration);logger.info(f"S{scene_num} Video clip ({src_clip.duration:.2f}s) shorter than target ({scene_dur:.2f}s).") current_scene_mvpy_clip=tmp_clip.set_duration(scene_dur) # Ensure target duration for concatenation if current_scene_mvpy_clip.size!=list(self.video_frame_size):current_scene_mvpy_clip=current_scene_mvpy_clip.resize(self.video_frame_size) except Exception as e:logger.error(f"S{scene_num} Video load error '{asset_path}':{e}",exc_info=True);continue finally: if src_clip and src_clip is not current_scene_mvpy_clip and hasattr(src_clip,'close'):src_clip.close() else: logger.warning(f"S{scene_num} Unknown asset type '{asset_type}'. Skip."); continue if current_scene_mvpy_clip and key_action: try: to_dur=min(current_scene_mvpy_clip.duration-0.5,current_scene_mvpy_clip.duration*0.8)if current_scene_mvpy_clip.duration>0.5 else current_scene_mvpy_clip.duration to_start=0.25 if to_dur > 0: txt_c=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(to_dur).set_start(to_start).set_position(('center',0.92),relative=True) current_scene_mvpy_clip=CompositeVideoClip([current_scene_mvpy_clip,txt_c],size=self.video_frame_size,use_bgclip=True) else: logger.warning(f"S{scene_num}: Text overlay duration is zero. Skip text.") except Exception as e:logger.error(f"S{scene_num} TextClip error:{e}. No text.",exc_info=True) if current_scene_mvpy_clip:processed_clips.append(current_scene_mvpy_clip);logger.info(f"S{scene_num} Processed. Dur:{current_scene_mvpy_clip.duration:.2f}s.") except Exception as e:logger.error(f"MAJOR Error S{scene_num} ({asset_path}):{e}",exc_info=True) finally: if current_scene_mvpy_clip and hasattr(current_scene_mvpy_clip,'close'): try: current_scene_mvpy_clip.close() except: pass if not processed_clips:logger.warning("No clips processed. Abort.");return None td=0.75 try: logger.info(f"Concatenating {len(processed_clips)} clips."); if len(processed_clips)>1:final_clip=concatenate_videoclips(processed_clips,padding=-td if td>0 else 0,method="compose") elif processed_clips:final_clip=processed_clips[0] if not final_clip:logger.error("Concatenation failed.");return None logger.info(f"Concatenated dur:{final_clip.duration:.2f}s") if td>0 and final_clip.duration>0: if final_clip.duration>td*2:final_clip=final_clip.fx(vfx.fadein,td).fx(vfx.fadeout,td) else:final_clip=final_clip.fx(vfx.fadein,min(td,final_clip.duration/2.0)) if overall_narration_path and os.path.exists(overall_narration_path) and final_clip.duration>0: try:narration_clip=AudioFileClip(overall_narration_path);final_clip=final_clip.set_audio(narration_clip);logger.info("Narration added.") except Exception as e:logger.error(f"Narration add error:{e}",exc_info=True) elif final_clip.duration<=0:logger.warning("Video no duration. No audio.") if final_clip and final_clip.duration>0: op=os.path.join(self.output_dir,output_filename);logger.info(f"Writing video:{op} (Dur:{final_clip.duration:.2f}s)") final_clip.write_videofile(op,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",ffmpeg_params=["-pix_fmt", "yuv420p"]) logger.info(f"Video created:{op}");return op else:logger.error("Final clip invalid. No write.");return None except Exception as e:logger.error(f"Video write error:{e}",exc_info=True);return None finally: logger.debug("Closing all MoviePy clips in `assemble_animatic_from_assets` finally block.") all_clips_to_close = processed_clips + ([narration_clip] if narration_clip else []) + ([final_clip] if final_clip else []) # Corrected variable name for clip_obj_to_close in all_clips_to_close: if clip_obj_to_close and hasattr(clip_obj_to_close, 'close'): try: clip_obj_to_close.close() except Exception as e_close: logger.warning(f"Ignoring error while closing a clip: {type(clip_obj_to_close).__name__} - {e_close}")