# core/visual_engine.py from PIL import Image, ImageDraw, ImageFont, ImageOps # --- 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 logger = logging.getLogger(__name__) logger.setLevel(logging.INFO) # --- SERVICE CLIENT IMPORTS (Keep as before) --- 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; RunwayMLClient = None # Placeholder try: # from runwayml import RunwayClient # Hypothetical actual import # RunwayMLClient = RunwayClient # RUNWAYML_SDK_IMPORTED = True logger.info("RunwayML SDK import is a placeholder.") except ImportError: logger.warning("RunwayML SDK (placeholder) not found. RunwayML disabled.") except Exception as e_runway_sdk: logger.warning(f"Error importing RunwayML SDK (placeholder): {e_runway_sdk}. RunwayML 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 # Placeholder client logger.info("VisualEngine initialized.") # --- API Key Setters (Keep as before) --- 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).") 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): # For RunwayML self.runway_api_key = k if k: # For Gen-4, we might not need an SDK client if using direct HTTP, or an SDK client might be initialized here # if RUNWAYML_SDK_IMPORTED and RunwayMLClient: # try: # # self.runway_client = RunwayMLClient(api_key=k) # Actual SDK client init # self.USE_RUNWAYML = True; logger.info("RunwayML Client (Placeholder SDK) Ready.") # except Exception as e: logger.error(f"RunwayML client init error: {e}", exc_info=True); self.USE_RUNWAYML = False # else: # No SDK, or direct HTTP calls are planned self.USE_RUNWAYML = True; logger.info("RunwayML API Key set. (SDK integration is placeholder).") else: self.USE_RUNWAYML = False; logger.info("RunwayML Disabled (no API key).") # --- Helper Methods _get_text_dimensions, _create_placeholder_image_content, _search_pexels_image (Keep as before) --- def _get_text_dimensions(self,tc,fo): di=fo.size if hasattr(fo,'size') else self.font_size_pil; return (0,di) if not tc else (lambda b:(b[2]-b[0],b[3]-b[1] if b[3]-b[1]>0 else di))(fo.getbbox(tc)) if hasattr(fo,'getbbox') else (lambda s:(s[0],s[1] if s[1]>0 else di))(fo.getsize(tc)) if hasattr(fo,'getsize') else (int(len(tc)*di*0.6),int(di*1.2)) def _create_placeholder_image_content(self,td,fn,sz=None): if sz is None: sz = self.video_frame_size; img=Image.new('RGB',sz,color=(20,20,40));d=ImageDraw.Draw(img);pd=25;mw=sz[0]-(2*pd);ls=[]; if not td: td="(Placeholder Image)" ws=td.split();cl="" for w in ws: tl=cl+w+" ";raw_w,_=self._get_text_dimensions(tl,self.font);w=raw_w if raw_w > 0 else len(tl)*(self.font_size_pil*0.6); if w<=mw:cl=tl;else: if cl:ls.append(cl.strip());cl=w+" " if cl.strip():ls.append(cl.strip()) if not ls and td:ls.append(td[:int(mw//(self._get_text_dimensions("A",self.font)[0]or 10))]+"..." if td else "(Text too long)");elif not ls:ls.append("(Placeholder Error)") _,slh=self._get_text_dimensions("Ay",self.font);slh=slh if slh>0 else self.font_size_pil+2;mld=min(len(ls),(sz[1]-(2*pd))//(slh+2)) if slh>0 else 1; if mld<=0:mld=1;yts=pd+(sz[1]-(2*pd)-mld*(slh+2))/2.0;yt=yts for i in range(mld):lc=ls[i];lw,_=self._get_text_dimensions(lc,self.font);xt=(sz[0]-lw)/2.0;d.text((xt,yt),lc,font=self.font,fill=(200,200,180));yt+=slh+2 if i==6 and mld>7:d.text((xt,yt),"...",font=self.font,fill=(200,200,180));break fp=os.path.join(self.output_dir,fn); try:img.save(fp);return fp except Exception as e:logger.error(f"Save placeholder img {fp}: {e}",exc_info=True);return None def _search_pexels_image(self, q, ofnb): 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=Image.open(io.BytesIO(ir.content)) if id.mode!='RGB':id=id.convert('RGB');id.save(fp);logger.info(f"Pexels saved: {fp}");return fp else: logger.info(f"No Pexels for: '{eq}'") except Exception as e:logger.error(f"Pexels error ('{q}'): {e}",exc_info=True);return None # --- RunwayML Video Generation (Gen-4 Aligned Placeholder) --- def _generate_video_clip_with_runwayml(self, text_prompt_for_motion, input_image_path, scene_identifier_filename_base, target_duration_seconds=5): """ Placeholder for Runway Gen-4. Requires an input image and a text prompt for motion. target_duration_seconds should ideally be 5 or 10 for Gen-4. """ if not self.USE_RUNWAYML or not self.runway_api_key: logger.warning("RunwayML not enabled/API key missing. Cannot generate video clip.") return None if not input_image_path or not os.path.exists(input_image_path): logger.error(f"Runway Gen-4 requires an input image. Path not provided or invalid: {input_image_path}") return None # Gen-4 produces 5s or 10s. We can aim for the closest or let user choose via app.py if more control is needed. # For simplicity, let's assume target_duration_seconds from Gemini/user is a suggestion. # Actual API call would specify duration if supported, or model has fixed outputs. runway_duration_param = 10 if target_duration_seconds > 7 else 5 # Example logic to map to 5s or 10s output_video_filename = scene_identifier_filename_base.replace(".png", f"_runway_gen4_d{runway_duration_param}s.mp4") output_video_filepath = os.path.join(self.output_dir, output_video_filename) logger.info(f"Attempting Runway Gen-4 (Placeholder) with image: {os.path.basename(input_image_path)}, motion prompt: '{text_prompt_for_motion[:100]}...', target duration: {runway_duration_param}s") # --- ACTUAL RUNWAY GEN-4 API/SDK CALL WOULD GO HERE --- # This would involve: # 1. Uploading input_image_path (if API requires it, or providing a URL). # 2. Submitting the job with text_prompt_for_motion and desired parameters (duration, seed, etc.). # 3. Polling for completion. # 4. Downloading the resulting video to output_video_filepath. # Example (very hypothetical SDK structure): # try: # if not self.runway_client: self.runway_client = RunwayMLClient(api_key=self.runway_api_key) # runway_task = self.runway_client.gen4.generate( # image_path=input_image_path, # text_prompt=text_prompt_for_motion, # duration_seconds=runway_duration_param, # Or let model default # # ... other Gen-4 parameters like seed, motion_score, upscale, etc. # ) # runway_task.wait_for_completion() # Blocks until done # if runway_task.status == 'succeeded': # runway_task.download_video(output_video_filepath) # logger.info(f"Runway Gen-4 video saved to: {output_video_filepath}") # return output_video_filepath # else: # logger.error(f"Runway Gen-4 task failed. Status: {runway_task.status}, Error: {runway_task.error_message}") # return None # except Exception as e_runway: # logger.error(f"Error during actual Runway Gen-4 call: {e_runway}", exc_info=True) # return None # --- END ACTUAL RUNWAY GEN-4 API/SDK CALL --- logger.warning("Using PLACEHOLDER video generation for Runway Gen-4.") # Create a dummy video using the input image as a static frame for the placeholder try: img_clip = ImageClip(input_image_path).set_duration(runway_duration_param) # Add a text overlay to indicate it's a placeholder txt = f"Runway Gen-4 Placeholder\nInput: {os.path.basename(input_image_path)}\nMotion: {text_prompt_for_motion[:50]}..." txt_clip = TextClip(txt, fontsize=24, color='white', font=self.video_overlay_font, bg_color='rgba(0,0,0,0.5)', size=(self.video_frame_size[0]*0.8, None), method='caption').set_duration(runway_duration_param).set_position('center') final_placeholder_clip = CompositeVideoClip([img_clip, txt_clip], size=img_clip.size) final_placeholder_clip.write_videofile(output_video_filepath, fps=fps, codec='libx264', preset='ultrafast', logger=None, threads=2) logger.info(f"Runway Gen-4 placeholder video saved: {output_video_filepath}") if hasattr(img_clip, 'close'): img_clip.close() if hasattr(txt_clip, 'close'): txt_clip.close() if hasattr(final_placeholder_clip, 'close'): final_placeholder_clip.close() return output_video_filepath except Exception as e_placeholder: logger.error(f"Failed to create Runway Gen-4 placeholder video: {e_placeholder}", exc_info=True) return None def _create_placeholder_video_content(self, text_description, filename, duration=4, size=None): # Generic placeholder if input_image not available # ... (Keep as before, used if Runway is selected but input image gen fails) ... if size is None: size = self.video_frame_size; fp = os.path.join(self.output_dir, filename); tc = None try: tc = TextClip(text_description, fontsize=50, color='white', font=self.video_overlay_font, bg_color='black', size=size, method='caption').set_duration(duration) 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() # --- generate_scene_asset (Updated for Gen-4 Workflow) --- def generate_scene_asset(self, image_generation_prompt_text, # For DALL-E / Pexels motion_prompt_text_for_video, # For Runway Gen-4 (motion only) scene_data, scene_identifier_filename_base, generate_as_video_clip=False, runway_target_duration=5): 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': 'Generation not attempted'} # STEP 1: Generate the input image (DALL-E/Pexels/Placeholder) regardless of final asset type if video is chosen. # This image will serve as the base for Runway Gen-4 if generate_as_video_clip is True. input_image_for_runway_path = None image_filename_with_ext = base_name + "_base_image.png" # Differentiate base image filename image_filepath = os.path.join(self.output_dir, image_filename_with_ext) temp_image_asset_info = {'error': True, 'prompt_used': image_generation_prompt_text, 'error_message': 'Base image generation not attempted'} if self.USE_AI_IMAGE_GENERATION and self.openai_api_key: max_r, att_n = 2, 0 for att_n in range(max_r): try: logger.info(f"Attempt {att_n+1} DALL-E (for base image): {image_generation_prompt_text[:100]}...") 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[:100]}...") ir = requests.get(iu, timeout=120); ir.raise_for_status() id_img = Image.open(io.BytesIO(ir.content)) # Renamed to avoid conflict if id_img.mode != 'RGB': id_img = id_img.convert('RGB') id_img.save(image_filepath); logger.info(f"DALL-E base image saved: {image_filepath}"); input_image_for_runway_path = image_filepath temp_image_asset_info = {'path': image_filepath, 'type': 'image', 'error': False, 'prompt_used': image_generation_prompt_text, 'revised_prompt': rp} break # Success except openai.RateLimitError as e: logger.warning(f"OpenAI Rate Limit {att_n+1}: {e}. Retry..."); time.sleep(5*(att_n+1)); temp_image_asset_info['error_message']=str(e) except Exception as e: logger.error(f"DALL-E error: {e}", exc_info=True); temp_image_asset_info['error_message']=str(e); break if temp_image_asset_info['error']: logger.warning(f"DALL-E failed after {att_n+1} attempts for base image.") if temp_image_asset_info['error'] and self.USE_PEXELS : # Try Pexels if DALL-E failed 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, image_filename_with_ext) # Pass base image filename if pp: input_image_for_runway_path = pp; temp_image_asset_info = {'path': pp, 'type': 'image', 'error': False, 'prompt_used': f"Pexels: {pqt}"} else: current_em = temp_image_asset_info.get('error_message',""); temp_image_asset_info['error_message']=(current_em + " Pexels failed.").strip() if temp_image_asset_info['error']: # Fallback to placeholder for base image logger.warning("Base image generation (DALL-E/Pexels) failed. Using placeholder for base image.") ppt = temp_image_asset_info.get('prompt_used', image_generation_prompt_text) php = self._create_placeholder_image_content(f"[Base Img Placeholder] {ppt[:100]}...", image_filename_with_ext) if php: input_image_for_runway_path = php; temp_image_asset_info = {'path': php, 'type': 'image', 'error': False, 'prompt_used': ppt} else: current_em=temp_image_asset_info.get('error_message',"");temp_image_asset_info['error_message']=(current_em + " Base placeholder failed.").strip() # STEP 2: If video clip is requested and base image was successfully created, generate video with RunwayML if generate_as_video_clip: if self.USE_RUNWAYML and input_image_for_runway_path: logger.info(f"Proceeding to Runway Gen-4 video clip generation for {base_name} using base image: {input_image_for_runway_path}") video_path = self._generate_video_clip_with_runwayml( text_prompt_for_motion=motion_prompt_text_for_video, # Use the motion-specific prompt input_image_path=input_image_for_runway_path, scene_identifier_filename_base=base_name, # Will append _runway_gen4.mp4 target_duration_seconds=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} return asset_info # Successfully generated video else: logger.warning(f"RunwayML video clip generation failed for {base_name}. Using the base image as fallback.") asset_info = temp_image_asset_info # Fallback to the base image asset_info['error'] = True # Indicate video step failed, though base image might be okay asset_info['error_message'] = "RunwayML video generation step failed; using base image." asset_info['type'] = 'image' # Explicitly set to image as it's the fallback return asset_info elif not self.USE_RUNWAYML: logger.warning("RunwayML selected but not enabled/configured. Using base image.") asset_info = temp_image_asset_info asset_info['error_message'] = "RunwayML disabled; using base image." asset_info['type'] = 'image' return asset_info else: # No input_image_for_runway_path logger.error("Cannot generate RunwayML video: base image generation failed entirely.") asset_info = temp_image_asset_info # This will have error=True asset_info['error_message'] = (asset_info.get('error_message',"") + " Base image failed, so Runway video not attempted.").strip() asset_info['type'] = 'image' # Even though it failed, its type was image return asset_info else: # Image was requested directly asset_info = temp_image_asset_info # Return the result of the base image generation return asset_info # --- generate_narration_audio (Keep as before) --- def generate_narration_audio(self, text_to_narrate, output_filename="narration_overall.mp3"): 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("11L .text_to_speech.stream()") elif hasattr(self.elevenlabs_client,'generate_stream'):asm=self.elevenlabs_client.generate_stream;logger.info("11L .generate_stream()") elif hasattr(self.elevenlabs_client,'generate'):logger.info("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 # --- assemble_animatic_from_assets (Keep robust version from previous step, ensuring C-contiguous array and debug saves) --- 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 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]) # CRITICAL DEBUG: Save image fed to NumPy array 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) # CRITICAL DEBUG: Save frame from MoviePy clip 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) 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) 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 # (current_scene_mvpy_clip.duration-to_dur)/2.0 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) 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() # This might close the clip if it's a VideoFileClip directly except: pass # Avoid error during cleanup 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"]) # Added pix_fmt 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.") # Close clips individually to catch errors without stopping others for clip_obj in processed_clips: if clip_obj and hasattr(clip_obj, 'close'): try: clip_obj.close() except Exception as e_close: logger.warning(f"Ignoring error closing a processed clip: {e_close}") if narration_clip and hasattr(narration_clip, 'close'): try: narration_clip.close() except Exception as e_close_audio: logger.warning(f"Ignoring error closing narration clip: {e_close_audio}") if final_clip and hasattr(final_clip, 'close'): # final_composite_clip_obj was renamed to final_clip try: final_clip.close() except Exception as e_close_final: logger.warning(f"Ignoring error closing final composite clip: {e_close_final}")