# core/visual_engine.py from PIL import Image, ImageDraw, ImageFont, ImageOps import base64 import mimetypes import numpy as np import os import openai # Ensure this is OpenAI v1.x.x+ import requests import io import time import random import logging # --- MoviePy Imports --- from moviepy.editor import (ImageClip, VideoFileClip, concatenate_videoclips, TextClip, CompositeVideoClip, AudioFileClip) import moviepy.video.fx.all as vfx # --- MONKEY PATCH for Pillow/MoviePy compatibility --- 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. MoviePy effects might fail or look different.") except Exception as e_monkey_patch: print(f"WARNING: An unexpected error occurred during Pillow ANTIALIAS monkey-patch: {e_monkey_patch}") logger = logging.getLogger(__name__) # Set a default level; can be overridden by the main app's logging config # logger.setLevel(logging.DEBUG) # Uncomment for very verbose output during development # --- External Service Client Imports --- ELEVENLABS_CLIENT_IMPORTED = False ElevenLabsAPIClient = None # Class placeholder Voice = None # Class placeholder VoiceSettings = None # Class placeholder 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 (SDK v1.x.x pattern) imported successfully.") except ImportError: logger.warning("ElevenLabs SDK not found (expected 'pip install elevenlabs>=1.0.0'). Audio generation will be disabled.") except Exception as e_eleven_import_general: logger.warning(f"General error importing ElevenLabs client components: {e_eleven_import_general}. Audio generation disabled.") RUNWAYML_SDK_IMPORTED = False RunwayMLAPIClientClass = None # Storing the class itself try: from runwayml import RunwayML as ImportedRunwayMLAPIClientClass # Actual SDK import RunwayMLAPIClientClass = ImportedRunwayMLAPIClientClass RUNWAYML_SDK_IMPORTED = True logger.info("RunwayML SDK (runwayml) 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_import_general: logger.warning(f"General error importing RunwayML SDK: {e_runway_sdk_import_general}. RunwayML features disabled.") class VisualEngine: DEFAULT_FONT_SIZE_PIL = 10 PREFERRED_FONT_SIZE_PIL = 20 VIDEO_OVERLAY_FONT_SIZE = 30 VIDEO_OVERLAY_FONT_COLOR = 'white' DEFAULT_MOVIEPY_FONT = 'DejaVu-Sans-Bold' # Common ImageMagick font name PREFERRED_MOVIEPY_FONT = 'Liberation-Sans-Bold' 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_pil_preference = "DejaVuSans-Bold.ttf" # More standard Linux font font_paths_to_try = [ self.font_filename_pil_preference, f"/usr/share/fonts/truetype/dejavu/{self.font_filename_pil_preference}", f"/usr/share/fonts/truetype/liberation/LiberationSans-Bold.ttf", # Alternative f"/System/Library/Fonts/Supplemental/Arial.ttf", f"C:/Windows/Fonts/arial.ttf", f"/usr/local/share/fonts/truetype/mycustomfonts/arial.ttf" # Previous custom ] self.resolved_font_path_pil = next((p for p in font_paths_to_try if os.path.exists(p)), None) self.active_font_pil = ImageFont.load_default() # Fallback default self.active_font_size_pil = self.DEFAULT_FONT_SIZE_PIL self.active_moviepy_font_name = self.DEFAULT_MOVIEPY_FONT if self.resolved_font_path_pil: try: self.active_font_pil = ImageFont.truetype(self.resolved_font_path_pil, self.PREFERRED_FONT_SIZE_PIL) self.active_font_size_pil = self.PREFERRED_FONT_SIZE_PIL logger.info(f"Pillow font loaded: {self.resolved_font_path_pil} at size {self.active_font_size_pil}.") # Determine MoviePy font based on loaded PIL font for consistency if "dejavu" in self.resolved_font_path_pil.lower(): self.active_moviepy_font_name = 'DejaVu-Sans-Bold' elif "liberation" in self.resolved_font_path_pil.lower(): self.active_moviepy_font_name = 'Liberation-Sans-Bold' except IOError as e_font_load_io: logger.error(f"Pillow font loading IOError for '{self.resolved_font_path_pil}': {e_font_load_io}. Using default font.") else: logger.warning("Preferred Pillow font not found in predefined paths. Using default font.") # Service API keys and flags self.openai_api_key = None; self.USE_AI_IMAGE_GENERATION = False self.dalle_model = "dall-e-3"; self.image_size_dalle3 = "1792x1024" # DALL-E 3 landscape self.video_frame_size = (1280, 720) # HD 16:9 self.elevenlabs_api_key = None; self.USE_ELEVENLABS = False; self.elevenlabs_client_instance = None self.elevenlabs_voice_id = default_elevenlabs_voice_id if VoiceSettings and ELEVENLABS_CLIENT_IMPORTED: self.elevenlabs_voice_settings_obj = VoiceSettings(stability=0.60, similarity_boost=0.80, style=0.15, use_speaker_boost=True) else: self.elevenlabs_voice_settings_obj = None self.pexels_api_key = None; self.USE_PEXELS = False self.runway_api_key = None; self.USE_RUNWAYML = False; self.runway_ml_sdk_client_instance = None # Instance of RunwayML() # Attempt to initialize RunwayML client if SDK is present and RUNWAYML_API_SECRET env var might be set if RUNWAYML_SDK_IMPORTED and RunwayMLAPIClientClass and os.getenv("RUNWAYML_API_SECRET"): try: self.runway_ml_sdk_client_instance = RunwayMLAPIClientClass() # SDK uses env var by default self.USE_RUNWAYML = True # Assume enabled if client initializes without error logger.info("RunwayML Client initialized using RUNWAYML_API_SECRET environment variable at startup.") except Exception as e_runway_init_at_startup: logger.error(f"Initial RunwayML client initialization failed (env var RUNWAYML_API_SECRET might be invalid or SDK issue): {e_runway_init_at_startup}") self.USE_RUNWAYML = False # Ensure it's disabled if init fails logger.info("VisualEngine initialized.") # --- API Key Setter Methods --- def set_openai_api_key(self, api_key_value): self.openai_api_key = api_key_value; self.USE_AI_IMAGE_GENERATION = bool(api_key_value) logger.info(f"DALL-E ({self.dalle_model}) service status: {'Ready' if self.USE_AI_IMAGE_GENERATION else 'Disabled (no API key)'}") def set_elevenlabs_api_key(self, api_key_value, voice_id_from_secret=None): self.elevenlabs_api_key = api_key_value if voice_id_from_secret: self.elevenlabs_voice_id = voice_id_from_secret if api_key_value and ELEVENLABS_CLIENT_IMPORTED and ElevenLabsAPIClient: try: self.elevenlabs_client_instance = ElevenLabsAPIClient(api_key=api_key_value) # Pass key directly self.USE_ELEVENLABS = bool(self.elevenlabs_client_instance) logger.info(f"ElevenLabs Client service status: {'Ready' if self.USE_ELEVENLABS else 'Failed Initialization'} (Using Voice ID: {self.elevenlabs_voice_id})") except Exception as e_11l_init: logger.error(f"ElevenLabs client initialization error: {e_11l_init}. Service Disabled.", exc_info=True) self.USE_ELEVENLABS = False; self.elevenlabs_client_instance = None else: self.USE_ELEVENLABS = False logger.info(f"ElevenLabs Service Disabled (API key not provided or SDK import issue).") def set_pexels_api_key(self, api_key_value): self.pexels_api_key = api_key_value; self.USE_PEXELS = bool(api_key_value) logger.info(f"Pexels Search service status: {'Ready' if self.USE_PEXELS else 'Disabled (no API key)'}") def set_runway_api_key(self, api_key_value): self.runway_api_key = api_key_value # Store the key itself if api_key_value: if RUNWAYML_SDK_IMPORTED and RunwayMLAPIClientClass: if not self.runway_ml_sdk_client_instance: # If not already initialized (e.g., by env var at startup) try: # The RunwayML Python SDK expects the API key via the RUNWAYML_API_SECRET env var. original_env_secret = os.getenv("RUNWAYML_API_SECRET") if not original_env_secret: # If env var not set, set it temporarily from passed key logger.info("Temporarily setting RUNWAYML_API_SECRET from provided key for SDK client initialization.") os.environ["RUNWAYML_API_SECRET"] = api_key_value self.runway_ml_sdk_client_instance = RunwayMLAPIClientClass() self.USE_RUNWAYML = True # Mark service as usable if client initializes logger.info("RunwayML Client initialized successfully using provided API key (via env var).") if not original_env_secret: # Clean up: remove env var if we set it del os.environ["RUNWAYML_API_SECRET"] logger.info("Cleared temporary RUNWAYML_API_SECRET environment variable.") except Exception as e_runway_client_setkey_init: logger.error(f"RunwayML Client initialization via set_runway_api_key failed: {e_runway_client_setkey_init}", exc_info=True) self.USE_RUNWAYML = False; self.runway_ml_sdk_client_instance = None # Ensure it's disabled else: # Client was already initialized self.USE_RUNWAYML = True # Service is usable logger.info("RunwayML Client was already initialized (likely from environment variable). API key stored.") else: # SDK not imported logger.warning("RunwayML SDK not imported. API key has been stored, but the current integration relies on the SDK. Service effectively disabled.") self.USE_RUNWAYML = False # Can't use without SDK if that's the implemented path else: # No API key provided self.USE_RUNWAYML = False; self.runway_ml_sdk_client_instance = None logger.info("RunwayML Service Disabled (no API key provided to set_runway_api_key).") # --- Helper Methods (_image_to_data_uri, _map_resolution_to_runway_ratio, etc.) --- 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() mime_map = {".png": "image/png", ".jpg": "image/jpeg", ".jpeg": "image/jpeg", ".webp": "image/webp"} mime_type = mime_map.get(ext, "application/octet-stream") if mime_type == "application/octet-stream": logger.warning(f"Could not determine MIME type for {image_path}, using default.") 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"Generated data URI for {os.path.basename(image_path)} (first 100 chars): {data_uri[:100]}...") return data_uri except FileNotFoundError: logger.error(f"Image file not found at {image_path} for data URI conversion."); return None except Exception as e: logger.error(f"Error converting image {image_path} to data URI: {e}", exc_info=True); return None def _map_resolution_to_runway_ratio(self, width, height): ratio_str = f"{width}:{height}" supported_ratios_gen4 = ["1280:720", "720:1280", "1104:832", "832:1104", "960:960", "1584:672"] if ratio_str in supported_ratios_gen4: return ratio_str logger.warning(f"Resolution {ratio_str} not directly in Gen-4 supported list. Defaulting to 1280:720 for RunwayML.") return "1280:720" def _get_text_dimensions(self, text_content, font_object_pil): default_h = getattr(font_object_pil, 'size', self.active_font_size_pil) if not text_content: return 0, default_h try: if hasattr(font_object_pil,'getbbox'): bbox=font_object_pil.getbbox(text_content);w=bbox[2]-bbox[0];h=bbox[3]-bbox[1]; return w, h if h > 0 else default_h elif hasattr(font_object_pil,'getsize'): w,h=font_object_pil.getsize(text_content); return w, h if h > 0 else default_h else: return int(len(text_content)*default_h*0.6),int(default_h*1.2) # Basic estimate except Exception as e_getdim: logger.warning(f"Error in _get_text_dimensions: {e_getdim}"); return int(len(text_content)*self.active_font_size_pil*0.6),int(self.active_font_size_pil*1.2) def _create_placeholder_image_content(self, text_description, filename, size=None): # <<< THIS IS THE CORRECTED VERSION OF THIS 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_for_placeholder = [] # Renamed to avoid conflict if not text_description: text_description = "(Placeholder Image)" words_list = text_description.split(); current_line_buffer = "" # Renamed for word_idx, word_item in enumerate(words_list): # Renamed prospective_add = word_item + (" " if word_idx < len(words_list) - 1 else "") test_line_candidate = current_line_buffer + prospective_add current_w_text, _ = self._get_text_dimensions(test_line_candidate, self.active_font_pil) if current_w_text == 0 and test_line_candidate.strip(): current_w_text = len(test_line_candidate) * (self.active_font_size_pil * 0.6) if current_w_text <= max_w: current_line_buffer = test_line_candidate else: if current_line_buffer.strip(): lines_for_placeholder.append(current_line_buffer.strip()) current_line_buffer = prospective_add if current_line_buffer.strip(): lines_for_placeholder.append(current_line_buffer.strip()) if not lines_for_placeholder and text_description: avg_char_w_est, _ = self._get_text_dimensions("W", self.active_font_pil); avg_char_w_est = avg_char_w_est or (self.active_font_size_pil * 0.6) chars_per_line_est = int(max_w / avg_char_w_est) if avg_char_w_est > 0 else 20 lines_for_placeholder.append(text_description[:chars_per_line_est] + ("..." if len(text_description) > chars_per_line_est else "")) elif not lines_for_placeholder: lines_for_placeholder.append("(Placeholder Error)") _, single_h = self._get_text_dimensions("Ay", self.active_font_pil); single_h = single_h if single_h > 0 else self.active_font_size_pil + 2 max_l = min(len(lines_for_placeholder), (size[1] - (2 * padding)) // (single_h + 2)) if single_h > 0 else 1; max_l = max(1, max_l) y_p = padding + (size[1] - (2 * padding) - max_l * (single_h + 2)) / 2.0 for i_line in range(max_l): # Renamed line_txt_content = lines_for_placeholder[i_line]; line_w_val, _ = self._get_text_dimensions(line_txt_content, self.active_font_pil) if line_w_val == 0 and line_txt_content.strip(): line_w_val = len(line_txt_content) * (self.active_font_size_pil * 0.6) x_p = (size[0] - line_w_val) / 2.0 try: d.text((x_p, y_p), line_txt_content, font=self.active_font_pil, fill=(200, 200, 180)) except Exception as e_draw_txt: logger.error(f"Pillow d.text error: {e_draw_txt} for '{line_txt_content}'") y_p += single_h + 2 if i_line == 6 and max_l > 7: try: d.text((x_p, y_p), "...", font=self.active_font_pil, fill=(200, 200, 180)) except Exception as e_elps: logger.error(f"Pillow d.text ellipsis error: {e_elps}"); break filepath_placeholder = os.path.join(self.output_dir, filename) # Renamed try: img.save(filepath_placeholder); return filepath_placeholder except Exception as e_save_ph: logger.error(f"Saving placeholder image '{filepath_placeholder}' error: {e_save_ph}", exc_info=True); return None def _search_pexels_image(self, query_str, output_fn_base): # (Corrected from previous response) if not self.USE_PEXELS or not self.pexels_api_key: return None http_headers = {"Authorization": self.pexels_api_key} http_params = {"query": query_str, "per_page": 1, "orientation": "landscape", "size": "large2x"} base_name_px, _ = os.path.splitext(output_fn_base) pexels_fn_str = base_name_px + f"_pexels_{random.randint(1000,9999)}.jpg" file_path_px = os.path.join(self.output_dir, pexels_fn_str) try: logger.info(f"Pexels: Searching for '{query_str}'") eff_query_px = " ".join(query_str.split()[:5]) http_params["query"] = eff_query_px response_px = requests.get("https://api.pexels.com/v1/search", headers=http_headers, params=http_params, timeout=20) response_px.raise_for_status() data_px = response_px.json() if data_px.get("photos") and len(data_px["photos"]) > 0: photo_details_px = data_px["photos"][0] photo_url_px = photo_details_px.get("src", {}).get("large2x") if not photo_url_px: logger.warning(f"Pexels: 'large2x' URL missing for '{eff_query_px}'. Details: {photo_details_px}"); return None image_response_px = requests.get(photo_url_px, timeout=60); image_response_px.raise_for_status() img_pil_data_px = Image.open(io.BytesIO(image_response_px.content)) if img_pil_data_px.mode != 'RGB': img_pil_data_px = img_pil_data_px.convert('RGB') img_pil_data_px.save(file_path_px); logger.info(f"Pexels: Image saved to {file_path_px}"); return file_path_px else: logger.info(f"Pexels: No photos for '{eff_query_px}'."); return None except requests.exceptions.RequestException as e_req_px: logger.error(f"Pexels: RequestException for '{query_str}': {e_req_px}", exc_info=False); return None except Exception as e_px_gen: logger.error(f"Pexels: General error for '{query_str}': {e_px_gen}", exc_info=True); return None 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 from before) if not self.USE_RUNWAYML or not self.runway_ml_sdk_client_instance: 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_str = self._image_to_data_uri(input_image_path) # Renamed if not image_data_uri_str: return None runway_dur = 10 if target_duration_seconds >= 8 else 5 runway_ratio = self._map_resolution_to_runway_ratio(self.video_frame_size[0], self.video_frame_size[1]) base_name_for_runway_vid, _ = os.path.splitext(scene_identifier_filename_base); output_vid_fn = base_name_for_runway_vid + f"_runway_gen4_d{runway_dur}s.mp4" # Renamed output_vid_fp = os.path.join(self.output_dir, output_vid_fn) # Renamed logger.info(f"Runway Gen-4 task: motion='{text_prompt_for_motion[:100]}...', img='{os.path.basename(input_image_path)}', dur={runway_dur}s, ratio='{runway_ratio}'") try: task_submitted_runway = self.runway_ml_sdk_client_instance.image_to_video.create(model='gen4_turbo', prompt_image=image_data_uri_str, prompt_text=text_prompt_for_motion, duration=runway_dur, ratio=runway_ratio) # Renamed task_id_runway = task_submitted_runway.id; logger.info(f"Runway Gen-4 task ID: {task_id_runway}. Polling...") # Renamed poll_sec=10; max_poll_count=36; poll_start_time = time.time() # Renamed while time.time() - poll_start_time < max_poll_count * poll_sec: time.sleep(poll_sec); task_details_runway = self.runway_ml_sdk_client_instance.tasks.retrieve(id=task_id_runway) # Renamed logger.info(f"Runway task {task_id_runway} status: {task_details_runway.status}") if task_details_runway.status == 'SUCCEEDED': output_url_runway = getattr(getattr(task_details_runway,'output',None),'url',None) or (getattr(task_details_runway,'artifacts',None) and task_details_runway.artifacts[0].url if task_details_runway.artifacts and hasattr(task_details_runway.artifacts[0],'url') else None) or (getattr(task_details_runway,'artifacts',None) and task_details_runway.artifacts[0].download_url if task_details_runway.artifacts and hasattr(task_details_runway.artifacts[0],'download_url') else None) # Renamed if not output_url_runway: logger.error(f"Runway task {task_id_runway} SUCCEEDED, but no output URL. Details: {vars(task_details_runway) if hasattr(task_details_runway,'__dict__') else task_details_runway}"); return None logger.info(f"Runway task {task_id_runway} SUCCEEDED. Downloading: {output_url_runway}") video_resp_get = requests.get(output_url_runway, stream=True, timeout=300); video_resp_get.raise_for_status() # Renamed with open(output_vid_fp,'wb') as f_vid: # Renamed for chunk_data in video_resp_get.iter_content(chunk_size=8192): f_vid.write(chunk_data) # Renamed logger.info(f"Runway Gen-4 video saved: {output_vid_fp}"); return output_vid_fp elif task_details_runway.status in ['FAILED','ABORTED','ERROR']: err_msg_runway = getattr(task_details_runway,'error_message',None) or getattr(getattr(task_details_runway,'output',None),'error',"Unknown Runway error.") # Renamed logger.error(f"Runway task {task_id_runway} status: {task_details_runway.status}. Error: {err_msg_runway}"); return None logger.warning(f"Runway task {task_id_runway} timed out."); return None except AttributeError as ae_sdk: logger.error(f"RunwayML SDK AttrError: {ae_sdk}. SDK/methods changed?", exc_info=True); return None # Renamed except Exception as e_runway_gen: logger.error(f"Runway Gen-4 API error: {e_runway_gen}", exc_info=True); return None # Renamed def _create_placeholder_video_content(self, text_desc_ph, filename_ph, duration_ph=4, size_ph=None): # Renamed variables # <<< THIS IS THE CORRECTED METHOD >>> if size_ph is None: size_ph = self.video_frame_size filepath_ph = os.path.join(self.output_dir, filename_ph) text_clip_ph = None # Initialize try: text_clip_ph = TextClip(text_desc_ph, fontsize=50, color='white', font=self.video_overlay_font, bg_color='black', size=size_ph, method='caption').set_duration(duration_ph) text_clip_ph.write_videofile(filepath_ph, fps=24, codec='libx264', preset='ultrafast', logger=None, threads=2) logger.info(f"Generic placeholder video created: {filepath_ph}") return filepath_ph except Exception as e_ph_vid: # Specific exception variable logger.error(f"Failed to create generic placeholder video '{filepath_ph}': {e_ph_vid}", exc_info=True) return None finally: if text_clip_ph and hasattr(text_clip_ph, 'close'): text_clip_ph.close() # --- generate_scene_asset (Main asset generation logic) --- def generate_scene_asset(self, image_generation_prompt_text, motion_prompt_text_for_video, scene_data_dict, scene_identifier_fn_base, # Renamed generate_as_video_clip_flag=False, runway_target_dur_val=5): # Renamed # (Logic mostly as before, ensuring base image is robustly generated first) base_name_asset, _ = os.path.splitext(scene_identifier_fn_base) # Renamed asset_info_result = {'path': None, 'type': 'none', 'error': True, 'prompt_used': image_generation_prompt_text, 'error_message': 'Asset generation init failed'} # Renamed path_for_input_image_runway = None # Renamed fn_for_base_image = base_name_asset + ("_base_for_video.png" if generate_as_video_clip_flag else ".png") # Renamed fp_for_base_image = os.path.join(self.output_dir, fn_for_base_image) # Renamed if self.USE_AI_IMAGE_GENERATION and self.openai_api_key: # (DALL-E logic with corrected try/except from previous response) max_r_dalle, att_n_dalle = 2,0; for att_n_dalle in range(max_r_dalle): att_c_dalle = att_n_dalle + 1 # Renamed try: logger.info(f"Att {att_c_dalle} DALL-E (base img): {image_generation_prompt_text[:70]}..."); oai_cl = openai.OpenAI(api_key=self.openai_api_key,timeout=90.0) # Renamed oai_r = oai_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") # Renamed oai_iu = oai_r.data[0].url; oai_rp = getattr(oai_r.data[0],'revised_prompt',None) # Renamed if oai_rp: logger.info(f"DALL-E revised: {oai_rp[:70]}...") oai_ir = requests.get(oai_iu,timeout=120); oai_ir.raise_for_status() # Renamed oai_id = Image.open(io.BytesIO(oai_ir.content)); # Renamed if oai_id.mode!='RGB': oai_id=oai_id.convert('RGB') oai_id.save(fp_for_base_image); logger.info(f"DALL-E base img saved: {fp_for_base_image}") path_for_input_image_runway=fp_for_base_image asset_info_result={'path':fp_for_base_image,'type':'image','error':False,'prompt_used':image_generation_prompt_text,'revised_prompt':oai_rp} break except openai.RateLimitError as e_oai_rl: logger.warning(f"OpenAI RateLimit Att {att_c_dalle}:{e_oai_rl}.Retry...");time.sleep(5*att_c_dalle);asset_info_result['error_message']=str(e_oai_rl) # Renamed except openai.APIError as e_oai_api: logger.error(f"OpenAI APIError Att {att_c_dalle}:{e_oai_api}");asset_info_result['error_message']=str(e_oai_api);break # Renamed except requests.exceptions.RequestException as e_oai_req: logger.error(f"Requests Err DALL-E Att {att_c_dalle}:{e_oai_req}");asset_info_result['error_message']=str(e_oai_req);break # Renamed except Exception as e_oai_gen: logger.error(f"General DALL-E Err Att {att_c_dalle}:{e_oai_gen}",exc_info=True);asset_info_result['error_message']=str(e_oai_gen);break # Renamed if asset_info_result['error']: logger.warning(f"DALL-E failed after {att_c_dalle} attempts for base img.") if asset_info_result['error'] and self.USE_PEXELS: logger.info("Trying Pexels for base img.");px_qt=scene_data_dict.get('pexels_search_query_감독',f"{scene_data_dict.get('emotional_beat','')} {scene_data_dict.get('setting_description','')}");px_pp=self._search_pexels_image(px_qt,fn_for_base_image); # Renamed variables if px_pp:path_for_input_image_runway=px_pp;asset_info_result={'path':px_pp,'type':'image','error':False,'prompt_used':f"Pexels:{px_qt}"} else:current_em_px=asset_info_result.get('error_message',"");asset_info_result['error_message']=(current_em_px+" Pexels failed for base.").strip() # Renamed if asset_info_result['error']: logger.warning("Base img (DALL-E/Pexels) failed. Using placeholder.");ph_ppt=asset_info_result.get('prompt_used',image_generation_prompt_text);ph_p=self._create_placeholder_image_content(f"[Base Placeholder]{ph_ppt[:70]}...",fn_for_base_image); # Renamed variables if ph_p:path_for_input_image_runway=ph_p;asset_info_result={'path':ph_p,'type':'image','error':False,'prompt_used':ph_ppt} else:current_em_ph=asset_info_result.get('error_message',"");asset_info_result['error_message']=(current_em_ph+" Base placeholder failed.").strip() # Renamed if generate_as_video_clip_flag: if not path_for_input_image_runway:logger.error("RunwayML video: base img failed.");asset_info_result['error']=True;asset_info_result['error_message']=(asset_info_result.get('error_message',"")+" Base img miss, Runway abort.").strip();asset_info_result['type']='none';return asset_info_result if self.USE_RUNWAYML: runway_video_p=self._generate_video_clip_with_runwayml(motion_prompt_text_for_video,path_for_input_image_runway,base_name_asset,runway_target_dur_val) # Renamed if runway_video_p and os.path.exists(runway_video_p):asset_info_result={'path':runway_video_p,'type':'video','error':False,'prompt_used':motion_prompt_text_for_video,'base_image_path':path_for_input_image_runway} else:logger.warning(f"RunwayML video failed for {base_name_asset}. Fallback to base img.");asset_info_result['error']=True;asset_info_result['error_message']=(asset_info_result.get('error_message',"Base img ok.")+" RunwayML video fail; use base img.").strip();asset_info_result['path']=path_for_input_image_runway;asset_info_result['type']='image';asset_info_result['prompt_used']=image_generation_prompt_text else:logger.warning("RunwayML selected but disabled. Use base img.");asset_info_result['error']=True;asset_info_result['error_message']=(asset_info_result.get('error_message',"Base img ok.")+" RunwayML disabled; use base img.").strip();asset_info_result['path']=path_for_input_image_runway;asset_info_result['type']='image';asset_info_result['prompt_used']=image_generation_prompt_text return asset_info_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_instance or not text_to_narrate: logger.info("11L skip."); return None # Check instance audio_fp_11l=os.path.join(self.output_dir,output_filename) # Renamed try: logger.info(f"11L audio (Voice:{self.elevenlabs_voice_id}): {text_to_narrate[:70]}..."); audio_sm_11l=None # Renamed if hasattr(self.elevenlabs_client_instance,'text_to_speech')and hasattr(self.elevenlabs_client_instance.text_to_speech,'stream'):audio_sm_11l=self.elevenlabs_client_instance.text_to_speech.stream;logger.info("Using 11L .text_to_speech.stream()") elif hasattr(self.elevenlabs_client_instance,'generate_stream'):audio_sm_11l=self.elevenlabs_client_instance.generate_stream;logger.info("Using 11L .generate_stream()") elif hasattr(self.elevenlabs_client_instance,'generate'):logger.info("Using 11L .generate()");eleven_vp=Voice(voice_id=str(self.elevenlabs_voice_id),settings=self.elevenlabs_voice_settings_obj)if Voice and self.elevenlabs_voice_settings_obj else str(self.elevenlabs_voice_id);eleven_ab=self.elevenlabs_client_instance.generate(text=text_to_narrate,voice=eleven_vp,model="eleven_multilingual_v2"); # Renamed with open(audio_fp_11l,"wb")as f_11l:f_11l.write(eleven_ab);logger.info(f"11L audio (non-stream): {audio_fp_11l}");return audio_fp_11l # Renamed else:logger.error("No 11L audio method.");return None if audio_sm_11l:eleven_vps={"voice_id":str(self.elevenlabs_voice_id)} # Renamed if self.elevenlabs_voice_settings_obj: if hasattr(self.elevenlabs_voice_settings_obj,'model_dump'):eleven_vps["voice_settings"]=self.elevenlabs_voice_settings_obj.model_dump() elif hasattr(self.elevenlabs_voice_settings_obj,'dict'):eleven_vps["voice_settings"]=self.elevenlabs_voice_settings_obj.dict() else:eleven_vps["voice_settings"]=self.elevenlabs_voice_settings_obj eleven_adi=audio_sm_11l(text=text_to_narrate,model_id="eleven_multilingual_v2",**eleven_vps) # Renamed with open(audio_fp_11l,"wb")as f_11l_stream: # Renamed for chunk_11l in eleven_adi: # Renamed if chunk_11l:f_11l_stream.write(chunk_11l) logger.info(f"11L audio (stream): {audio_fp_11l}");return audio_fp_11l except Exception as e_11labs_audio:logger.error(f"11L audio error: {e_11labs_audio}",exc_info=True);return None # Renamed 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 with robust image processing, C-contiguous array, debug saves, and pix_fmt) # ... (This extensive method is assumed to be largely correct from the previous iteration focusing on blank video issues) # ... (The core of that robust version should be retained here) ... # For brevity, I'm not re-pasting the entire assemble_animatic_from_assets if it was correct before this syntax error hunt. # If it also needs review, let me know. The key is that the *input* (asset_data_list) # from the corrected generate_scene_asset is now more reliable. if not asset_data_list: logger.warning("No assets for animatic."); return None processed_moviepy_clips_list = []; narration_audio_clip_mvpy = None; final_video_output_clip = None # Renamed variables logger.info(f"Assembling from {len(asset_data_list)} assets. Target Frame: {self.video_frame_size}.") for i_asset, asset_info_item_loop in enumerate(asset_data_list): # Renamed loop variables path_of_asset, type_of_asset, duration_for_scene = asset_info_item_loop.get('path'), asset_info_item_loop.get('type'), asset_info_item_loop.get('duration', 4.5) num_of_scene, action_in_key = asset_info_item_loop.get('scene_num', i_asset + 1), asset_info_item_loop.get('key_action', '') logger.info(f"S{num_of_scene}: Path='{path_of_asset}', Type='{type_of_asset}', Dur='{duration_for_scene}'s") if not (path_of_asset and os.path.exists(path_of_asset)): logger.warning(f"S{num_of_scene}: Not found '{path_of_asset}'. Skip."); continue if duration_for_scene <= 0: logger.warning(f"S{num_of_scene}: Invalid duration ({duration_for_scene}s). Skip."); continue active_scene_clip = None # Clip for this iteration try: if type_of_asset == 'image': # (Robust image processing from previous version, including debug saves) opened_pil_img = Image.open(path_of_asset); logger.debug(f"S{num_of_scene}: Loaded img. Mode:{opened_pil_img.mode}, Size:{opened_pil_img.size}") converted_img_rgba = opened_pil_img.convert('RGBA') if opened_pil_img.mode != 'RGBA' else opened_pil_img.copy() thumbnailed_img = converted_img_rgba.copy(); resample_f = Image.Resampling.LANCZOS if hasattr(Image.Resampling,'LANCZOS') else Image.BILINEAR; thumbnailed_img.thumbnail(self.video_frame_size,resample_f) rgba_canvas = Image.new('RGBA',self.video_frame_size,(0,0,0,0)); pos_x,pos_y=(self.video_frame_size[0]-thumbnailed_img.width)//2,(self.video_frame_size[1]-thumbnailed_img.height)//2 rgba_canvas.paste(thumbnailed_img,(pos_x,pos_y),thumbnailed_img) final_rgb_img_pil = Image.new("RGB",self.video_frame_size,(0,0,0)); final_rgb_img_pil.paste(rgba_canvas,mask=rgba_canvas.split()[3]) debug_path_img_pre_numpy = os.path.join(self.output_dir,f"debug_PRE_NUMPY_S{num_of_scene}.png"); final_rgb_img_pil.save(debug_path_img_pre_numpy); logger.info(f"DEBUG: Saved PRE_NUMPY_S{num_of_scene} to {debug_path_img_pre_numpy}") numpy_frame_arr = np.array(final_rgb_img_pil,dtype=np.uint8); if not numpy_frame_arr.flags['C_CONTIGUOUS']: numpy_frame_arr=np.ascontiguousarray(numpy_frame_arr,dtype=np.uint8) logger.debug(f"S{num_of_scene}: NumPy for MoviePy. Shape:{numpy_frame_arr.shape}, DType:{numpy_frame_arr.dtype}, C-Contig:{numpy_frame_arr.flags['C_CONTIGUOUS']}") if numpy_frame_arr.size==0 or numpy_frame_arr.ndim!=3 or numpy_frame_arr.shape[2]!=3: logger.error(f"S{num_of_scene}: Invalid NumPy array for MoviePy. Skip."); continue base_image_clip = ImageClip(numpy_frame_arr,transparent=False).set_duration(duration_for_scene) debug_path_moviepy_frame=os.path.join(self.output_dir,f"debug_MOVIEPY_FRAME_S{num_of_scene}.png"); base_image_clip.save_frame(debug_path_moviepy_frame,t=0.1); logger.info(f"DEBUG: Saved MOVIEPY_FRAME_S{num_of_scene} to {debug_path_moviepy_frame}") fx_image_clip = base_image_clip try: scale_end_kb=random.uniform(1.03,1.08); fx_image_clip=base_image_clip.fx(vfx.resize,lambda t_val:1+(scale_end_kb-1)*(t_val/duration_for_scene) if duration_for_scene>0 else 1).set_position('center') except Exception as e_kb_fx: logger.error(f"S{num_of_scene} Ken Burns error: {e_kb_fx}",exc_info=False) active_scene_clip = fx_image_clip elif type_of_asset == 'video': # (Video processing logic from previous version) source_video_clip_obj=None try: source_video_clip_obj=VideoFileClip(path_of_asset,target_resolution=(self.video_frame_size[1],self.video_frame_size[0])if self.video_frame_size else None, audio=False) temp_video_clip_obj_loop=source_video_clip_obj if source_video_clip_obj.duration!=duration_for_scene: if source_video_clip_obj.duration>duration_for_scene:temp_video_clip_obj_loop=source_video_clip_obj.subclip(0,duration_for_scene) else: if duration_for_scene/source_video_clip_obj.duration > 1.5 and source_video_clip_obj.duration>0.1:temp_video_clip_obj_loop=source_video_clip_obj.loop(duration=duration_for_scene) else:temp_video_clip_obj_loop=source_video_clip_obj.set_duration(source_video_clip_obj.duration);logger.info(f"S{num_of_scene} Video clip ({source_video_clip_obj.duration:.2f}s) shorter than target ({duration_for_scene:.2f}s).") active_scene_clip=temp_video_clip_obj_loop.set_duration(duration_for_scene) if active_scene_clip.size!=list(self.video_frame_size):active_scene_clip=active_scene_clip.resize(self.video_frame_size) except Exception as e_vid_load_loop:logger.error(f"S{num_of_scene} Video load error '{path_of_asset}':{e_vid_load_loop}",exc_info=True);continue finally: if source_video_clip_obj and source_video_clip_obj is not active_scene_clip and hasattr(source_video_clip_obj,'close'):source_video_clip_obj.close() else: logger.warning(f"S{num_of_scene} Unknown asset type '{type_of_asset}'. Skip."); continue if active_scene_clip and action_in_key: # Text Overlay try: dur_text_overlay=min(active_scene_clip.duration-0.5,active_scene_clip.duration*0.8)if active_scene_clip.duration>0.5 else active_scene_clip.duration start_text_overlay=0.25 if dur_text_overlay > 0: text_clip_for_overlay=TextClip(f"Scene {num_of_scene}\n{action_in_key}",fontsize=self.VIDEO_OVERLAY_FONT_SIZE,color=self.VIDEO_OVERLAY_FONT_COLOR,font=self.active_moviepy_font_name,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(dur_text_overlay).set_start(start_text_overlay).set_position(('center',0.92),relative=True) active_scene_clip=CompositeVideoClip([active_scene_clip,text_clip_for_overlay],size=self.video_frame_size,use_bgclip=True) else: logger.warning(f"S{num_of_scene}: Text overlay duration zero. Skip text.") except Exception as e_txt_comp:logger.error(f"S{num_of_scene} TextClip error:{e_txt_comp}. No text.",exc_info=True) if active_scene_clip:processed_moviepy_clips_list.append(active_scene_clip);logger.info(f"S{num_of_scene} Processed. Dur:{active_scene_clip.duration:.2f}s.") except Exception as e_asset_loop_main:logger.error(f"MAJOR Error processing asset for S{num_of_scene} ({path_of_asset}):{e_asset_loop_main}",exc_info=True) finally: if active_scene_clip and hasattr(active_scene_clip,'close'): try: active_scene_clip.close() except: pass # Ignore errors during cleanup if not processed_moviepy_clips_list:logger.warning("No clips processed for animatic. Aborting.");return None transition_duration_val=0.75 # Renamed try: logger.info(f"Concatenating {len(processed_moviepy_clips_list)} clips for final animatic."); if len(processed_moviepy_clips_list)>1:final_video_output_clip=concatenate_videoclips(processed_moviepy_clips_list,padding=-transition_duration_val if transition_duration_val>0 else 0,method="compose") elif processed_moviepy_clips_list:final_video_output_clip=processed_moviepy_clips_list[0] if not final_video_output_clip:logger.error("Concatenation resulted in a None clip. Aborting.");return None logger.info(f"Concatenated animatic duration:{final_video_output_clip.duration:.2f}s") if transition_duration_val>0 and final_video_output_clip.duration>0: if final_video_output_clip.duration>transition_duration_val*2:final_video_output_clip=final_video_output_clip.fx(vfx.fadein,transition_duration_val).fx(vfx.fadeout,transition_duration_val) else:final_video_output_clip=final_video_output_clip.fx(vfx.fadein,min(transition_duration_val,final_video_output_clip.duration/2.0)) if overall_narration_path and os.path.exists(overall_narration_path) and final_video_output_clip.duration>0: try:narration_audio_clip_mvpy=AudioFileClip(overall_narration_path);final_video_output_clip=final_video_output_clip.set_audio(narration_audio_clip_mvpy);logger.info("Overall narration added to animatic.") except Exception as e_narr_add:logger.error(f"Error adding narration to animatic:{e_narr_add}",exc_info=True) elif final_video_output_clip.duration<=0:logger.warning("Animatic has no duration. Audio not added.") if final_video_output_clip and final_video_output_clip.duration>0: final_output_path_str=os.path.join(self.output_dir,output_filename);logger.info(f"Writing final animatic video to:{final_output_path_str} (Duration:{final_video_output_clip.duration:.2f}s)") # Renamed final_video_output_clip.write_videofile(final_output_path_str,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"Animatic video created successfully:{final_output_path_str}");return final_output_path_str else:logger.error("Final animatic clip is invalid or has zero duration. Cannot write video file.");return None except Exception as e_vid_write_final:logger.error(f"Error during final animatic video file writing or composition:{e_vid_write_final}",exc_info=True);return None # Renamed finally: logger.debug("Closing all MoviePy clips in `assemble_animatic_from_assets` main finally block.") clips_for_final_closure = processed_moviepy_clips_list + ([narration_audio_clip_mvpy] if narration_audio_clip_mvpy else []) + ([final_video_output_clip] if final_video_output_clip else []) # Renamed for clip_item_to_close in clips_for_final_closure: # Renamed if clip_item_to_close and hasattr(clip_item_to_close, 'close'): try: clip_item_to_close.close() except Exception as e_final_clip_close: logger.warning(f"Ignoring error while closing a MoviePy clip: {type(clip_item_to_close).__name__} - {e_final_clip_close}")