# core/visual_engine.py from PIL import Image, ImageDraw, ImageFont, ImageOps import base64 import mimetypes import numpy as np import os import openai import requests import io import time import random import logging from moviepy.editor import (ImageClip, VideoFileClip, concatenate_videoclips, TextClip, CompositeVideoClip, AudioFileClip) import moviepy.video.fx.all as vfx 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 version lacks common Resampling or ANTIALIAS. MoviePy effects might fail.") except Exception as e_mp: print(f"WARNING: ANTIALIAS monkey-patch error: {e_mp}") logger = logging.getLogger(__name__) # logger.setLevel(logging.DEBUG) 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_11l_imp: logger.warning(f"ElevenLabs client import failed: {e_11l_imp}. Audio disabled.") RUNWAYML_SDK_IMPORTED = False; RunwayMLAPIClientClass = None try: from runwayml import RunwayML as ImportedRunwayMLAPIClientClass RunwayMLAPIClientClass = ImportedRunwayMLAPIClientClass; RUNWAYML_SDK_IMPORTED = True logger.info("RunwayML SDK imported.") except Exception as e_rwy_imp: logger.warning(f"RunwayML SDK import failed: {e_rwy_imp}. RunwayML 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'; PREFERRED_MOVIEPY_FONT = 'Liberation-Sans-Bold' def __init__(self, output_dir="temp_cinegen_media", default_elevenlabs_voice_id="Rachel"): self.output_dir = output_dir try: os.makedirs(self.output_dir, exist_ok=True) logger.info(f"VisualEngine output directory set/ensured: {os.path.abspath(self.output_dir)}") except Exception as e_mkdir: logger.error(f"CRITICAL: Failed to create output directory '{self.output_dir}': {e_mkdir}", exc_info=True) # This is a critical failure; the app might not be able to save any files. # Consider raising the exception or setting a clear failure state for the engine. # raise OSError(f"Could not create output directory: {self.output_dir}") from e_mkdir self.font_filename_pil_preference = "DejaVuSans-Bold.ttf" font_paths = [ 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", f"/System/Library/Fonts/Supplemental/Arial.ttf", f"C:/Windows/Fonts/arial.ttf", f"/usr/local/share/fonts/truetype/mycustomfonts/arial.ttf"] self.resolved_font_path_pil = next((p for p in font_paths if os.path.exists(p)), None) self.active_font_pil = ImageFont.load_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}.") self.active_moviepy_font_name = 'DejaVu-Sans-Bold' if "dejavu" in self.resolved_font_path_pil.lower() else ('Liberation-Sans-Bold' if "liberation" in self.resolved_font_path_pil.lower() else self.DEFAULT_MOVIEPY_FONT) except IOError as e_font: logger.error(f"Pillow font IOError '{self.resolved_font_path_pil}': {e_font}. Using default.") else: logger.warning("Preferred Pillow font not found. Using default.") 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_instance = None self.elevenlabs_voice_id = default_elevenlabs_voice_id logger.info(f"VisualEngine __init__: ElevenLabs Voice ID initially set to: {self.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 if RUNWAYML_SDK_IMPORTED and RunwayMLAPIClientClass and os.getenv("RUNWAYML_API_SECRET"): try: self.runway_ml_sdk_client_instance = RunwayMLAPIClientClass(); self.USE_RUNWAYML = True; logger.info("RunwayML Client init from env var at startup.") except Exception as e_rwy_init: logger.error(f"Initial RunwayML client init failed: {e_rwy_init}"); self.USE_RUNWAYML = False logger.info("VisualEngine __init__ sequence complete.") 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 status: {'Ready' if self.USE_AI_IMAGE_GENERATION else 'Disabled'}") 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; logger.info(f"11L Voice ID updated via set_elevenlabs_api_key to: {self.elevenlabs_voice_id}") if api_key_value and ELEVENLABS_CLIENT_IMPORTED and ElevenLabsAPIClient: try: self.elevenlabs_client_instance = ElevenLabsAPIClient(api_key=api_key_value); self.USE_ELEVENLABS = bool(self.elevenlabs_client_instance); logger.info(f"11L Client: {'Ready' if self.USE_ELEVENLABS else 'Failed'} (Using Voice: {self.elevenlabs_voice_id})") except Exception as e_11l_setkey_init: logger.error(f"11L client init error: {e_11l_setkey_init}. Disabled.", exc_info=True); self.USE_ELEVENLABS=False; self.elevenlabs_client_instance=None else: self.USE_ELEVENLABS = False; logger.info(f"11L Disabled (API key not provided or SDK 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 status: {'Ready' if self.USE_PEXELS else 'Disabled'}") def set_runway_api_key(self, api_key_value): self.runway_api_key = api_key_value if api_key_value: if RUNWAYML_SDK_IMPORTED and RunwayMLAPIClientClass: if not self.runway_ml_sdk_client_instance: try: original_env_secret = os.getenv("RUNWAYML_API_SECRET") if not original_env_secret: os.environ["RUNWAYML_API_SECRET"] = api_key_value; logger.info("Temporarily set RUNWAYML_API_SECRET from provided key for SDK client init.") self.runway_ml_sdk_client_instance = RunwayMLAPIClientClass(); self.USE_RUNWAYML = True; logger.info("RunwayML Client initialized successfully via set_runway_api_key.") if not original_env_secret: del os.environ["RUNWAYML_API_SECRET"]; logger.info("Cleared temporary RUNWAYML_API_SECRET environment variable.") except Exception as e_runway_setkey_init: logger.error(f"RunwayML Client initialization in set_runway_api_key failed: {e_runway_setkey_init}", exc_info=True); self.USE_RUNWAYML=False;self.runway_ml_sdk_client_instance=None else: self.USE_RUNWAYML = True; logger.info("RunwayML Client was already initialized (likely from environment variable). API key stored.") else: logger.warning("RunwayML SDK not imported. API key stored, but current integration relies on SDK. Service effectively disabled."); self.USE_RUNWAYML = False else: self.USE_RUNWAYML = False; self.runway_ml_sdk_client_instance = None; logger.info("RunwayML Service Disabled (no API key provided).") 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} from ext '{ext}', using default {mime_type}.") with open(image_path, "rb") as image_file_handle: image_binary_data = image_file_handle.read() encoded_base64_string = base64.b64encode(image_binary_data).decode('utf-8') data_uri_string = f"data:{mime_type};base64,{encoded_base64_string}"; logger.debug(f"Data URI for {os.path.basename(image_path)} (MIME:{mime_type}): {data_uri_string[:100]}..."); return data_uri_string except FileNotFoundError: logger.error(f"Img not found {image_path} for data URI."); return None except Exception as e_data_uri: logger.error(f"Error converting {image_path} to data URI:{e_data_uri}", 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"Res {ratio_str} not in Gen-4 list. Default 1280:720 for Runway.");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) 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): if size is None: size = self.video_frame_size img = Image.new('RGB', size, color=(20, 20, 40)); d_draw = ImageDraw.Draw(img); padding = 25 max_w_text = size[0] - (2 * padding); lines_out = [] if not text_description: text_description = "(Placeholder Image)" words_in_desc = text_description.split(); current_line_buf = "" for word_idx_loop, word_val in enumerate(words_in_desc): prospective_add_str = word_val + (" " if word_idx_loop < len(words_in_desc) - 1 else "") test_line_str = current_line_buf + prospective_add_str current_w_val, _ = self._get_text_dimensions(test_line_str, self.active_font_pil) if current_w_val == 0 and test_line_str.strip(): current_w_val = len(test_line_str) * (self.active_font_size_pil * 0.6) if current_w_val <= max_w_text: current_line_buf = test_line_str else: if current_line_buf.strip(): lines_out.append(current_line_buf.strip()) current_line_buf = prospective_add_str if current_line_buf.strip(): lines_out.append(current_line_buf.strip()) if not lines_out and text_description: avg_char_w_val, _ = self._get_text_dimensions("W", self.active_font_pil); avg_char_w_val = avg_char_w_val or (self.active_font_size_pil * 0.6) chars_p_line = int(max_w_text / avg_char_w_val) if avg_char_w_val > 0 else 20 lines_out.append(text_description[:chars_p_line] + ("..." if len(text_description) > chars_p_line else "")) elif not lines_out: lines_out.append("(Placeholder Error)") _, single_line_h_val = self._get_text_dimensions("Ay", self.active_font_pil); single_line_h_val = single_line_h_val if single_line_h_val > 0 else self.active_font_size_pil + 2 max_lines_disp = min(len(lines_out), (size[1] - (2 * padding)) // (single_line_h_val + 2)) if single_line_h_val > 0 else 1; max_lines_disp = max(1, max_lines_disp) y_pos_text = padding + (size[1] - (2 * padding) - max_lines_disp * (single_line_h_val + 2)) / 2.0 for i_ln in range(max_lines_disp): line_content_str = lines_out[i_ln]; line_w_px, _ = self._get_text_dimensions(line_content_str, self.active_font_pil) if line_w_px == 0 and line_content_str.strip(): line_w_px = len(line_content_str) * (self.active_font_size_pil * 0.6) x_pos_text = (size[0] - line_w_px) / 2.0 try: d_draw.text((x_pos_text, y_pos_text), line_content_str, font=self.active_font_pil, fill=(200, 200, 180)) except Exception as e_draw_ph: logger.error(f"Pillow d.text error: {e_draw_ph} for '{line_content_str}'") y_pos_text += single_line_h_val + 2 if i_ln == 6 and max_lines_disp > 7: try: d_draw.text((x_pos_text, y_pos_text), "...", font=self.active_font_pil, fill=(200, 200, 180)) except Exception as e_elps_ph: logger.error(f"Pillow d.text ellipsis error: {e_elps_ph}"); break filepath_ph_img = os.path.join(self.output_dir, filename) try: img.save(filepath_ph_img); return filepath_ph_img except Exception as e_save_ph_img: logger.error(f"Saving placeholder image '{filepath_ph_img}' error: {e_save_ph_img}", exc_info=True); return None def _search_pexels_image(self, query_str_px, output_fn_base_px): if not self.USE_PEXELS or not self.pexels_api_key: return None http_headers_px = {"Authorization": self.pexels_api_key} http_params_px = {"query": query_str_px, "per_page": 1, "orientation": "landscape", "size": "large2x"} base_name_for_pexels_img, _ = os.path.splitext(output_fn_base_px) pexels_filename_output = base_name_for_pexels_img + f"_pexels_{random.randint(1000,9999)}.jpg" filepath_for_pexels_img = os.path.join(self.output_dir, pexels_filename_output) try: logger.info(f"Pexels: Searching for '{query_str_px}'") effective_query_for_pexels = " ".join(query_str_px.split()[:5]) http_params_px["query"] = effective_query_for_pexels response_from_pexels = requests.get("https://api.pexels.com/v1/search", headers=http_headers_px, params=http_params_px, timeout=20) response_from_pexels.raise_for_status() data_from_pexels = response_from_pexels.json() if data_from_pexels.get("photos") and len(data_from_pexels["photos"]) > 0: photo_details_item_px = data_from_pexels["photos"][0] photo_url_item_px = photo_details_item_px.get("src", {}).get("large2x") if not photo_url_item_px: logger.warning(f"Pexels: 'large2x' URL missing for '{effective_query_for_pexels}'. Details: {photo_details_item_px}"); return None image_response_get_px = requests.get(photo_url_item_px, timeout=60); image_response_get_px.raise_for_status() img_pil_data_from_pexels = Image.open(io.BytesIO(image_response_get_px.content)) if img_pil_data_from_pexels.mode != 'RGB': img_pil_data_from_pexels = img_pil_data_from_pexels.convert('RGB') img_pil_data_from_pexels.save(filepath_for_pexels_img); logger.info(f"Pexels: Image saved to {filepath_for_pexels_img}"); return filepath_for_pexels_img else: logger.info(f"Pexels: No photos for '{effective_query_for_pexels}'."); return None except requests.exceptions.RequestException as e_req_px_loop: logger.error(f"Pexels: RequestException for '{query_str_px}': {e_req_px_loop}", exc_info=False); return None except Exception as e_px_gen_loop: logger.error(f"Pexels: General error for '{query_str_px}': {e_px_gen_loop}", exc_info=True); return None def _generate_video_clip_with_runwayml(self, motion_prompt_rwy, input_img_path_rwy, scene_id_base_fn_rwy, duration_s_rwy=5): if not self.USE_RUNWAYML or not self.runway_ml_sdk_client_instance: logger.warning("RunwayML skip: Not enabled/client not init."); return None if not input_img_path_rwy or not os.path.exists(input_img_path_rwy): logger.error(f"Runway Gen-4 needs input img. Invalid: {input_img_path_rwy}"); return None img_data_uri_rwy = self._image_to_data_uri(input_img_path_rwy) if not img_data_uri_rwy: return None rwy_actual_dur = 10 if duration_s_rwy >= 8 else 5; rwy_actual_ratio = self._map_resolution_to_runway_ratio(self.video_frame_size[0],self.video_frame_size[1]) rwy_fn_base, _ = os.path.splitext(scene_id_base_fn_rwy); rwy_output_fn = rwy_fn_base + f"_runway_gen4_d{rwy_actual_dur}s.mp4"; rwy_output_fp = os.path.join(self.output_dir,rwy_output_fn) logger.info(f"Runway Gen-4 task: motion='{motion_prompt_rwy[:70]}...', img='{os.path.basename(input_img_path_rwy)}', dur={rwy_actual_dur}s, ratio='{rwy_actual_ratio}'") try: rwy_submitted_task = self.runway_ml_sdk_client_instance.image_to_video.create(model='gen4_turbo',prompt_image=img_data_uri_rwy,prompt_text=motion_prompt_rwy,duration=rwy_actual_dur,ratio=rwy_actual_ratio) rwy_task_id_val = rwy_submitted_task.id; logger.info(f"Runway task ID: {rwy_task_id_val}. Polling...") poll_interval_val=10;max_poll_attempts=36;poll_start_timestamp=time.time() while time.time()-poll_start_timestamp < max_poll_attempts*poll_interval_val: time.sleep(poll_interval_val);rwy_task_details_obj=self.runway_ml_sdk_client_instance.tasks.retrieve(id=rwy_task_id_val) logger.info(f"Runway task {rwy_task_id_val} status: {rwy_task_details_obj.status}") if rwy_task_details_obj.status=='SUCCEEDED': rwy_video_output_url=getattr(getattr(rwy_task_details_obj,'output',None),'url',None) or (getattr(rwy_task_details_obj,'artifacts',None)and rwy_task_details_obj.artifacts and hasattr(rwy_task_details_obj.artifacts[0],'url')and rwy_task_details_obj.artifacts[0].url) or (getattr(rwy_task_details_obj,'artifacts',None)and rwy_task_details_obj.artifacts and hasattr(rwy_task_details_obj.artifacts[0],'download_url')and rwy_task_details_obj.artifacts[0].download_url) if not rwy_video_output_url:logger.error(f"Runway task {rwy_task_id_val} SUCCEEDED, no output URL. Details:{vars(rwy_task_details_obj)if hasattr(rwy_task_details_obj,'__dict__')else rwy_task_details_obj}");return None logger.info(f"Runway task {rwy_task_id_val} SUCCEEDED. Downloading: {rwy_video_output_url}") runway_video_response=requests.get(rwy_video_output_url,stream=True,timeout=300);runway_video_response.raise_for_status() with open(rwy_output_fp,'wb')as f_out_vid: for data_chunk_vid in runway_video_response.iter_content(chunk_size=8192): f_out_vid.write(data_chunk_vid) logger.info(f"Runway Gen-4 video saved: {rwy_output_fp}");return rwy_output_fp elif rwy_task_details_obj.status in['FAILED','ABORTED','ERROR']: runway_error_detail=getattr(rwy_task_details_obj,'error_message',None)or getattr(getattr(rwy_task_details_obj,'output',None),'error',"Unknown Runway error.") logger.error(f"Runway task {rwy_task_id_val} status:{rwy_task_details_obj.status}. Error:{runway_error_detail}");return None logger.warning(f"Runway task {rwy_task_id_val} timed out.");return None except AttributeError as e_rwy_sdk_attr: logger.error(f"RunwayML SDK AttrError:{e_rwy_sdk_attr}. SDK methods changed?",exc_info=True);return None except Exception as e_rwy_general: logger.error(f"Runway Gen-4 API error:{e_rwy_general}",exc_info=True);return None def _create_placeholder_video_content(self, text_desc_ph, filename_ph, duration_ph=4, size_ph=None): if size_ph is None: size_ph = self.video_frame_size filepath_ph_vid = os.path.join(self.output_dir, filename_ph) text_clip_object_ph = None try: text_clip_object_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_object_ph.write_videofile(filepath_ph_vid, fps=24, codec='libx264', preset='ultrafast', logger=None, threads=2) logger.info(f"Generic placeholder video created: {filepath_ph_vid}") return filepath_ph_vid except Exception as e_placeholder_video_creation: logger.error(f"Failed to create generic placeholder video '{filepath_ph_vid}': {e_placeholder_video_creation}", exc_info=True) return None finally: if text_clip_object_ph and hasattr(text_clip_object_ph, 'close'): try: text_clip_object_ph.close() except Exception as e_close_placeholder_clip: logger.warning(f"Ignoring error closing placeholder TextClip: {e_close_placeholder_clip}") def generate_scene_asset(self, image_generation_prompt_text, motion_prompt_text_for_video, scene_data_dict, scene_identifier_fn_base, generate_as_video_clip_flag=False, runway_target_dur_val=5): base_name_current_asset, _ = os.path.splitext(scene_identifier_fn_base) asset_info_return_obj = {'path': None, 'type': 'none', 'error': True, 'prompt_used': image_generation_prompt_text, 'error_message': 'Asset generation init failed'} path_to_input_image_for_runway = None filename_for_base_image_output = base_name_current_asset + ("_base_for_video.png" if generate_as_video_clip_flag else ".png") filepath_for_base_image_output = os.path.join(self.output_dir, filename_for_base_image_output) if self.USE_AI_IMAGE_GENERATION and self.openai_api_key: max_retries_dalle, current_attempt_dalle = 2,0; for idx_dalle_attempt in range(max_retries_dalle): current_attempt_dalle = idx_dalle_attempt + 1 try: logger.info(f"Att {current_attempt_dalle} DALL-E (base img): {image_generation_prompt_text[:70]}..."); oai_client = openai.OpenAI(api_key=self.openai_api_key,timeout=90.0); oai_response = oai_client.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"); oai_image_url = oai_response.data[0].url; oai_revised_prompt = getattr(oai_response.data[0],'revised_prompt',None); if oai_revised_prompt: logger.info(f"DALL-E revised: {oai_revised_prompt[:70]}...") oai_image_get_response = requests.get(oai_image_url,timeout=120); oai_image_get_response.raise_for_status(); oai_pil_image = Image.open(io.BytesIO(oai_image_get_response.content)); if oai_pil_image.mode!='RGB': oai_pil_image=oai_pil_image.convert('RGB') oai_pil_image.save(filepath_for_base_image_output); logger.info(f"DALL-E base img saved: {filepath_for_base_image_output}"); path_to_input_image_for_runway=filepath_for_base_image_output; asset_info_return_obj={'path':filepath_for_base_image_output,'type':'image','error':False,'prompt_used':image_generation_prompt_text,'revised_prompt':oai_revised_prompt}; break except openai.RateLimitError as e_dalle_rl: logger.warning(f"OpenAI RateLimit Att {current_attempt_dalle}:{e_dalle_rl}.Retry...");time.sleep(5*current_attempt_dalle);asset_info_return_obj['error_message']=str(e_dalle_rl) except openai.APIError as e_dalle_api: logger.error(f"OpenAI APIError Att {current_attempt_dalle}:{e_dalle_api}");asset_info_return_obj['error_message']=str(e_dalle_api);break except requests.exceptions.RequestException as e_dalle_req: logger.error(f"Requests Err DALL-E Att {current_attempt_dalle}:{e_dalle_req}");asset_info_return_obj['error_message']=str(e_dalle_req);break except Exception as e_dalle_gen: logger.error(f"General DALL-E Err Att {current_attempt_dalle}:{e_dalle_gen}",exc_info=True);asset_info_return_obj['error_message']=str(e_dalle_gen);break if asset_info_return_obj['error']: logger.warning(f"DALL-E failed after {current_attempt_dalle} attempts for base img.") if asset_info_return_obj['error'] and self.USE_PEXELS: logger.info("Trying Pexels for base img.");pexels_query_text_val = scene_data_dictionary.get('pexels_search_query_감독',f"{scene_data_dictionary.get('emotional_beat','')} {scene_data_dictionary.get('setting_description','')}");pexels_path_result = self._search_pexels_image(pexels_query_text_val, filename_for_base_image_output); if pexels_path_result:path_to_input_image_for_runway=pexels_path_result;asset_info_return_obj={'path':pexels_path_result,'type':'image','error':False,'prompt_used':f"Pexels:{pexels_query_text_val}"} else:current_error_msg_pexels=asset_info_return_obj.get('error_message',"");asset_info_return_obj['error_message']=(current_error_msg_pexels+" Pexels failed for base.").strip() if asset_info_return_obj['error']: logger.warning("Base img (DALL-E/Pexels) failed. Using placeholder.");placeholder_prompt_text_val =asset_info_return_obj.get('prompt_used',image_generation_prompt_text);placeholder_path_result=self._create_placeholder_image_content(f"[Base Placeholder]{placeholder_prompt_text_val[:70]}...",filename_for_base_image_output); if placeholder_path_result:path_to_input_image_for_runway=placeholder_path_result;asset_info_return_obj={'path':placeholder_path_result,'type':'image','error':False,'prompt_used':placeholder_prompt_text_val} else:current_error_msg_ph=asset_info_return_obj.get('error_message',"");asset_info_return_obj['error_message']=(current_error_msg_ph+" Base placeholder failed.").strip() if generate_as_video_clip_flag: if not path_to_input_image_for_runway:logger.error("RunwayML video: base img failed.");asset_info_return_obj['error']=True;asset_info_return_obj['error_message']=(asset_info_return_obj.get('error_message',"")+" Base img miss, Runway abort.").strip();asset_info_return_obj['type']='none';return asset_info_return_obj if self.USE_RUNWAYML: runway_generated_video_path=self._generate_video_clip_with_runwayml(motion_prompt_text_for_video,path_to_input_image_for_runway,base_name_current_asset,runway_target_duration_val) if runway_generated_video_path and os.path.exists(runway_generated_video_path):asset_info_return_obj={'path':runway_generated_video_path,'type':'video','error':False,'prompt_used':motion_prompt_text_for_video,'base_image_path':path_to_input_image_for_runway} else:logger.warning(f"RunwayML video failed for {base_name_current_asset}. Fallback to base img.");asset_info_return_obj['error']=True;asset_info_return_obj['error_message']=(asset_info_return_obj.get('error_message',"Base img ok.")+" RunwayML video fail; use base img.").strip();asset_info_return_obj['path']=path_to_input_image_for_runway;asset_info_return_obj['type']='image';asset_info_return_obj['prompt_used']=image_generation_prompt_text else:logger.warning("RunwayML selected but disabled. Use base img.");asset_info_return_obj['error']=True;asset_info_return_obj['error_message']=(asset_info_return_obj.get('error_message',"Base img ok.")+" RunwayML disabled; use base img.").strip();asset_info_return_obj['path']=path_to_input_image_for_runway;asset_info_return_obj['type']='image';asset_info_return_obj['prompt_used']=image_generation_prompt_text return asset_info_return_obj def generate_narration_audio(self, narration_text, output_fn="narration_overall.mp3"): if not self.USE_ELEVENLABS or not self.elevenlabs_client_instance or not narration_text: logger.info("11L conditions not met. Skip audio."); return None narration_fp = os.path.join(self.output_dir, output_fn) try: logger.info(f"11L audio (Voice:{self.elevenlabs_voice_id}): \"{narration_text[:70]}...\"") stream_method = None if hasattr(self.elevenlabs_client_instance,'text_to_speech') and hasattr(self.elevenlabs_client_instance.text_to_speech,'stream'): stream_method=self.elevenlabs_client_instance.text_to_speech.stream; logger.info("Using 11L .text_to_speech.stream()") elif hasattr(self.elevenlabs_client_instance,'generate_stream'): stream_method=self.elevenlabs_client_instance.generate_stream; logger.info("Using 11L .generate_stream()") elif hasattr(self.elevenlabs_client_instance,'generate'): logger.info("Using 11L .generate() (non-streaming).") voice_p = 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) audio_b = self.elevenlabs_client_instance.generate(text=narration_text,voice=voice_p,model="eleven_multilingual_v2") with open(narration_fp,"wb") as f_audio: f_audio.write(audio_b); logger.info(f"11L audio (non-stream): {narration_fp}"); return narration_fp else: logger.error("No recognized 11L audio method."); return None if stream_method: voice_stream_params={"voice_id":str(self.elevenlabs_voice_id)} if self.elevenlabs_voice_settings_obj: if hasattr(self.elevenlabs_voice_settings_obj,'model_dump'): voice_stream_params["voice_settings"]=self.elevenlabs_voice_settings_obj.model_dump() elif hasattr(self.elevenlabs_voice_settings_obj,'dict'): voice_stream_params["voice_settings"]=self.elevenlabs_voice_settings_obj.dict() else: voice_stream_params["voice_settings"]=self.elevenlabs_voice_settings_obj audio_iter = stream_method(text=narration_text,model_id="eleven_multilingual_v2",**voice_stream_params) with open(narration_fp,"wb") as f_audio_stream: for chunk_item in audio_iter: if chunk_item: f_audio_stream.write(chunk_item) logger.info(f"11L audio (stream): {narration_fp}"); return narration_fp except AttributeError as e_11l_attr: logger.error(f"11L SDK AttrError: {e_11l_attr}. SDK/methods changed?", exc_info=True); return None except Exception as e_11l_gen: logger.error(f"11L audio gen error: {e_11l_gen}", exc_info=True); return None def assemble_animatic_from_assets(self, asset_data_list, overall_narration_path=None, output_filename="final_video.mp4", fps=24): if not asset_data_list: logger.warning("No assets for animatic."); return None processed_moviepy_clips_list = []; narration_audio_clip_mvpy = None; final_video_output_clip = None 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): 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 try: if type_of_asset == 'image': pil_img_original = Image.open(path_of_asset); logger.debug(f"S{num_of_scene} (0-Load): Original. Mode:{pil_img_original.mode}, Size:{pil_img_original.size}"); pil_img_original.save(os.path.join(self.output_dir,f"debug_0_ORIGINAL_S{num_of_scene}.png")) img_rgba_intermediate = pil_img_original.convert('RGBA') if pil_img_original.mode != 'RGBA' else pil_img_original.copy().convert('RGBA'); logger.debug(f"S{num_of_scene} (1-ToRGBA): Mode:{img_rgba_intermediate.mode}, Size:{img_rgba_intermediate.size}"); img_rgba_intermediate.save(os.path.join(self.output_dir,f"debug_1_AS_RGBA_S{num_of_scene}.png")) thumbnailed_img_rgba = img_rgba_intermediate.copy(); resample_filter_pil = Image.Resampling.LANCZOS if hasattr(Image.Resampling,'LANCZOS') else Image.BILINEAR; thumbnailed_img_rgba.thumbnail(self.video_frame_size, resample_filter_pil); logger.debug(f"S{num_of_scene} (2-Thumbnail): Mode:{thumbnailed_img_rgba.mode}, Size:{thumbnailed_img_rgba.size}"); thumbnailed_img_rgba.save(os.path.join(self.output_dir,f"debug_2_THUMBNAIL_RGBA_S{num_of_scene}.png")) canvas_for_compositing_rgba = Image.new('RGBA', self.video_frame_size, (0,0,0,0)); pos_x_paste = (self.video_frame_size[0] - thumbnailed_img_rgba.width) // 2; pos_y_paste = (self.video_frame_size[1] - thumbnailed_img_rgba.height) // 2; canvas_for_compositing_rgba.paste(thumbnailed_img_rgba, (pos_x_paste, pos_y_paste), thumbnailed_img_rgba); logger.debug(f"S{num_of_scene} (3-PasteOnRGBA): Mode:{canvas_for_compositing_rgba.mode}, Size:{canvas_for_compositing_rgba.size}"); canvas_for_compositing_rgba.save(os.path.join(self.output_dir,f"debug_3_COMPOSITED_RGBA_S{num_of_scene}.png")) final_rgb_image_for_pil = Image.new("RGB", self.video_frame_size, (0, 0, 0)); if canvas_for_compositing_rgba.mode == 'RGBA': final_rgb_image_for_pil.paste(canvas_for_compositing_rgba, mask=canvas_for_compositing_rgba.split()[3]) else: final_rgb_image_for_pil.paste(canvas_for_compositing_rgba) logger.debug(f"S{num_of_scene} (4-ToRGB): Final RGB. Mode:{final_rgb_image_for_pil.mode}, Size:{final_rgb_image_for_pil.size}") debug_path_img_pre_numpy = os.path.join(self.output_dir,f"debug_4_PRE_NUMPY_RGB_S{num_of_scene}.png"); final_rgb_image_for_pil.save(debug_path_img_pre_numpy); logger.info(f"CRITICAL DEBUG: Saved PRE_NUMPY_RGB_S{num_of_scene} to {debug_path_img_pre_numpy}") numpy_frame_arr = np.array(final_rgb_image_for_pil, dtype=np.uint8) # Explicit dtype 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} (5-NumPy): Final NumPy. Shape:{numpy_frame_arr.shape}, DType:{numpy_frame_arr.dtype}, Flags:{numpy_frame_arr.flags}") 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 shape/size ({numpy_frame_arr.shape}). Skipping."); continue base_image_clip_mvpy = ImageClip(numpy_frame_arr, transparent=False, ismask=False).set_duration(duration_for_scene) logger.debug(f"S{num_of_scene} (6-ImageClip): Base ImageClip. Duration: {base_image_clip_mvpy.duration}") debug_path_moviepy_frame = os.path.join(self.output_dir,f"debug_7_MOVIEPY_FRAME_S{num_of_scene}.png") # <<< CORRECTED TRY-EXCEPT FOR save_frame >>> try: save_frame_time = min(0.1, base_image_clip_mvpy.duration / 2 if base_image_clip_mvpy.duration > 0 else 0.1) base_image_clip_mvpy.save_frame(debug_path_moviepy_frame, t=save_frame_time) logger.info(f"CRITICAL DEBUG: Saved frame FROM MOVIEPY ImageClip S{num_of_scene} to {debug_path_moviepy_frame}") except Exception as e_save_mvpy_frame: logger.error(f"DEBUG: Error saving frame FROM MOVIEPY ImageClip S{num_of_scene}: {e_save_mvpy_frame}", exc_info=True) fx_image_clip_mvpy = base_image_clip_mvpy try: # Ken Burns scale_end_kb_val = random.uniform(1.03, 1.08) if duration_for_scene > 0: fx_image_clip_mvpy = base_image_clip_mvpy.fx(vfx.resize, lambda t_val: 1 + (scale_end_kb_val - 1) * (t_val / duration_for_scene)).set_position('center'); logger.debug(f"S{num_of_scene} (8-KenBurns): Ken Burns applied.") else: logger.warning(f"S{num_of_scene}: Duration zero, skipping Ken Burns.") except Exception as e_kb_fx_loop: logger.error(f"S{num_of_scene} Ken Burns error: {e_kb_fx_loop}", exc_info=False) active_scene_clip = fx_image_clip_mvpy elif type_of_asset == 'video': # (Video processing logic as before) source_video_clip_obj=None try: logger.debug(f"S{num_of_scene}: Loading VIDEO asset: {path_of_asset}") 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) logger.debug(f"S{num_of_scene}: Video asset processed. Final duration: {active_scene_clip.duration:.2f}s") 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'): try: source_video_clip_obj.close() except Exception as e_close_src_vid: logger.warning(f"S{num_of_scene}: Error closing source VideoFileClip: {e_close_src_vid}") else: logger.warning(f"S{num_of_scene} Unknown asset type '{type_of_asset}'. Skipping."); continue if active_scene_clip and action_in_key: # Text Overlay try: dur_text_overlay_val=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 if active_scene_clip.duration > 0 else 0) start_text_overlay_val=0.25 if active_scene_clip.duration > 0.5 else 0 if dur_text_overlay_val > 0: text_clip_for_overlay_obj=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_val).set_start(start_text_overlay_val).set_position(('center',0.92),relative=True) active_scene_clip=CompositeVideoClip([active_scene_clip,text_clip_for_overlay_obj],size=self.video_frame_size,use_bgclip=True) logger.debug(f"S{num_of_scene}: Text overlay composited.") else: logger.warning(f"S{num_of_scene}: Text overlay duration zero or negative ({dur_text_overlay_val}). Skipping text overlay.") except Exception as e_txt_comp_loop:logger.error(f"S{num_of_scene} TextClip compositing error:{e_txt_comp_loop}. Proceeding without text for this scene.",exc_info=True) if active_scene_clip: processed_moviepy_clips_list.append(active_scene_clip); logger.info(f"S{num_of_scene}: Asset successfully processed. Clip duration: {active_scene_clip.duration:.2f}s. Added to final list.") except Exception as e_asset_loop_main_exc: logger.error(f"MAJOR UNHANDLED ERROR processing asset for S{num_of_scene} (Path: {path_of_asset}): {e_asset_loop_main_exc}", exc_info=True) finally: if active_scene_clip and hasattr(active_scene_clip,'close'): try: active_scene_clip.close() except Exception as e_close_active_err: logger.warning(f"S{num_of_scene}: Error closing active_scene_clip in error handler: {e_close_active_err}") if not processed_moviepy_clips_list: logger.warning("No MoviePy clips were successfully processed. Aborting animatic assembly before concatenation."); return None transition_duration_val=0.75 try: logger.info(f"Concatenating {len(processed_moviepy_clips_list)} processed 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 base 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)) logger.debug("Applied fade in/out effects to final composite clip.") 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); logger.info(f"Adding overall narration. Video duration: {final_video_output_clip.duration:.2f}s, Narration duration: {narration_audio_clip_mvpy.duration:.2f}s"); final_video_output_clip=final_video_output_clip.set_audio(narration_audio_clip_mvpy); logger.info("Overall narration successfully added to animatic.") except Exception as e_narr_add_final:logger.error(f"Error adding overall narration to animatic:{e_narr_add_final}",exc_info=True) elif final_video_output_clip.duration <= 0: logger.warning("Animatic has zero or negative duration before adding audio. Audio will not be 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} (Target Duration: {final_video_output_clip.duration:.2f}s)") num_threads = os.cpu_count(); num_threads = num_threads if isinstance(num_threads, int) and num_threads >= 1 else 2 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=num_threads, 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_op: logger.error(f"Error during final animatic video file writing or composition stage: {e_vid_write_final_op}", exc_info=True); return None finally: logger.debug("Closing all MoviePy clips in `assemble_animatic_from_assets` main finally block.") all_clips_for_closure = processed_moviepy_clips_list[:] if narration_audio_clip_mvpy and hasattr(narration_audio_clip_mvpy, 'close'): all_clips_for_closure.append(narration_audio_clip_mvpy) if final_video_output_clip and hasattr(final_video_output_clip, 'close'): all_clips_for_closure.append(final_video_output_clip) for clip_to_close_item_final in all_clips_for_closure: if clip_to_close_item_final and hasattr(clip_to_close_item_final, 'close'): try: clip_to_close_item_final.close() except Exception as e_final_clip_close_op: logger.warning(f"Ignoring error while closing a MoviePy clip ({type(clip_to_close_item_final).__name__}): {e_final_clip_close_op}")