mgbam commited on
Commit
7b5fcd5
·
verified ·
1 Parent(s): a48cea9

Update core/visual_engine.py

Browse files
Files changed (1) hide show
  1. core/visual_engine.py +664 -448
core/visual_engine.py CHANGED
@@ -1,41 +1,39 @@
1
  # core/visual_engine.py
2
  from PIL import Image, ImageDraw, ImageFont, ImageOps
3
- import base64
4
- import mimetypes
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5
  import numpy as np
6
  import os
7
- import openai # Ensure this is OpenAI v1.x.x+
8
  import requests
9
  import io
10
  import time
11
  import random
12
  import logging
13
 
14
- # --- MoviePy Imports ---
15
- from moviepy.editor import (ImageClip, VideoFileClip, concatenate_videoclips, TextClip,
16
- CompositeVideoClip, AudioFileClip)
17
- import moviepy.video.fx.all as vfx
18
-
19
- # --- MONKEY PATCH for Pillow/MoviePy compatibility ---
20
- try:
21
- if hasattr(Image, 'Resampling') and hasattr(Image.Resampling, 'LANCZOS'): # Pillow 9+
22
- if not hasattr(Image, 'ANTIALIAS'): Image.ANTIALIAS = Image.Resampling.LANCZOS
23
- elif hasattr(Image, 'LANCZOS'): # Pillow 8
24
- if not hasattr(Image, 'ANTIALIAS'): Image.ANTIALIAS = Image.LANCZOS
25
- elif not hasattr(Image, 'ANTIALIAS'):
26
- print("WARNING: Pillow version lacks common Resampling attributes or ANTIALIAS. MoviePy effects might fail or look different.")
27
- except Exception as e_monkey_patch:
28
- print(f"WARNING: An unexpected error occurred during Pillow ANTIALIAS monkey-patch: {e_monkey_patch}")
29
-
30
  logger = logging.getLogger(__name__)
31
- # Set a default level; can be overridden by the main app's logging config
32
- # logger.setLevel(logging.DEBUG) # Uncomment for very verbose output during development
33
 
34
- # --- External Service Client Imports ---
35
  ELEVENLABS_CLIENT_IMPORTED = False
36
- ElevenLabsAPIClient = None # Class placeholder
37
- Voice = None # Class placeholder
38
- VoiceSettings = None # Class placeholder
39
  try:
40
  from elevenlabs.client import ElevenLabs as ImportedElevenLabsClient
41
  from elevenlabs import Voice as ImportedVoice, VoiceSettings as ImportedVoiceSettings
@@ -43,476 +41,694 @@ try:
43
  Voice = ImportedVoice
44
  VoiceSettings = ImportedVoiceSettings
45
  ELEVENLABS_CLIENT_IMPORTED = True
46
- logger.info("ElevenLabs client components (SDK v1.x.x pattern) imported successfully.")
47
- except ImportError:
48
- logger.warning("ElevenLabs SDK not found (expected 'pip install elevenlabs>=1.0.0'). Audio generation will be disabled.")
49
- except Exception as e_eleven_import_general:
50
- logger.warning(f"General error importing ElevenLabs client components: {e_eleven_import_general}. Audio generation disabled.")
51
 
 
52
  RUNWAYML_SDK_IMPORTED = False
53
- RunwayMLAPIClientClass = None # Storing the class itself
54
  try:
55
- from runwayml import RunwayML as ImportedRunwayMLAPIClientClass # Actual SDK import
56
- RunwayMLAPIClientClass = ImportedRunwayMLAPIClientClass
57
- RUNWAYML_SDK_IMPORTED = True
58
- logger.info("RunwayML SDK (runwayml) imported successfully.")
59
  except ImportError:
60
- logger.warning("RunwayML SDK not found (pip install runwayml). RunwayML video generation will be disabled.")
61
- except Exception as e_runway_sdk_import_general:
62
- logger.warning(f"General error importing RunwayML SDK: {e_runway_sdk_import_general}. RunwayML features disabled.")
63
 
64
 
65
  class VisualEngine:
66
- DEFAULT_FONT_SIZE_PIL = 10
67
- PREFERRED_FONT_SIZE_PIL = 20
68
- VIDEO_OVERLAY_FONT_SIZE = 30
69
- VIDEO_OVERLAY_FONT_COLOR = 'white'
70
- DEFAULT_MOVIEPY_FONT = 'DejaVu-Sans-Bold' # Common ImageMagick font name
71
- PREFERRED_MOVIEPY_FONT = 'Liberation-Sans-Bold'
72
-
73
  def __init__(self, output_dir="temp_cinegen_media", default_elevenlabs_voice_id="Rachel"):
74
  self.output_dir = output_dir
75
  os.makedirs(self.output_dir, exist_ok=True)
76
-
77
- self.font_filename_pil_preference = "DejaVuSans-Bold.ttf" # More standard Linux font
78
  font_paths_to_try = [
79
- self.font_filename_pil_preference,
80
- f"/usr/share/fonts/truetype/dejavu/{self.font_filename_pil_preference}",
81
- f"/usr/share/fonts/truetype/liberation/LiberationSans-Bold.ttf", # Alternative
82
  f"/System/Library/Fonts/Supplemental/Arial.ttf",
83
  f"C:/Windows/Fonts/arial.ttf",
84
- f"/usr/local/share/fonts/truetype/mycustomfonts/arial.ttf" # Previous custom
85
  ]
86
- self.resolved_font_path_pil = next((p for p in font_paths_to_try if os.path.exists(p)), None)
87
-
88
- self.active_font_pil = ImageFont.load_default() # Fallback default
89
- self.active_font_size_pil = self.DEFAULT_FONT_SIZE_PIL
90
- self.active_moviepy_font_name = self.DEFAULT_MOVIEPY_FONT
91
 
92
- if self.resolved_font_path_pil:
93
- try:
94
- self.active_font_pil = ImageFont.truetype(self.resolved_font_path_pil, self.PREFERRED_FONT_SIZE_PIL)
95
- self.active_font_size_pil = self.PREFERRED_FONT_SIZE_PIL
96
- logger.info(f"Pillow font loaded: {self.resolved_font_path_pil} at size {self.active_font_size_pil}.")
97
- # Determine MoviePy font based on loaded PIL font for consistency
98
- if "dejavu" in self.resolved_font_path_pil.lower(): self.active_moviepy_font_name = 'DejaVu-Sans-Bold'
99
- elif "liberation" in self.resolved_font_path_pil.lower(): self.active_moviepy_font_name = 'Liberation-Sans-Bold'
100
- except IOError as e_font_load_io:
101
- logger.error(f"Pillow font loading IOError for '{self.resolved_font_path_pil}': {e_font_load_io}. Using default font.")
102
- else:
103
- logger.warning("Preferred Pillow font not found in predefined paths. Using default font.")
104
-
105
- # Service API keys and flags
106
- self.openai_api_key = None; self.USE_AI_IMAGE_GENERATION = False
107
- self.dalle_model = "dall-e-3"; self.image_size_dalle3 = "1792x1024" # DALL-E 3 landscape
108
- self.video_frame_size = (1280, 720) # HD 16:9
109
-
110
- self.elevenlabs_api_key = None; self.USE_ELEVENLABS = False; self.elevenlabs_client_instance = None
 
 
111
  self.elevenlabs_voice_id = default_elevenlabs_voice_id
112
  if VoiceSettings and ELEVENLABS_CLIENT_IMPORTED:
113
- self.elevenlabs_voice_settings_obj = VoiceSettings(stability=0.60, similarity_boost=0.80, style=0.15, use_speaker_boost=True)
114
- else: self.elevenlabs_voice_settings_obj = None
115
-
116
- self.pexels_api_key = None; self.USE_PEXELS = False
117
- self.runway_api_key = None; self.USE_RUNWAYML = False; self.runway_ml_sdk_client_instance = None # Instance of RunwayML()
118
-
119
- # Attempt to initialize RunwayML client if SDK is present and RUNWAYML_API_SECRET env var might be set
120
- if RUNWAYML_SDK_IMPORTED and RunwayMLAPIClientClass and os.getenv("RUNWAYML_API_SECRET"):
121
- try:
122
- self.runway_ml_sdk_client_instance = RunwayMLAPIClientClass() # SDK uses env var by default
123
- self.USE_RUNWAYML = True # Assume enabled if client initializes without error
124
- logger.info("RunwayML Client initialized using RUNWAYML_API_SECRET environment variable at startup.")
125
- except Exception as e_runway_init_at_startup:
126
- logger.error(f"Initial RunwayML client initialization failed (env var RUNWAYML_API_SECRET might be invalid or SDK issue): {e_runway_init_at_startup}")
127
- self.USE_RUNWAYML = False # Ensure it's disabled if init fails
128
-
129
  logger.info("VisualEngine initialized.")
130
 
131
- # --- API Key Setter Methods ---
132
- def set_openai_api_key(self, api_key_value):
133
- self.openai_api_key = api_key_value; self.USE_AI_IMAGE_GENERATION = bool(api_key_value)
134
- logger.info(f"DALL-E ({self.dalle_model}) service status: {'Ready' if self.USE_AI_IMAGE_GENERATION else 'Disabled (no API key)'}")
135
 
136
- def set_elevenlabs_api_key(self, api_key_value, voice_id_from_secret=None):
137
- self.elevenlabs_api_key = api_key_value
138
- if voice_id_from_secret: self.elevenlabs_voice_id = voice_id_from_secret
139
-
140
- if api_key_value and ELEVENLABS_CLIENT_IMPORTED and ElevenLabsAPIClient:
141
  try:
142
- self.elevenlabs_client_instance = ElevenLabsAPIClient(api_key=api_key_value) # Pass key directly
143
- self.USE_ELEVENLABS = bool(self.elevenlabs_client_instance)
144
- logger.info(f"ElevenLabs Client service status: {'Ready' if self.USE_ELEVENLABS else 'Failed Initialization'} (Using Voice ID: {self.elevenlabs_voice_id})")
145
- except Exception as e_11l_init:
146
- logger.error(f"ElevenLabs client initialization error: {e_11l_init}. Service Disabled.", exc_info=True)
147
- self.USE_ELEVENLABS = False; self.elevenlabs_client_instance = None
148
  else:
149
  self.USE_ELEVENLABS = False
150
- logger.info(f"ElevenLabs Service Disabled (API key not provided or SDK import issue).")
151
 
152
- def set_pexels_api_key(self, api_key_value):
153
- self.pexels_api_key = api_key_value; self.USE_PEXELS = bool(api_key_value)
154
- logger.info(f"Pexels Search service status: {'Ready' if self.USE_PEXELS else 'Disabled (no API key)'}")
 
155
 
156
- def set_runway_api_key(self, api_key_value):
157
- self.runway_api_key = api_key_value # Store the key itself
158
- if api_key_value:
159
- if RUNWAYML_SDK_IMPORTED and RunwayMLAPIClientClass:
160
- if not self.runway_ml_sdk_client_instance: # If not already initialized (e.g., by env var at startup)
161
- try:
162
- # The RunwayML Python SDK expects the API key via the RUNWAYML_API_SECRET env var.
163
- original_env_secret = os.getenv("RUNWAYML_API_SECRET")
164
- if not original_env_secret: # If env var not set, set it temporarily from passed key
165
- logger.info("Temporarily setting RUNWAYML_API_SECRET from provided key for SDK client initialization.")
166
- os.environ["RUNWAYML_API_SECRET"] = api_key_value
167
-
168
- self.runway_ml_sdk_client_instance = RunwayMLAPIClientClass()
169
- self.USE_RUNWAYML = True # Mark service as usable if client initializes
170
- logger.info("RunwayML Client initialized successfully using provided API key (via env var).")
171
-
172
- if not original_env_secret: # Clean up: remove env var if we set it
173
- del os.environ["RUNWAYML_API_SECRET"]
174
- logger.info("Cleared temporary RUNWAYML_API_SECRET environment variable.")
175
- except Exception as e_runway_client_setkey_init:
176
- logger.error(f"RunwayML Client initialization via set_runway_api_key failed: {e_runway_client_setkey_init}", exc_info=True)
177
- self.USE_RUNWAYML = False; self.runway_ml_sdk_client_instance = None # Ensure it's disabled
178
- else: # Client was already initialized
179
- self.USE_RUNWAYML = True # Service is usable
180
- logger.info("RunwayML Client was already initialized (likely from environment variable). API key stored.")
181
- else: # SDK not imported
182
- logger.warning("RunwayML SDK not imported. API key has been stored, but the current integration relies on the SDK. Service effectively disabled.")
183
- self.USE_RUNWAYML = False # Can't use without SDK if that's the implemented path
184
- else: # No API key provided
185
- self.USE_RUNWAYML = False; self.runway_ml_sdk_client_instance = None
186
- logger.info("RunwayML Service Disabled (no API key provided to set_runway_api_key).")
187
-
188
- # --- Helper Methods (_image_to_data_uri, _map_resolution_to_runway_ratio, etc.) ---
189
- def _image_to_data_uri(self, image_path):
190
- try:
191
- mime_type, _ = mimetypes.guess_type(image_path)
192
- if not mime_type:
193
- ext = os.path.splitext(image_path)[1].lower()
194
- mime_map = {".png": "image/png", ".jpg": "image/jpeg", ".jpeg": "image/jpeg", ".webp": "image/webp"}
195
- mime_type = mime_map.get(ext, "application/octet-stream")
196
- if mime_type == "application/octet-stream": logger.warning(f"Could not determine MIME type for {image_path}, using default.")
197
- with open(image_path, "rb") as image_file: encoded_string = base64.b64encode(image_file.read()).decode('utf-8')
198
- data_uri = f"data:{mime_type};base64,{encoded_string}"
199
- logger.debug(f"Generated data URI for {os.path.basename(image_path)} (first 100 chars): {data_uri[:100]}...")
200
- return data_uri
201
- except FileNotFoundError: logger.error(f"Image file not found at {image_path} for data URI conversion."); return None
202
- except Exception as e: logger.error(f"Error converting image {image_path} to data URI: {e}", exc_info=True); return None
203
-
204
- def _map_resolution_to_runway_ratio(self, width, height):
205
- ratio_str = f"{width}:{height}"
206
- supported_ratios_gen4 = ["1280:720", "720:1280", "1104:832", "832:1104", "960:960", "1584:672"]
207
- if ratio_str in supported_ratios_gen4: return ratio_str
208
- logger.warning(f"Resolution {ratio_str} not directly in Gen-4 supported list. Defaulting to 1280:720 for RunwayML.")
209
- return "1280:720"
210
 
211
- def _get_text_dimensions(self, text_content, font_object_pil):
212
- default_h = getattr(font_object_pil, 'size', self.active_font_size_pil)
213
- if not text_content: return 0, default_h
 
214
  try:
215
- 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
216
- elif hasattr(font_object_pil,'getsize'): w,h=font_object_pil.getsize(text_content); return w, h if h > 0 else default_h
217
- else: return int(len(text_content)*default_h*0.6),int(default_h*1.2) # Basic estimate
218
- 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)
 
 
 
 
 
 
 
 
 
219
 
220
  def _create_placeholder_image_content(self, text_description, filename, size=None):
221
- # <<< THIS IS THE CORRECTED VERSION OF THIS METHOD >>>
222
- if size is None: size = self.video_frame_size
223
- img = Image.new('RGB', size, color=(20, 20, 40)); d = ImageDraw.Draw(img); padding = 25
224
- max_w = size[0] - (2 * padding); lines_for_placeholder = [] # Renamed to avoid conflict
225
- if not text_description: text_description = "(Placeholder Image)"
226
- words_list = text_description.split(); current_line_buffer = "" # Renamed
227
- for word_idx, word_item in enumerate(words_list): # Renamed
228
- prospective_add = word_item + (" " if word_idx < len(words_list) - 1 else "")
229
- test_line_candidate = current_line_buffer + prospective_add
230
- current_w_text, _ = self._get_text_dimensions(test_line_candidate, self.active_font_pil)
231
- if current_w_text == 0 and test_line_candidate.strip(): current_w_text = len(test_line_candidate) * (self.active_font_size_pil * 0.6)
232
-
233
- if current_w_text <= max_w: current_line_buffer = test_line_candidate
 
 
 
234
  else:
235
- if current_line_buffer.strip(): lines_for_placeholder.append(current_line_buffer.strip())
236
- current_line_buffer = prospective_add
237
- if current_line_buffer.strip(): lines_for_placeholder.append(current_line_buffer.strip())
238
-
239
- if not lines_for_placeholder and text_description:
240
- 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)
241
- chars_per_line_est = int(max_w / avg_char_w_est) if avg_char_w_est > 0 else 20
242
- lines_for_placeholder.append(text_description[:chars_per_line_est] + ("..." if len(text_description) > chars_per_line_est else ""))
243
- elif not lines_for_placeholder: lines_for_placeholder.append("(Placeholder Error)")
244
-
245
- _, 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
246
- 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)
247
-
248
- y_p = padding + (size[1] - (2 * padding) - max_l * (single_h + 2)) / 2.0
249
- for i_line in range(max_l): # Renamed
250
- line_txt_content = lines_for_placeholder[i_line]; line_w_val, _ = self._get_text_dimensions(line_txt_content, self.active_font_pil)
251
- if line_w_val == 0 and line_txt_content.strip(): line_w_val = len(line_txt_content) * (self.active_font_size_pil * 0.6)
252
- x_p = (size[0] - line_w_val) / 2.0
253
- try: d.text((x_p, y_p), line_txt_content, font=self.active_font_pil, fill=(200, 200, 180))
254
- except Exception as e_draw_txt: logger.error(f"Pillow d.text error: {e_draw_txt} for '{line_txt_content}'")
255
- y_p += single_h + 2
256
- if i_line == 6 and max_l > 7:
257
- try: d.text((x_p, y_p), "...", font=self.active_font_pil, fill=(200, 200, 180))
258
- except Exception as e_elps: logger.error(f"Pillow d.text ellipsis error: {e_elps}"); break
259
- filepath_placeholder = os.path.join(self.output_dir, filename) # Renamed
260
- try: img.save(filepath_placeholder); return filepath_placeholder
261
- except Exception as e_save_ph: logger.error(f"Saving placeholder image '{filepath_placeholder}' error: {e_save_ph}", exc_info=True); return None
262
-
263
- def _search_pexels_image(self, query_str, output_fn_base):
264
- # (Corrected from previous response)
265
- if not self.USE_PEXELS or not self.pexels_api_key: return None
266
- http_headers = {"Authorization": self.pexels_api_key}
267
- http_params = {"query": query_str, "per_page": 1, "orientation": "landscape", "size": "large2x"}
268
- base_name_px, _ = os.path.splitext(output_fn_base)
269
- pexels_fn_str = base_name_px + f"_pexels_{random.randint(1000,9999)}.jpg"
270
- file_path_px = os.path.join(self.output_dir, pexels_fn_str)
 
 
 
 
 
 
 
 
 
271
  try:
272
- logger.info(f"Pexels: Searching for '{query_str}'")
273
- eff_query_px = " ".join(query_str.split()[:5])
274
- http_params["query"] = eff_query_px
275
- response_px = requests.get("https://api.pexels.com/v1/search", headers=http_headers, params=http_params, timeout=20)
276
- response_px.raise_for_status()
277
- data_px = response_px.json()
278
- if data_px.get("photos") and len(data_px["photos"]) > 0:
279
- photo_details_px = data_px["photos"][0]
280
- photo_url_px = photo_details_px.get("src", {}).get("large2x")
281
- if not photo_url_px: logger.warning(f"Pexels: 'large2x' URL missing for '{eff_query_px}'. Details: {photo_details_px}"); return None
282
- image_response_px = requests.get(photo_url_px, timeout=60); image_response_px.raise_for_status()
283
- img_pil_data_px = Image.open(io.BytesIO(image_response_px.content))
284
- if img_pil_data_px.mode != 'RGB': img_pil_data_px = img_pil_data_px.convert('RGB')
285
- img_pil_data_px.save(file_path_px); logger.info(f"Pexels: Image saved to {file_path_px}"); return file_path_px
286
- else: logger.info(f"Pexels: No photos for '{eff_query_px}'."); return None
287
- except requests.exceptions.RequestException as e_req_px: logger.error(f"Pexels: RequestException for '{query_str}': {e_req_px}", exc_info=False); return None
288
- except Exception as e_px_gen: logger.error(f"Pexels: General error for '{query_str}': {e_px_gen}", exc_info=True); return None
289
 
290
- def _generate_video_clip_with_runwayml(self, text_prompt_for_motion, input_image_path, scene_identifier_filename_base, target_duration_seconds=5):
291
- # (Updated RunwayML integration from before)
292
- 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
293
- 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
294
- image_data_uri_str = self._image_to_data_uri(input_image_path) # Renamed
295
- if not image_data_uri_str: return None
296
- runway_dur = 10 if target_duration_seconds >= 8 else 5
297
- runway_ratio = self._map_resolution_to_runway_ratio(self.video_frame_size[0], self.video_frame_size[1])
298
- 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
299
- output_vid_fp = os.path.join(self.output_dir, output_vid_fn) # Renamed
300
- 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}'")
301
  try:
302
- 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
303
- task_id_runway = task_submitted_runway.id; logger.info(f"Runway Gen-4 task ID: {task_id_runway}. Polling...") # Renamed
304
- poll_sec=10; max_poll_count=36; poll_start_time = time.time() # Renamed
305
- while time.time() - poll_start_time < max_poll_count * poll_sec:
306
- time.sleep(poll_sec); task_details_runway = self.runway_ml_sdk_client_instance.tasks.retrieve(id=task_id_runway) # Renamed
307
- logger.info(f"Runway task {task_id_runway} status: {task_details_runway.status}")
308
- if task_details_runway.status == 'SUCCEEDED':
309
- 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
310
- 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
311
- logger.info(f"Runway task {task_id_runway} SUCCEEDED. Downloading: {output_url_runway}")
312
- video_resp_get = requests.get(output_url_runway, stream=True, timeout=300); video_resp_get.raise_for_status() # Renamed
313
- with open(output_vid_fp,'wb') as f_vid: # Renamed
314
- for chunk_data in video_resp_get.iter_content(chunk_size=8192): f_vid.write(chunk_data) # Renamed
315
- logger.info(f"Runway Gen-4 video saved: {output_vid_fp}"); return output_vid_fp
316
- elif task_details_runway.status in ['FAILED','ABORTED','ERROR']:
317
- err_msg_runway = getattr(task_details_runway,'error_message',None) or getattr(getattr(task_details_runway,'output',None),'error',"Unknown Runway error.") # Renamed
318
- logger.error(f"Runway task {task_id_runway} status: {task_details_runway.status}. Error: {err_msg_runway}"); return None
319
- logger.warning(f"Runway task {task_id_runway} timed out."); return None
320
- except AttributeError as ae_sdk: logger.error(f"RunwayML SDK AttrError: {ae_sdk}. SDK/methods changed?", exc_info=True); return None # Renamed
321
- except Exception as e_runway_gen: logger.error(f"Runway Gen-4 API error: {e_runway_gen}", exc_info=True); return None # Renamed
322
-
323
- def _create_placeholder_video_content(self, text_desc_ph, filename_ph, duration_ph=4, size_ph=None): # Renamed variables
324
- # <<< THIS IS THE CORRECTED METHOD >>>
325
- if size_ph is None: size_ph = self.video_frame_size
326
- filepath_ph = os.path.join(self.output_dir, filename_ph)
327
- text_clip_ph = None # Initialize
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
328
  try:
329
- text_clip_ph = TextClip(text_desc_ph, fontsize=50, color='white', font=self.video_overlay_font,
330
- bg_color='black', size=size_ph, method='caption').set_duration(duration_ph)
331
- text_clip_ph.write_videofile(filepath_ph, fps=24, codec='libx264', preset='ultrafast', logger=None, threads=2)
332
- logger.info(f"Generic placeholder video created: {filepath_ph}")
333
- return filepath_ph
334
- except Exception as e_ph_vid: # Specific exception variable
335
- logger.error(f"Failed to create generic placeholder video '{filepath_ph}': {e_ph_vid}", exc_info=True)
 
 
 
 
 
 
 
 
 
 
336
  return None
337
  finally:
338
- if text_clip_ph and hasattr(text_clip_ph, 'close'):
339
- text_clip_ph.close()
340
-
341
- # --- generate_scene_asset (Main asset generation logic) ---
342
- def generate_scene_asset(self, image_generation_prompt_text, motion_prompt_text_for_video,
343
- scene_data_dict, scene_identifier_fn_base, # Renamed
344
- generate_as_video_clip_flag=False, runway_target_dur_val=5): # Renamed
345
- # (Logic mostly as before, ensuring base image is robustly generated first)
346
- base_name_asset, _ = os.path.splitext(scene_identifier_fn_base) # Renamed
347
- asset_info_result = {'path': None, 'type': 'none', 'error': True, 'prompt_used': image_generation_prompt_text, 'error_message': 'Asset generation init failed'} # Renamed
348
- path_for_input_image_runway = None # Renamed
349
- fn_for_base_image = base_name_asset + ("_base_for_video.png" if generate_as_video_clip_flag else ".png") # Renamed
350
- fp_for_base_image = os.path.join(self.output_dir, fn_for_base_image) # Renamed
351
-
352
- if self.USE_AI_IMAGE_GENERATION and self.openai_api_key:
353
- # (DALL-E logic with corrected try/except from previous response)
354
- max_r_dalle, att_n_dalle = 2,0;
355
- for att_n_dalle in range(max_r_dalle):
356
- att_c_dalle = att_n_dalle + 1 # Renamed
357
- try:
358
- logger.info(f"Att {att_c_dalle} DALL-E (base img): {image_generation_prompt_text[:70]}...");
359
- oai_cl = openai.OpenAI(api_key=self.openai_api_key,timeout=90.0) # Renamed
360
- 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
361
- oai_iu = oai_r.data[0].url; oai_rp = getattr(oai_r.data[0],'revised_prompt',None) # Renamed
362
- if oai_rp: logger.info(f"DALL-E revised: {oai_rp[:70]}...")
363
- oai_ir = requests.get(oai_iu,timeout=120); oai_ir.raise_for_status() # Renamed
364
- oai_id = Image.open(io.BytesIO(oai_ir.content)); # Renamed
365
- if oai_id.mode!='RGB': oai_id=oai_id.convert('RGB')
366
- oai_id.save(fp_for_base_image); logger.info(f"DALL-E base img saved: {fp_for_base_image}")
367
- path_for_input_image_runway=fp_for_base_image
368
- asset_info_result={'path':fp_for_base_image,'type':'image','error':False,'prompt_used':image_generation_prompt_text,'revised_prompt':oai_rp}
369
- break
370
- 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
371
- 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
372
- 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
373
- 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
374
- if asset_info_result['error']: logger.warning(f"DALL-E failed after {att_c_dalle} attempts for base img.")
375
-
376
- if asset_info_result['error'] and self.USE_PEXELS:
377
- 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
378
- 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}"}
379
- else:current_em_px=asset_info_result.get('error_message',"");asset_info_result['error_message']=(current_em_px+" Pexels failed for base.").strip() # Renamed
380
 
381
- if asset_info_result['error']:
382
- 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
383
- if ph_p:path_for_input_image_runway=ph_p;asset_info_result={'path':ph_p,'type':'image','error':False,'prompt_used':ph_ppt}
384
- else:current_em_ph=asset_info_result.get('error_message',"");asset_info_result['error_message']=(current_em_ph+" Base placeholder failed.").strip() # Renamed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
385
 
386
- if generate_as_video_clip_flag:
387
- 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
388
- if self.USE_RUNWAYML:
389
- 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
390
- 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}
391
- 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
392
- 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
393
- return asset_info_result
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
394
 
395
- def generate_narration_audio(self, text_to_narrate, output_filename="narration_overall.mp3"):
396
- # (Keep as before)
397
- if not self.USE_ELEVENLABS or not self.elevenlabs_client_instance or not text_to_narrate: logger.info("11L skip."); return None # Check instance
398
- audio_fp_11l=os.path.join(self.output_dir,output_filename) # Renamed
399
- try: logger.info(f"11L audio (Voice:{self.elevenlabs_voice_id}): {text_to_narrate[:70]}..."); audio_sm_11l=None # Renamed
400
- 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()")
401
- elif hasattr(self.elevenlabs_client_instance,'generate_stream'):audio_sm_11l=self.elevenlabs_client_instance.generate_stream;logger.info("Using 11L .generate_stream()")
402
- 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
403
- 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
404
- else:logger.error("No 11L audio method.");return None
405
- if audio_sm_11l:eleven_vps={"voice_id":str(self.elevenlabs_voice_id)} # Renamed
406
- if self.elevenlabs_voice_settings_obj:
407
- if hasattr(self.elevenlabs_voice_settings_obj,'model_dump'):eleven_vps["voice_settings"]=self.elevenlabs_voice_settings_obj.model_dump()
408
- elif hasattr(self.elevenlabs_voice_settings_obj,'dict'):eleven_vps["voice_settings"]=self.elevenlabs_voice_settings_obj.dict()
409
- else:eleven_vps["voice_settings"]=self.elevenlabs_voice_settings_obj
410
- eleven_adi=audio_sm_11l(text=text_to_narrate,model_id="eleven_multilingual_v2",**eleven_vps) # Renamed
411
- with open(audio_fp_11l,"wb")as f_11l_stream: # Renamed
412
- for chunk_11l in eleven_adi: # Renamed
413
- if chunk_11l:f_11l_stream.write(chunk_11l)
414
- logger.info(f"11L audio (stream): {audio_fp_11l}");return audio_fp_11l
415
- except Exception as e_11labs_audio:logger.error(f"11L audio error: {e_11labs_audio}",exc_info=True);return None # Renamed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
416
 
417
  def assemble_animatic_from_assets(self, asset_data_list, overall_narration_path=None, output_filename="final_video.mp4", fps=24):
418
- # (Keep as in the version with robust image processing, C-contiguous array, debug saves, and pix_fmt)
419
- # ... (This extensive method is assumed to be largely correct from the previous iteration focusing on blank video issues)
420
- # ... (The core of that robust version should be retained here) ...
421
- # For brevity, I'm not re-pasting the entire assemble_animatic_from_assets if it was correct before this syntax error hunt.
422
- # If it also needs review, let me know. The key is that the *input* (asset_data_list)
423
- # from the corrected generate_scene_asset is now more reliable.
424
- if not asset_data_list: logger.warning("No assets for animatic."); return None
425
- processed_moviepy_clips_list = []; narration_audio_clip_mvpy = None; final_video_output_clip = None # Renamed variables
426
- logger.info(f"Assembling from {len(asset_data_list)} assets. Target Frame: {self.video_frame_size}.")
427
-
428
- for i_asset, asset_info_item_loop in enumerate(asset_data_list): # Renamed loop variables
429
- 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)
430
- num_of_scene, action_in_key = asset_info_item_loop.get('scene_num', i_asset + 1), asset_info_item_loop.get('key_action', '')
431
- logger.info(f"S{num_of_scene}: Path='{path_of_asset}', Type='{type_of_asset}', Dur='{duration_for_scene}'s")
432
-
433
- 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
434
- if duration_for_scene <= 0: logger.warning(f"S{num_of_scene}: Invalid duration ({duration_for_scene}s). Skip."); continue
435
-
436
- active_scene_clip = None # Clip for this iteration
 
 
 
 
 
437
  try:
438
- if type_of_asset == 'image':
439
- # (Robust image processing from previous version, including debug saves)
440
- 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}")
441
- converted_img_rgba = opened_pil_img.convert('RGBA') if opened_pil_img.mode != 'RGBA' else opened_pil_img.copy()
442
- 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)
443
- 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
444
- rgba_canvas.paste(thumbnailed_img,(pos_x,pos_y),thumbnailed_img)
445
- 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])
446
- 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}")
447
- numpy_frame_arr = np.array(final_rgb_img_pil,dtype=np.uint8);
448
- if not numpy_frame_arr.flags['C_CONTIGUOUS']: numpy_frame_arr=np.ascontiguousarray(numpy_frame_arr,dtype=np.uint8)
449
- 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']}")
450
- 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
451
- base_image_clip = ImageClip(numpy_frame_arr,transparent=False).set_duration(duration_for_scene)
452
- 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}")
453
- fx_image_clip = base_image_clip
454
- 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')
455
- except Exception as e_kb_fx: logger.error(f"S{num_of_scene} Ken Burns error: {e_kb_fx}",exc_info=False)
456
- active_scene_clip = fx_image_clip
457
- elif type_of_asset == 'video':
458
- # (Video processing logic from previous version)
459
- source_video_clip_obj=None
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
460
  try:
461
- 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)
462
- temp_video_clip_obj_loop=source_video_clip_obj
463
- if source_video_clip_obj.duration!=duration_for_scene:
464
- if source_video_clip_obj.duration>duration_for_scene:temp_video_clip_obj_loop=source_video_clip_obj.subclip(0,duration_for_scene)
 
 
 
 
 
465
  else:
466
- 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)
467
- 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).")
468
- active_scene_clip=temp_video_clip_obj_loop.set_duration(duration_for_scene)
469
- if active_scene_clip.size!=list(self.video_frame_size):active_scene_clip=active_scene_clip.resize(self.video_frame_size)
470
- 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
 
 
 
 
 
 
471
  finally:
472
- 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()
473
- else: logger.warning(f"S{num_of_scene} Unknown asset type '{type_of_asset}'. Skip."); continue
474
-
475
- if active_scene_clip and action_in_key: # Text Overlay
 
 
 
476
  try:
477
- 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
478
- start_text_overlay=0.25
479
- if dur_text_overlay > 0:
480
- 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)
481
- active_scene_clip=CompositeVideoClip([active_scene_clip,text_clip_for_overlay],size=self.video_frame_size,use_bgclip=True)
482
- else: logger.warning(f"S{num_of_scene}: Text overlay duration zero. Skip text.")
483
- except Exception as e_txt_comp:logger.error(f"S{num_of_scene} TextClip error:{e_txt_comp}. No text.",exc_info=True)
484
- 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.")
485
- 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)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
486
  finally:
487
- if active_scene_clip and hasattr(active_scene_clip,'close'):
488
- try: active_scene_clip.close()
489
- except: pass # Ignore errors during cleanup
 
 
490
 
491
- if not processed_moviepy_clips_list:logger.warning("No clips processed for animatic. Aborting.");return None
492
- transition_duration_val=0.75 # Renamed
 
 
493
  try:
494
- logger.info(f"Concatenating {len(processed_moviepy_clips_list)} clips for final animatic.");
495
- 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")
496
- elif processed_moviepy_clips_list:final_video_output_clip=processed_moviepy_clips_list[0]
497
- if not final_video_output_clip:logger.error("Concatenation resulted in a None clip. Aborting.");return None
498
- logger.info(f"Concatenated animatic duration:{final_video_output_clip.duration:.2f}s")
499
- if transition_duration_val>0 and final_video_output_clip.duration>0:
500
- 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)
501
- else:final_video_output_clip=final_video_output_clip.fx(vfx.fadein,min(transition_duration_val,final_video_output_clip.duration/2.0))
502
- if overall_narration_path and os.path.exists(overall_narration_path) and final_video_output_clip.duration>0:
503
- 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.")
504
- except Exception as e_narr_add:logger.error(f"Error adding narration to animatic:{e_narr_add}",exc_info=True)
505
- elif final_video_output_clip.duration<=0:logger.warning("Animatic has no duration. Audio not added.")
506
- if final_video_output_clip and final_video_output_clip.duration>0:
507
- 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
508
- 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"])
509
- logger.info(f"Animatic video created successfully:{final_output_path_str}");return final_output_path_str
510
- else:logger.error("Final animatic clip is invalid or has zero duration. Cannot write video file.");return None
511
- 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
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
512
  finally:
513
- logger.debug("Closing all MoviePy clips in `assemble_animatic_from_assets` main finally block.")
514
- 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
515
- for clip_item_to_close in clips_for_final_closure: # Renamed
516
- if clip_item_to_close and hasattr(clip_item_to_close, 'close'):
517
- try: clip_item_to_close.close()
518
- 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}")
 
 
 
1
  # core/visual_engine.py
2
  from PIL import Image, ImageDraw, ImageFont, ImageOps
3
+ # --- MONKEY PATCH FOR Image.ANTIALIAS ---
4
+ try:
5
+ if hasattr(Image, 'Resampling') and hasattr(Image.Resampling, 'LANCZOS'): # Pillow 9+
6
+ if not hasattr(Image, 'ANTIALIAS'):
7
+ Image.ANTIALIAS = Image.Resampling.LANCZOS
8
+ elif hasattr(Image, 'LANCZOS'): # Pillow 8
9
+ if not hasattr(Image, 'ANTIALIAS'):
10
+ Image.ANTIALIAS = Image.LANCZOS
11
+ elif not hasattr(Image, 'ANTIALIAS'):
12
+ print("WARNING: Pillow version lacks common Resampling attributes or ANTIALIAS. Video effects might fail.")
13
+ except Exception as e_mp:
14
+ print(f"WARNING: ANTIALIAS monkey-patch error: {e_mp}")
15
+ # --- END MONKEY PATCH ---
16
+
17
+ from moviepy.editor import (ImageClip, VideoFileClip, concatenate_videoclips, TextClip,
18
+ CompositeVideoClip, AudioFileClip)
19
+ import moviepy.video.fx.all as vfx
20
  import numpy as np
21
  import os
22
+ import openai
23
  import requests
24
  import io
25
  import time
26
  import random
27
  import logging
28
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
29
  logger = logging.getLogger(__name__)
30
+ logger.setLevel(logging.INFO)
 
31
 
32
+ # --- ElevenLabs Client Import ---
33
  ELEVENLABS_CLIENT_IMPORTED = False
34
+ ElevenLabsAPIClient = None
35
+ Voice = None
36
+ VoiceSettings = None
37
  try:
38
  from elevenlabs.client import ElevenLabs as ImportedElevenLabsClient
39
  from elevenlabs import Voice as ImportedVoice, VoiceSettings as ImportedVoiceSettings
 
41
  Voice = ImportedVoice
42
  VoiceSettings = ImportedVoiceSettings
43
  ELEVENLABS_CLIENT_IMPORTED = True
44
+ logger.info("ElevenLabs client components imported.")
45
+ except Exception as e_eleven:
46
+ logger.warning(f"ElevenLabs client import failed: {e_eleven}. Audio disabled.")
 
 
47
 
48
+ # --- RunwayML Client Import (Placeholder) ---
49
  RUNWAYML_SDK_IMPORTED = False
50
+ RunwayMLClient = None
51
  try:
52
+ logger.info("RunwayML SDK import is a placeholder.")
 
 
 
53
  except ImportError:
54
+ logger.warning("RunwayML SDK (placeholder) not found. RunwayML disabled.")
55
+ except Exception as e_runway_sdk:
56
+ logger.warning(f"Error importing RunwayML SDK (placeholder): {e_runway_sdk}. RunwayML disabled.")
57
 
58
 
59
  class VisualEngine:
 
 
 
 
 
 
 
60
  def __init__(self, output_dir="temp_cinegen_media", default_elevenlabs_voice_id="Rachel"):
61
  self.output_dir = output_dir
62
  os.makedirs(self.output_dir, exist_ok=True)
63
+ self.font_filename = "DejaVuSans-Bold.ttf"
 
64
  font_paths_to_try = [
65
+ self.font_filename,
66
+ f"/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf",
67
+ f"/usr/share/fonts/truetype/liberation/LiberationSans-Bold.ttf",
68
  f"/System/Library/Fonts/Supplemental/Arial.ttf",
69
  f"C:/Windows/Fonts/arial.ttf",
70
+ f"/usr/local/share/fonts/truetype/mycustomfonts/arial.ttf"
71
  ]
72
+ self.font_path_pil = next((p for p in font_paths_to_try if os.path.exists(p)), None)
73
+ self.font_size_pil = 20
74
+ self.video_overlay_font_size = 30
75
+ self.video_overlay_font_color = 'white'
76
+ self.video_overlay_font = 'DejaVu-Sans-Bold'
77
 
78
+ try:
79
+ if self.font_path_pil:
80
+ self.font = ImageFont.truetype(self.font_path_pil, self.font_size_pil)
81
+ logger.info(f"Pillow font loaded: {self.font_path_pil}.")
82
+ else:
83
+ self.font = ImageFont.load_default()
84
+ logger.warning("Using default Pillow font.")
85
+ self.font_size_pil = 10
86
+ except IOError as e_font:
87
+ logger.error(f"Pillow font loading IOError: {e_font}. Using default.")
88
+ self.font = ImageFont.load_default()
89
+ self.font_size_pil = 10
90
+
91
+ self.openai_api_key = None
92
+ self.USE_AI_IMAGE_GENERATION = False
93
+ self.dalle_model = "dall-e-3"
94
+ self.image_size_dalle3 = "1792x1024"
95
+ self.video_frame_size = (1280, 720)
96
+ self.elevenlabs_api_key = None
97
+ self.USE_ELEVENLABS = False
98
+ self.elevenlabs_client = None
99
  self.elevenlabs_voice_id = default_elevenlabs_voice_id
100
  if VoiceSettings and ELEVENLABS_CLIENT_IMPORTED:
101
+ self.elevenlabs_voice_settings = VoiceSettings(
102
+ stability=0.60,
103
+ similarity_boost=0.80,
104
+ style=0.15,
105
+ use_speaker_boost=True
106
+ )
107
+ else:
108
+ self.elevenlabs_voice_settings = None
109
+ self.pexels_api_key = None
110
+ self.USE_PEXELS = False
111
+ self.runway_api_key = None
112
+ self.USE_RUNWAYML = False
113
+ self.runway_client = None
 
 
 
114
  logger.info("VisualEngine initialized.")
115
 
116
+ def set_openai_api_key(self, k):
117
+ self.openai_api_key = k
118
+ self.USE_AI_IMAGE_GENERATION = bool(k)
119
+ logger.info(f"DALL-E ({self.dalle_model}) {'Ready.' if k else 'Disabled.'}")
120
 
121
+ def set_elevenlabs_api_key(self, api_key, voice_id_from_secret=None):
122
+ self.elevenlabs_api_key = api_key
123
+ if voice_id_from_secret:
124
+ self.elevenlabs_voice_id = voice_id_from_secret
125
+ if api_key and ELEVENLABS_CLIENT_IMPORTED and ElevenLabsAPIClient:
126
  try:
127
+ self.elevenlabs_client = ElevenLabsAPIClient(api_key=api_key)
128
+ self.USE_ELEVENLABS = bool(self.elevenlabs_client)
129
+ logger.info(f"ElevenLabs Client {'Ready' if self.USE_ELEVENLABS else 'Failed Init'} (Voice ID: {self.elevenlabs_voice_id}).")
130
+ except Exception as e:
131
+ logger.error(f"ElevenLabs client init error: {e}. Disabled.", exc_info=True)
132
+ self.USE_ELEVENLABS = False
133
  else:
134
  self.USE_ELEVENLABS = False
135
+ logger.info("ElevenLabs Disabled (no key or SDK).")
136
 
137
+ def set_pexels_api_key(self, k):
138
+ self.pexels_api_key = k
139
+ self.USE_PEXELS = bool(k)
140
+ logger.info(f"Pexels Search {'Ready.' if k else 'Disabled.'}")
141
 
142
+ def set_runway_api_key(self, k):
143
+ self.runway_api_key = k
144
+ if k and RUNWAYML_SDK_IMPORTED and RunwayMLClient:
145
+ try:
146
+ self.USE_RUNWAYML = True
147
+ logger.info(f"RunwayML Client (Placeholder SDK) {'Ready.' if self.USE_RUNWAYML else 'Failed Init.'}")
148
+ except Exception as e:
149
+ logger.error(f"RunwayML client (Placeholder SDK) init error: {e}. Disabled.", exc_info=True)
150
+ self.USE_RUNWAYML = False
151
+ elif k:
152
+ self.USE_RUNWAYML = True
153
+ logger.info("RunwayML API Key set (direct API or placeholder).")
154
+ else:
155
+ self.USE_RUNWAYML = False
156
+ logger.info("RunwayML Disabled (no API key).")
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
157
 
158
+ def _get_text_dimensions(self, text_content, font_obj):
159
+ default_line_height = getattr(font_obj, 'size', self.font_size_pil)
160
+ if not text_content:
161
+ return 0, default_line_height
162
  try:
163
+ if hasattr(font_obj, 'getbbox'):
164
+ bbox = font_obj.getbbox(text_content)
165
+ width = bbox[2] - bbox[0]
166
+ height = bbox[3] - bbox[1]
167
+ return width, height if height > 0 else default_line_height
168
+ elif hasattr(font_obj, 'getsize'):
169
+ width, height = font_obj.getsize(text_content)
170
+ return width, height if height > 0 else default_line_height
171
+ else:
172
+ return int(len(text_content) * default_line_height * 0.6), int(default_line_height * 1.2)
173
+ except Exception as e:
174
+ logger.warning(f"Error in _get_text_dimensions for '{text_content[:20]}...': {e}")
175
+ return int(len(text_content) * self.font_size_pil * 0.6), int(self.font_size_pil * 1.2)
176
 
177
  def _create_placeholder_image_content(self, text_description, filename, size=None):
178
+ if size is None:
179
+ size = self.video_frame_size
180
+ img = Image.new('RGB', size, color=(20, 20, 40))
181
+ draw = ImageDraw.Draw(img)
182
+ padding = 25
183
+ max_text_width = size[0] - (2 * padding)
184
+ lines = []
185
+ if not text_description:
186
+ text_description = "(Placeholder: No text description provided)"
187
+ words = text_description.split()
188
+ current_line = ""
189
+ for word in words:
190
+ test_line = current_line + word + " "
191
+ line_width_test, _ = self._get_text_dimensions(test_line.strip(), self.font)
192
+ if line_width_test <= max_text_width:
193
+ current_line = test_line
194
  else:
195
+ if current_line.strip():
196
+ lines.append(current_line.strip())
197
+ word_width, _ = self._get_text_dimensions(word, self.font)
198
+ if word_width > max_text_width:
199
+ avg_char_w = self._get_text_dimensions("A", self.font)[0] or 10
200
+ chars_that_fit = int(max_text_width / avg_char_w) if avg_char_w > 0 else 10
201
+ if len(word) > chars_that_fit:
202
+ lines.append(word[:chars_that_fit-3] + "...")
203
+ else:
204
+ lines.append(word)
205
+ current_line = ""
206
+ else:
207
+ current_line = word + " "
208
+ if current_line.strip():
209
+ lines.append(current_line.strip())
210
+ if not lines and text_description:
211
+ avg_char_w = self._get_text_dimensions("A", self.font)[0] or 10
212
+ chars_that_fit = int(max_text_width / avg_char_w) if avg_char_w > 0 else 10
213
+ if len(text_description) > chars_that_fit:
214
+ lines.append(text_description[:chars_that_fit-3] + "...")
215
+ else:
216
+ lines.append(text_description)
217
+ elif not lines:
218
+ lines.append("(Placeholder Text Error)")
219
+ _, single_line_height = self._get_text_dimensions("Ay", self.font)
220
+ single_line_height = single_line_height if single_line_height > 0 else (self.font_size_pil + 2)
221
+ line_spacing = 2
222
+ max_lines_to_display = min(len(lines), (size[1] - (2 * padding)) // (single_line_height + line_spacing)) if single_line_height > 0 else 1
223
+ if max_lines_to_display <= 0:
224
+ max_lines_to_display = 1
225
+ total_text_block_height = max_lines_to_display * single_line_height + (max_lines_to_display - 1) * line_spacing
226
+ y_text_start = padding + (size[1] - (2 * padding) - total_text_block_height) / 2.0
227
+ current_y = y_text_start
228
+ for i in range(max_lines_to_display):
229
+ line_content = lines[i]
230
+ line_width_actual, _ = self._get_text_dimensions(line_content, self.font)
231
+ x_text = max(padding, (size[0] - line_width_actual) / 2.0)
232
+ draw.text((x_text, current_y), line_content, font=self.font, fill=(200, 200, 180))
233
+ current_y += single_line_height + line_spacing
234
+ if i == 6 and max_lines_to_display > 7 and len(lines) > max_lines_to_display:
235
+ ellipsis_width, _ = self._get_text_dimensions("...", self.font)
236
+ x_ellipsis = max(padding, (size[0] - ellipsis_width) / 2.0)
237
+ draw.text((x_ellipsis, current_y), "...", font=self.font, fill=(200, 200, 180))
238
+ break
239
+ filepath = os.path.join(self.output_dir, filename)
240
  try:
241
+ img.save(filepath)
242
+ return filepath
243
+ except Exception as e:
244
+ logger.error(f"Error saving placeholder image {filepath}: {e}", exc_info=True)
245
+ return None
 
 
 
 
 
 
 
 
 
 
 
 
246
 
247
+ def _search_pexels_image(self, query, output_filename_base):
248
+ if not self.USE_PEXELS or not self.pexels_api_key:
249
+ return None
250
+ headers = {"Authorization": self.pexels_api_key}
251
+ params = {"query": query, "per_page": 1, "orientation": "landscape", "size": "large2x"}
252
+ base_name, _ = os.path.splitext(output_filename_base)
253
+ pexels_filename = base_name + f"_pexels_{random.randint(1000,9999)}.jpg"
254
+ filepath = os.path.join(self.output_dir, pexels_filename)
 
 
 
255
  try:
256
+ logger.info(f"Pexels search: '{query}'")
257
+ effective_query = " ".join(query.split()[:5])
258
+ params["query"] = effective_query
259
+ response = requests.get("https://api.pexels.com/v1/search", headers=headers, params=params, timeout=20)
260
+ response.raise_for_status()
261
+ data = response.json()
262
+ if data.get("photos") and len(data["photos"]) > 0:
263
+ photo_details = data["photos"][0]
264
+ photo_url = photo_details["src"]["large2x"]
265
+ logger.info(f"Downloading Pexels image from: {photo_url}")
266
+ image_response = requests.get(photo_url, timeout=60)
267
+ image_response.raise_for_status()
268
+ img_data = Image.open(io.BytesIO(image_response.content))
269
+ if img_data.mode != 'RGB':
270
+ logger.debug(f"Pexels image mode is {img_data.mode}, converting to RGB.")
271
+ img_data = img_data.convert('RGB')
272
+ img_data.save(filepath)
273
+ logger.info(f"Pexels image saved successfully: {filepath}")
274
+ return filepath
275
+ else:
276
+ logger.info(f"No photos found on Pexels for query: '{effective_query}'")
277
+ return None
278
+ except requests.exceptions.RequestException as e_req:
279
+ logger.error(f"Pexels request error for query '{query}': {e_req}", exc_info=True)
280
+ except json.JSONDecodeError as e_json:
281
+ logger.error(f"Pexels JSON decode error for query '{query}': {e_json}", exc_info=True)
282
+ except Exception as e:
283
+ logger.error(f"General Pexels error for query '{query}': {e}", exc_info=True)
284
+ return None
285
+
286
+ def _generate_video_clip_with_runwayml(self, pt, iip, sifnb, tds=5):
287
+ if not self.USE_RUNWAYML or not self.runway_api_key:
288
+ logger.warning("RunwayML disabled.")
289
+ return None
290
+ if not iip or not os.path.exists(iip):
291
+ logger.error(f"Runway Gen-4 needs input image. Path invalid: {iip}")
292
+ return None
293
+ runway_dur = 10 if tds > 7 else 5
294
+ ovfn = sifnb.replace(".png", f"_runway_gen4_d{runway_dur}s.mp4")
295
+ ovfp = os.path.join(self.output_dir, ovfn)
296
+ logger.info(f"Runway Gen-4 (Placeholder) img: {os.path.basename(iip)}, motion: '{pt[:100]}...', dur: {runway_dur}s")
297
+ logger.warning("Using PLACEHOLDER video for Runway Gen-4.")
298
+ img_clip = None
299
+ txt_c = None
300
+ final_ph_clip = None
301
  try:
302
+ img_clip = ImageClip(iip).set_duration(runway_dur)
303
+ txt = f"Runway Gen-4 Placeholder\nInput: {os.path.basename(iip)}\nMotion: {pt[:50]}..."
304
+ txt_c = TextClip(
305
+ txt,
306
+ fontsize=24,
307
+ color='white',
308
+ font=self.video_overlay_font,
309
+ bg_color='rgba(0,0,0,0.5)',
310
+ size=(self.video_frame_size[0] * 0.8, None),
311
+ method='caption'
312
+ ).set_duration(runway_dur).set_position('center')
313
+ final_ph_clip = CompositeVideoClip([img_clip, txt_c], size=img_clip.size)
314
+ final_ph_clip.write_videofile(ovfp, fps=24, codec='libx264', preset='ultrafast', logger=None, threads=2)
315
+ logger.info(f"Runway Gen-4 placeholder video: {ovfp}")
316
+ return ovfp
317
+ except Exception as e:
318
+ logger.error(f"Runway Gen-4 placeholder error: {e}", exc_info=True)
319
  return None
320
  finally:
321
+ if img_clip and hasattr(img_clip, 'close'):
322
+ img_clip.close()
323
+ if txt_c and hasattr(txt_c, 'close'):
324
+ txt_c.close()
325
+ if final_ph_clip and hasattr(final_ph_clip, 'close'):
326
+ final_ph_clip.close()
327
+
328
+ def _create_placeholder_video_content(self, text_description, filename, duration=4, size=None):
329
+ if size is None:
330
+ size = self.video_frame_size
331
+ filepath = os.path.join(self.output_dir, filename)
332
+ txt_clip = None
333
+ try:
334
+ txt_clip = TextClip(
335
+ text_description,
336
+ fontsize=50,
337
+ color='white',
338
+ font=self.video_overlay_font,
339
+ bg_color='black',
340
+ size=size,
341
+ method='caption'
342
+ ).set_duration(duration)
343
+ txt_clip.write_videofile(
344
+ filepath,
345
+ fps=24,
346
+ codec='libx264',
347
+ preset='ultrafast',
348
+ logger=None,
349
+ threads=2
350
+ )
351
+ logger.info(f"Generic placeholder video created successfully: {filepath}")
352
+ return filepath
353
+ except Exception as e:
354
+ logger.error(f"Failed to create generic placeholder video {filepath}: {e}", exc_info=True)
355
+ return None
356
+ finally:
357
+ if txt_clip and hasattr(txt_clip, 'close'):
358
+ try:
359
+ txt_clip.close()
360
+ except Exception as e_close:
361
+ logger.warning(f"Error closing TextClip in _create_placeholder_video_content: {e_close}")
 
362
 
363
+ def generate_scene_asset(self, image_generation_prompt_text, motion_prompt_text_for_video,
364
+ scene_data, scene_identifier_filename_base,
365
+ generate_as_video_clip=False, runway_target_duration=5):
366
+ base_name = scene_identifier_filename_base
367
+ asset_info = {
368
+ 'path': None,
369
+ 'type': 'none',
370
+ 'error': True,
371
+ 'prompt_used': image_generation_prompt_text,
372
+ 'error_message': 'Generation not attempted'
373
+ }
374
+ input_image_for_runway_path = None
375
+ image_filename_for_base = base_name + "_base_image.png"
376
+ temp_image_asset_info = {
377
+ 'error': True,
378
+ 'prompt_used': image_generation_prompt_text,
379
+ 'error_message': 'Base image generation not attempted'
380
+ }
381
 
382
+ if self.USE_AI_IMAGE_GENERATION and self.openai_api_key:
383
+ max_r, att_n = 2, 0
384
+ for att_n in range(max_r):
385
+ try:
386
+ img_fp_dalle = os.path.join(self.output_dir, image_filename_for_base)
387
+ logger.info(f"Attempt {att_n+1} DALL-E (base img): {image_generation_prompt_text[:100]}...")
388
+ cl = openai.OpenAI(api_key=self.openai_api_key, timeout=90.0)
389
+ r = cl.images.generate(
390
+ model=self.dalle_model,
391
+ prompt=image_generation_prompt_text,
392
+ n=1,
393
+ size=self.image_size_dalle3,
394
+ quality="hd",
395
+ response_format="url",
396
+ style="vivid"
397
+ )
398
+ iu = r.data[0].url
399
+ rp = getattr(r.data[0], 'revised_prompt', None)
400
+ if rp:
401
+ logger.info(f"DALL-E revised: {rp[:100]}...")
402
+ ir = requests.get(iu, timeout=120)
403
+ ir.raise_for_status()
404
+ id_img = Image.open(io.BytesIO(ir.content))
405
+ if id_img.mode != 'RGB':
406
+ id_img = id_img.convert('RGB')
407
+ id_img.save(img_fp_dalle)
408
+ logger.info(f"DALL-E base image: {img_fp_dalle}")
409
+ input_image_for_runway_path = img_fp_dalle
410
+ temp_image_asset_info = {
411
+ 'path': img_fp_dalle,
412
+ 'type': 'image',
413
+ 'error': False,
414
+ 'prompt_used': image_generation_prompt_text,
415
+ 'revised_prompt': rp
416
+ }
417
+ break
418
+ except openai.RateLimitError as e:
419
+ logger.warning(f"OpenAI Rate Limit {att_n+1}: {e}. Retry...")
420
+ time.sleep(5 * (att_n + 1))
421
+ temp_image_asset_info['error_message'] = str(e)
422
+ except Exception as e:
423
+ logger.error(f"DALL-E error: {e}", exc_info=True)
424
+ temp_image_asset_info['error_message'] = str(e)
425
+ break
426
+ if temp_image_asset_info['error']:
427
+ logger.warning(f"DALL-E failed after {att_n+1} attempts for base image.")
428
+
429
+ if temp_image_asset_info['error'] and self.USE_PEXELS:
430
+ pqt = scene_data.get('pexels_search_query_감독',
431
+ f"{scene_data.get('emotional_beat','')} {scene_data.get('setting_description','')}")
432
+ pp = self._search_pexels_image(pqt, image_filename_for_base)
433
+ if pp:
434
+ input_image_for_runway_path = pp
435
+ temp_image_asset_info = {
436
+ 'path': pp,
437
+ 'type': 'image',
438
+ 'error': False,
439
+ 'prompt_used': f"Pexels: {pqt}"
440
+ }
441
+ else:
442
+ current_em = temp_image_asset_info.get('error_message', "")
443
+ temp_image_asset_info['error_message'] = (current_em + " Pexels failed.").strip()
444
+
445
+ if temp_image_asset_info['error']:
446
+ logger.warning("Base image (DALL-E/Pexels) failed. Placeholder base image.")
447
+ ppt = temp_image_asset_info.get('prompt_used', image_generation_prompt_text)
448
+ php = self._create_placeholder_image_content(f"[Base Img Placeholder] {ppt[:100]}...", image_filename_for_base)
449
+ if php:
450
+ input_image_for_runway_path = php
451
+ temp_image_asset_info = {
452
+ 'path': php,
453
+ 'type': 'image',
454
+ 'error': False,
455
+ 'prompt_used': ppt
456
+ }
457
+ else:
458
+ current_em = temp_image_asset_info.get('error_message', "")
459
+ temp_image_asset_info['error_message'] = (current_em + " Base placeholder failed.").strip()
460
+
461
+ if generate_as_video_clip:
462
+ if self.USE_RUNWAYML and input_image_for_runway_path:
463
+ video_path = self._generate_video_clip_with_runwayml(
464
+ motion_prompt_text_for_video,
465
+ input_image_for_runway_path,
466
+ base_name,
467
+ runway_target_duration
468
+ )
469
+ if video_path and os.path.exists(video_path):
470
+ return {
471
+ 'path': video_path,
472
+ 'type': 'video',
473
+ 'error': False,
474
+ 'prompt_used': motion_prompt_text_for_video,
475
+ 'base_image_path': input_image_for_runway_path
476
+ }
477
+ else:
478
+ asset_info = temp_image_asset_info
479
+ asset_info['error'] = True
480
+ asset_info['error_message'] = "RunwayML video gen failed; using base image."
481
+ asset_info['type'] = 'image'
482
+ return asset_info
483
+ elif not self.USE_RUNWAYML:
484
+ asset_info = temp_image_asset_info
485
+ asset_info['error_message'] = "RunwayML disabled; using base image."
486
+ asset_info['type'] = 'image'
487
+ return asset_info
488
+ else:
489
+ asset_info = temp_image_asset_info
490
+ asset_info['error_message'] = (asset_info.get('error_message', "") + " Base image failed, Runway video not attempted.").strip()
491
+ asset_info['type'] = 'image'
492
+ return asset_info
493
+ else:
494
+ return temp_image_asset_info
495
 
496
+ def generate_narration_audio(self, ttn, ofn="narration_overall.mp3"):
497
+ if not self.USE_ELEVENLABS or not self.elevenlabs_client or not ttn:
498
+ logger.info("11L skip.")
499
+ return None
500
+ afp = os.path.join(self.output_dir, ofn)
501
+ try:
502
+ logger.info(f"11L audio (Voice:{self.elevenlabs_voice_id}): {ttn[:70]}...")
503
+ asm = None
504
+ if hasattr(self.elevenlabs_client, 'text_to_speech') and hasattr(self.elevenlabs_client.text_to_speech, 'stream'):
505
+ asm = self.elevenlabs_client.text_to_speech.stream
506
+ logger.info("Using 11L .text_to_speech.stream()")
507
+ elif hasattr(self.elevenlabs_client, 'generate_stream'):
508
+ asm = self.elevenlabs_client.generate_stream
509
+ logger.info("Using 11L .generate_stream()")
510
+ elif hasattr(self.elevenlabs_client, 'generate'):
511
+ logger.info("Using 11L .generate()")
512
+ vp = Voice(voice_id=str(self.elevenlabs_voice_id),
513
+ settings=self.elevenlabs_voice_settings) if Voice and self.elevenlabs_voice_settings else str(self.elevenlabs_voice_id)
514
+ ab = self.elevenlabs_client.generate(text=ttn, voice=vp, model="eleven_multilingual_v2")
515
+ with open(afp, "wb") as f:
516
+ f.write(ab)
517
+ logger.info(f"11L audio (non-stream): {afp}")
518
+ return afp
519
+ else:
520
+ logger.error("No 11L audio method.")
521
+ return None
522
+
523
+ if asm:
524
+ vps = {"voice_id": str(self.elevenlabs_voice_id)}
525
+ if self.elevenlabs_voice_settings:
526
+ if hasattr(self.elevenlabs_voice_settings, 'model_dump'):
527
+ vps["voice_settings"] = self.elevenlabs_voice_settings.model_dump()
528
+ elif hasattr(self.elevenlabs_voice_settings, 'dict'):
529
+ vps["voice_settings"] = self.elevenlabs_voice_settings.dict()
530
+ else:
531
+ vps["voice_settings"] = self.elevenlabs_voice_settings
532
+ adi = asm(text=ttn, model_id="eleven_multilingual_v2", **vps)
533
+ with open(afp, "wb") as f:
534
+ for c in adi:
535
+ if c:
536
+ f.write(c)
537
+ logger.info(f"11L audio (stream): {afp}")
538
+ return afp
539
+ except Exception as e:
540
+ logger.error(f"11L audio error: {e}", exc_info=True)
541
+ return None
542
 
543
  def assemble_animatic_from_assets(self, asset_data_list, overall_narration_path=None, output_filename="final_video.mp4", fps=24):
544
+ if not asset_data_list:
545
+ logger.warning("No assets for animatic.")
546
+ return None
547
+ processed_clips = []
548
+ narration_clip = None
549
+ final_clip = None
550
+ logger.info(f"Assembling from {len(asset_data_list)} assets. Frame: {self.video_frame_size}.")
551
+
552
+ for i, asset_info in enumerate(asset_data_list):
553
+ asset_path = asset_info.get('path')
554
+ asset_type = asset_info.get('type')
555
+ scene_dur = asset_info.get('duration', 4.5)
556
+ scene_num = asset_info.get('scene_num', i + 1)
557
+ key_action = asset_info.get('key_action', '')
558
+ logger.info(f"S{scene_num}: Path='{asset_path}', Type='{asset_type}', Dur='{scene_dur}'s")
559
+
560
+ if not (asset_path and os.path.exists(asset_path)):
561
+ logger.warning(f"S{scene_num}: Not found '{asset_path}'. Skip.")
562
+ continue
563
+ if scene_dur <= 0:
564
+ logger.warning(f"S{scene_num}: Invalid duration ({scene_dur}s). Skip.")
565
+ continue
566
+
567
+ current_scene_mvpy_clip = None
568
  try:
569
+ if asset_type == 'image':
570
+ pil_img = Image.open(asset_path)
571
+ logger.debug(f"S{scene_num}: Loaded img. Mode:{pil_img.mode}, Size:{pil_img.size}")
572
+ img_rgba = pil_img.convert('RGBA') if pil_img.mode != 'RGBA' else pil_img.copy()
573
+ thumb = img_rgba.copy()
574
+ rf = Image.Resampling.LANCZOS if hasattr(Image.Resampling, 'LANCZOS') else Image.BILINEAR
575
+ thumb.thumbnail(self.video_frame_size, rf)
576
+ cv_rgba = Image.new('RGBA', self.video_frame_size, (0, 0, 0, 0))
577
+ xo = (self.video_frame_size[0] - thumb.width) // 2
578
+ yo = (self.video_frame_size[1] - thumb.height) // 2
579
+ cv_rgba.paste(thumb, (xo, yo), thumb)
580
+ final_rgb_pil = Image.new("RGB", self.video_frame_size, (0, 0, 0))
581
+ final_rgb_pil.paste(cv_rgba, mask=cv_rgba.split()[3])
582
+ dbg_path = os.path.join(self.output_dir, f"debug_PRE_NUMPY_S{scene_num}.png")
583
+ final_rgb_pil.save(dbg_path)
584
+ logger.info(f"DEBUG: Saved PRE_NUMPY_S{scene_num} to {dbg_path}")
585
+ frame_np = np.array(final_rgb_pil, dtype=np.uint8)
586
+ if not frame_np.flags['C_CONTIGUOUS']:
587
+ frame_np = np.ascontiguousarray(frame_np, dtype=np.uint8)
588
+ logger.debug(f"S{scene_num}: NumPy for MoviePy. Shape:{frame_np.shape}, DType:{frame_np.dtype}, C-Contig:{frame_np.flags['C_CONTIGUOUS']}")
589
+ if frame_np.size == 0 or frame_np.ndim != 3 or frame_np.shape[2] != 3:
590
+ logger.error(f"S{scene_num}: Invalid NumPy. Skip.")
591
+ continue
592
+ clip_base = ImageClip(frame_np, transparent=False).set_duration(scene_dur)
593
+ mvpy_dbg_path = os.path.join(self.output_dir, f"debug_MOVIEPY_FRAME_S{scene_num}.png")
594
+ clip_base.save_frame(mvpy_dbg_path, t=0.1)
595
+ logger.info(f"DEBUG: Saved MOVIEPY_FRAME_S{scene_num} to {mvpy_dbg_path}")
596
+ clip_fx = clip_base
597
+ try:
598
+ es = random.uniform(1.03, 1.08)
599
+ clip_fx = clip_base.fx(
600
+ vfx.resize,
601
+ lambda t: 1 + (es - 1) * (t / scene_dur) if scene_dur > 0 else 1
602
+ ).set_position('center')
603
+ except Exception as e:
604
+ logger.error(f"S{scene_num} Ken Burns error: {e}", exc_info=False)
605
+ current_scene_mvpy_clip = clip_fx
606
+ elif asset_type == 'video':
607
+ src_clip = None
608
  try:
609
+ src_clip = VideoFileClip(
610
+ asset_path,
611
+ target_resolution=(self.video_frame_size[1], self.video_frame_size[0]) if self.video_frame_size else None,
612
+ audio=False
613
+ )
614
+ tmp_clip = src_clip
615
+ if src_clip.duration != scene_dur:
616
+ if src_clip.duration > scene_dur:
617
+ tmp_clip = src_clip.subclip(0, scene_dur)
618
  else:
619
+ if scene_dur / src_clip.duration > 1.5 and src_clip.duration > 0.1:
620
+ tmp_clip = src_clip.loop(duration=scene_dur)
621
+ else:
622
+ tmp_clip = src_clip.set_duration(src_clip.duration)
623
+ logger.info(f"S{scene_num} Video clip ({src_clip.duration:.2f}s) shorter than target ({scene_dur:.2f}s).")
624
+ current_scene_mvpy_clip = tmp_clip.set_duration(scene_dur)
625
+ if current_scene_mvpy_clip.size != list(self.video_frame_size):
626
+ current_scene_mvpy_clip = current_scene_mvpy_clip.resize(self.video_frame_size)
627
+ except Exception as e:
628
+ logger.error(f"S{scene_num} Video load error '{asset_path}':{e}", exc_info=True)
629
+ continue
630
  finally:
631
+ if src_clip and src_clip is not current_scene_mvpy_clip and hasattr(src_clip, 'close'):
632
+ src_clip.close()
633
+ else:
634
+ logger.warning(f"S{scene_num} Unknown asset type '{asset_type}'. Skip.")
635
+ continue
636
+
637
+ if current_scene_mvpy_clip and key_action:
638
  try:
639
+ 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
640
+ to_start = 0.25
641
+ txt_c = TextClip(
642
+ f"Scene {scene_num}\n{key_action}",
643
+ fontsize=self.video_overlay_font_size,
644
+ color=self.video_overlay_font_color,
645
+ font=self.video_overlay_font,
646
+ bg_color='rgba(10,10,20,0.7)',
647
+ method='caption',
648
+ align='West',
649
+ size=(self.video_frame_size[0] * 0.9, None),
650
+ kerning=-1,
651
+ stroke_color='black',
652
+ stroke_width=1.5
653
+ ).set_duration(to_dur).set_start(to_start).set_position(('center', 0.92), relative=True)
654
+ current_scene_mvpy_clip = CompositeVideoClip(
655
+ [current_scene_mvpy_clip, txt_c],
656
+ size=self.video_frame_size,
657
+ use_bgclip=True
658
+ )
659
+ except Exception as e:
660
+ logger.error(f"S{scene_num} TextClip error:{e}. No text.", exc_info=True)
661
+
662
+ if current_scene_mvpy_clip:
663
+ processed_clips.append(current_scene_mvpy_clip)
664
+ logger.info(f"S{scene_num} Processed. Dur:{current_scene_mvpy_clip.duration:.2f}s.")
665
+ except Exception as e:
666
+ logger.error(f"MAJOR Error S{scene_num} ({asset_path}):{e}", exc_info=True)
667
  finally:
668
+ if current_scene_mvpy_clip and hasattr(current_scene_mvpy_clip, 'close'):
669
+ try:
670
+ current_scene_mvpy_clip.close()
671
+ except:
672
+ pass
673
 
674
+ if not processed_clips:
675
+ logger.warning("No clips processed. Abort.")
676
+ return None
677
+ td = 0.75
678
  try:
679
+ logger.info(f"Concatenating {len(processed_clips)} clips.")
680
+ if len(processed_clips) > 1:
681
+ final_clip = concatenate_videoclips(processed_clips, padding=-td if td > 0 else 0, method="compose")
682
+ elif processed_clips:
683
+ final_clip = processed_clips[0]
684
+ if not final_clip:
685
+ logger.error("Concatenation failed.")
686
+ return None
687
+ logger.info(f"Concatenated dur:{final_clip.duration:.2f}s")
688
+ if td > 0 and final_clip.duration > 0:
689
+ if final_clip.duration > td * 2:
690
+ final_clip = final_clip.fx(vfx.fadein, td).fx(vfx.fadeout, td)
691
+ else:
692
+ final_clip = final_clip.fx(vfx.fadein, min(td, final_clip.duration / 2.0))
693
+ if overall_narration_path and os.path.exists(overall_narration_path) and final_clip.duration > 0:
694
+ try:
695
+ narration_clip = AudioFileClip(overall_narration_path)
696
+ final_clip = final_clip.set_audio(narration_clip)
697
+ logger.info("Narration added.")
698
+ except Exception as e:
699
+ logger.error(f"Narration add error:{e}", exc_info=True)
700
+ elif final_clip.duration <= 0:
701
+ logger.warning("Video no duration. No audio.")
702
+ if final_clip and final_clip.duration > 0:
703
+ op = os.path.join(self.output_dir, output_filename)
704
+ logger.info(f"Writing video:{op} (Dur:{final_clip.duration:.2f}s)")
705
+ final_clip.write_videofile(
706
+ op,
707
+ fps=fps,
708
+ codec='libx264',
709
+ preset='medium',
710
+ audio_codec='aac',
711
+ temp_audiofile=os.path.join(self.output_dir, f'temp-audio-{os.urandom(4).hex()}.m4a'),
712
+ remove_temp=True,
713
+ threads=os.cpu_count() or 2,
714
+ logger='bar',
715
+ bitrate="5000k",
716
+ ffmpeg_params=["-pix_fmt", "yuv420p"]
717
+ )
718
+ logger.info(f"Video created:{op}")
719
+ return op
720
+ else:
721
+ logger.error("Final clip invalid. No write.")
722
+ return None
723
+ except Exception as e:
724
+ logger.error(f"Video write error:{e}", exc_info=True)
725
+ return None
726
  finally:
727
+ logger.debug("Closing all MoviePy clips in `assemble_animatic_from_assets` finally block.")
728
+ clips_to_close = processed_clips + ([narration_clip] if narration_clip else []) + ([final_clip] if final_clip else [])
729
+ for clip_obj in clips_to_close:
730
+ if clip_obj and hasattr(clip_obj, 'close'):
731
+ try:
732
+ clip_obj.close()
733
+ except Exception as e_close:
734
+ logger.warning(f"Ignoring error while closing a clip: {e_close}")