CingenAI / core /visual_engine.py
mgbam's picture
Update core/visual_engine.py
3084a6c verified
raw
history blame
38.3 kB
# 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
# --- MoviePy Imports ---
from moviepy.editor import (ImageClip, VideoFileClip, concatenate_videoclips, TextClip,
CompositeVideoClip, AudioFileClip)
import moviepy.video.fx.all as vfx
# --- MONKEY PATCH for Pillow/MoviePy compatibility ---
try:
if hasattr(Image, 'Resampling') and hasattr(Image.Resampling, 'LANCZOS'): # Pillow 9+
if not hasattr(Image, 'ANTIALIAS'): Image.ANTIALIAS = Image.Resampling.LANCZOS
elif hasattr(Image, 'LANCZOS'): # Pillow 8
if not hasattr(Image, 'ANTIALIAS'): Image.ANTIALIAS = Image.LANCZOS
elif not hasattr(Image, 'ANTIALIAS'):
print("WARNING: Pillow version lacks common Resampling attributes or ANTIALIAS. MoviePy effects might fail or look different.")
except Exception as e_monkey_patch:
print(f"WARNING: An unexpected error occurred during Pillow ANTIALIAS monkey-patch: {e_monkey_patch}")
logger = logging.getLogger(__name__)
# logger.setLevel(logging.DEBUG) # Uncomment for very verbose debugging
# --- External Service Client Imports ---
ELEVENLABS_CLIENT_IMPORTED = False
ElevenLabsAPIClient = None
Voice = None
VoiceSettings = None
try:
from elevenlabs.client import ElevenLabs as ImportedElevenLabsClient
from elevenlabs import Voice as ImportedVoice, VoiceSettings as ImportedVoiceSettings
ElevenLabsAPIClient = ImportedElevenLabsClient
Voice = ImportedVoice
VoiceSettings = ImportedVoiceSettings
ELEVENLABS_CLIENT_IMPORTED = True
logger.info("ElevenLabs client components imported successfully.")
except ImportError:
logger.warning("ElevenLabs SDK not found (pip install elevenlabs). Audio generation will be disabled.")
except Exception as e_eleven_import:
logger.warning(f"Error importing ElevenLabs client components: {e_eleven_import}. Audio generation disabled.")
RUNWAYML_SDK_IMPORTED = False
RunwayMLAPIClient = None
try:
from runwayml import RunwayML as ImportedRunwayMLClient
RunwayMLAPIClient = ImportedRunwayMLClient
RUNWAYML_SDK_IMPORTED = True
logger.info("RunwayML SDK imported successfully.")
except ImportError:
logger.warning("RunwayML SDK not found (pip install runwayml). RunwayML video generation will be disabled.")
except Exception as e_runway_sdk_import:
logger.warning(f"Error importing RunwayML SDK: {e_runway_sdk_import}. RunwayML features disabled.")
class VisualEngine:
DEFAULT_FONT_SIZE_PIL = 10
PREFERRED_FONT_SIZE_PIL = 20
VIDEO_OVERLAY_FONT_SIZE = 30
VIDEO_OVERLAY_FONT_COLOR = 'white'
DEFAULT_MOVIEPY_FONT = 'DejaVu-Sans-Bold'
PREFERRED_MOVIEPY_FONT = 'Liberation-Sans-Bold'
def __init__(self, output_dir="temp_cinegen_media", default_elevenlabs_voice_id="Rachel"):
self.output_dir = output_dir
os.makedirs(self.output_dir, exist_ok=True)
self.font_filename_pil = "DejaVuSans-Bold.ttf"
font_paths_to_try = [ self.font_filename_pil, f"/usr/share/fonts/truetype/dejavu/{self.font_filename_pil}", 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.font_path_pil_resolved = next((p for p in font_paths_to_try if os.path.exists(p)), None)
self.font_pil = ImageFont.load_default(); self.current_font_size_pil = self.DEFAULT_FONT_SIZE_PIL
if self.font_path_pil_resolved:
try: self.font_pil = ImageFont.truetype(self.font_path_pil_resolved, self.PREFERRED_FONT_SIZE_PIL); self.current_font_size_pil = self.PREFERRED_FONT_SIZE_PIL; logger.info(f"Pillow font: {self.font_path_pil_resolved} sz {self.current_font_size_pil}."); self.video_overlay_font = 'DejaVu-Sans-Bold' if "dejavu" in self.font_path_pil_resolved.lower() else ('Liberation-Sans-Bold' if "liberation" in self.font_path_pil_resolved.lower() else self.DEFAULT_MOVIEPY_FONT)
except IOError as e_font_load: logger.error(f"Pillow font IOError '{self.font_path_pil_resolved}': {e_font_load}. Default.")
else: logger.warning("Custom Pillow font not found. 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 = None; self.elevenlabs_voice_id = default_elevenlabs_voice_id
if VoiceSettings and ELEVENLABS_CLIENT_IMPORTED: self.elevenlabs_voice_settings = VoiceSettings(stability=0.60, similarity_boost=0.80, style=0.15, use_speaker_boost=True)
else: self.elevenlabs_voice_settings = None
self.pexels_api_key = None; self.USE_PEXELS = False
self.runway_api_key = None; self.USE_RUNWAYML = False; self.runway_ml_client_instance = None
if RUNWAYML_SDK_IMPORTED and RunwayMLAPIClient and os.getenv("RUNWAYML_API_SECRET"):
try: self.runway_ml_client_instance = RunwayMLAPIClient(); self.USE_RUNWAYML = True; logger.info("RunwayML Client init from env var at startup.")
except Exception as e_runway_init_startup: logger.error(f"Initial RunwayML client init failed: {e_runway_init_startup}"); self.USE_RUNWAYML = False
logger.info("VisualEngine initialized.")
def set_openai_api_key(self, api_key): self.openai_api_key = api_key; self.USE_AI_IMAGE_GENERATION = bool(api_key); logger.info(f"DALL-E status: {'Ready' if self.USE_AI_IMAGE_GENERATION else 'Disabled'}")
def set_elevenlabs_api_key(self, api_key, voice_id_from_secret=None):
self.elevenlabs_api_key = api_key
if voice_id_from_secret: self.elevenlabs_voice_id = voice_id_from_secret
if api_key and ELEVENLABS_CLIENT_IMPORTED and ElevenLabsAPIClient:
try: self.elevenlabs_client = ElevenLabsAPIClient(api_key=api_key); self.USE_ELEVENLABS = bool(self.elevenlabs_client); logger.info(f"11L Client: {'Ready' if self.USE_ELEVENLABS else 'Failed'} (Voice: {self.elevenlabs_voice_id})")
except Exception as e: logger.error(f"11L client init error: {e}. Disabled.", exc_info=True); self.USE_ELEVENLABS=False; self.elevenlabs_client=None
else: self.USE_ELEVENLABS = False; logger.info(f"11L Disabled (key/SDK).")
def set_pexels_api_key(self, api_key): self.pexels_api_key = api_key; self.USE_PEXELS = bool(api_key); logger.info(f"Pexels status: {'Ready' if self.USE_PEXELS else 'Disabled'}")
def set_runway_api_key(self, api_key):
self.runway_api_key = api_key
if api_key:
if RUNWAYML_SDK_IMPORTED and RunwayMLAPIClient:
if not self.runway_ml_client_instance:
try:
original_env_secret = os.getenv("RUNWAYML_API_SECRET")
if not original_env_secret: os.environ["RUNWAYML_API_SECRET"] = api_key; logger.info("Temp set RUNWAYML_API_SECRET for SDK.")
self.runway_ml_client_instance = RunwayMLAPIClient(); self.USE_RUNWAYML = True; logger.info("RunwayML Client init via set_runway_api_key.")
if not original_env_secret: del os.environ["RUNWAYML_API_SECRET"]; logger.info("Cleared temp RUNWAYML_API_SECRET.")
except Exception as e: logger.error(f"RunwayML Client init in set_runway_api_key fail: {e}", exc_info=True); self.USE_RUNWAYML=False;self.runway_ml_client_instance=None
else: self.USE_RUNWAYML = True; logger.info("RunwayML Client already init.")
else: logger.warning("RunwayML SDK not imported. Service disabled."); self.USE_RUNWAYML = False
else: self.USE_RUNWAYML = False; self.runway_ml_client_instance = None; logger.info("RunwayML Disabled (no API key).")
def _image_to_data_uri(self, image_path):
# (Implementation from before)
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"};mime_type=mime_map.get(ext,"application/octet-stream");
if mime_type=="application/octet-stream":logger.warning(f"Unknown MIME for {image_path}, using {mime_type}.")
with open(image_path,"rb")as image_file:encoded_string=base64.b64encode(image_file.read()).decode('utf-8')
data_uri=f"data:{mime_type};base64,{encoded_string}";logger.debug(f"Data URI for {os.path.basename(image_path)} (start): {data_uri[:100]}...");return data_uri
except FileNotFoundError:logger.error(f"Img not found {image_path} for data URI.");return None
except Exception as e:logger.error(f"Error converting {image_path} to data URI:{e}",exc_info=True);return None
def _map_resolution_to_runway_ratio(self, width, height):
# (Implementation from before)
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.");return "1280:720"
def _get_text_dimensions(self, text_content, font_object):
# (Implementation from before)
dch=getattr(font_object,'size',self.current_font_size_pil);
if not text_content:return 0,dch
try:
if hasattr(font_object,'getbbox'):bb=font_object.getbbox(text_content);w=bb[2]-bb[0];h=bb[3]-bb[1];return w,h if h>0 else dch
elif hasattr(font_object,'getsize'):w,h=font_object.getsize(text_content);return w,h if h>0 else dch
else:return int(len(text_content)*dch*0.6),int(dch*1.2)
except Exception as e:logger.warning(f"Error in _get_text_dimensions:{e}");return int(len(text_content)*self.current_font_size_pil*0.6),int(self.current_font_size_pil*1.2)
def _create_placeholder_image_content(self,text_description,filename,size=None):
# <<< CORRECTED VERSION OF THIS METHOD >>>
if size is None: size = self.video_frame_size
img = Image.new('RGB', size, color=(20, 20, 40)); d = ImageDraw.Draw(img); padding = 25
max_w = size[0] - (2 * padding); lines = []
if not text_description: text_description = "(Placeholder Image)"
words = text_description.split(); current_line_text = ""
for word_idx, word in enumerate(words):
prospective_addition = word + (" " if word_idx < len(words) - 1 else "")
test_line_text = current_line_text + prospective_addition
current_w, _ = self._get_text_dimensions(test_line_text, self.font_pil)
if current_w == 0 and test_line_text.strip(): current_w = len(test_line_text) * (self.current_font_size_pil * 0.6)
if current_w <= max_w: current_line_text = test_line_text
else:
if current_line_text.strip(): lines.append(current_line_text.strip())
current_line_text = prospective_addition
if current_line_text.strip(): lines.append(current_line_text.strip())
if not lines and text_description:
avg_char_w, _ = self._get_text_dimensions("W", self.font_pil); avg_char_w = avg_char_w or (self.current_font_size_pil * 0.6)
chars_per_line = int(max_w / avg_char_w) if avg_char_w > 0 else 20
lines.append(text_description[:chars_per_line] + ("..." if len(text_description) > chars_per_line else ""))
elif not lines: lines.append("(Placeholder Error)")
_, single_line_h = self._get_text_dimensions("Ay", self.font_pil); single_line_h = single_line_h if single_line_h > 0 else self.current_font_size_pil + 2
max_lines = min(len(lines), (size[1] - (2 * padding)) // (single_line_h + 2)) if single_line_h > 0 else 1
max_lines = max(1, max_lines)
y_pos = padding + (size[1] - (2 * padding) - max_lines * (single_line_h + 2)) / 2.0
for i in range(max_lines):
line_text = lines[i]; line_w, _ = self._get_text_dimensions(line_text, self.font_pil)
if line_w == 0 and line_text.strip(): line_w = len(line_text) * (self.current_font_size_pil * 0.6)
x_pos = (size[0] - line_w) / 2.0
try: d.text((x_pos, y_pos), line_text, font=self.font_pil, fill=(200, 200, 180))
except Exception as e_draw: logger.error(f"Pillow d.text error: {e_draw} for '{line_text}'")
y_pos += single_line_h + 2
if i == 6 and max_lines > 7:
try: d.text((x_pos, y_pos), "...", font=self.font_pil, fill=(200, 200, 180))
except Exception as e_elip: logger.error(f"Pillow d.text ellipsis error: {e_elip}"); break
filepath = os.path.join(self.output_dir, filename)
try: img.save(filepath); return filepath
except Exception as e_save: logger.error(f"Saving placeholder image '{filepath}' error: {e_save}", exc_info=True); return None
def _search_pexels_image(self, query, output_filename_base):
# <<< CORRECTED VERSION OF THIS METHOD >>>
if not self.USE_PEXELS or not self.pexels_api_key: return None
headers = {"Authorization": self.pexels_api_key}
params = {"query": query, "per_page": 1, "orientation": "landscape", "size": "large2x"}
base_name_for_pexels, _ = os.path.splitext(output_filename_base)
pexels_filename = base_name_for_pexels + f"_pexels_{random.randint(1000,9999)}.jpg"
filepath = os.path.join(self.output_dir, pexels_filename)
try:
logger.info(f"Pexels: Searching for '{query}'")
effective_query = " ".join(query.split()[:5])
params["query"] = effective_query
response = requests.get("https://api.pexels.com/v1/search", headers=headers, params=params, timeout=20)
response.raise_for_status()
data = response.json()
if data.get("photos") and len(data["photos"]) > 0:
photo_details = data["photos"][0]
photo_url = photo_details.get("src", {}).get("large2x")
if not photo_url: logger.warning(f"Pexels: 'large2x' URL missing for '{effective_query}'. Details: {photo_details}"); return None
image_response = requests.get(photo_url, timeout=60); image_response.raise_for_status()
img_data_pil = Image.open(io.BytesIO(image_response.content))
if img_data_pil.mode != 'RGB': img_data_pil = img_data_pil.convert('RGB')
img_data_pil.save(filepath); logger.info(f"Pexels: Image saved to {filepath}"); return filepath
else: logger.info(f"Pexels: No photos for '{effective_query}'."); return None
except requests.exceptions.RequestException as e_req: logger.error(f"Pexels: RequestException for '{query}': {e_req}", exc_info=False); return None
except Exception as e: logger.error(f"Pexels: General error for '{query}': {e}", exc_info=True); return None
def _generate_video_clip_with_runwayml(self, text_prompt_for_motion, input_image_path, scene_identifier_filename_base, target_duration_seconds=5):
# (Implementation from previous response, with Runway SDK calls)
if not self.USE_RUNWAYML or not self.runway_ml_client_instance: logger.warning("RunwayML not enabled/client not init. Skip video."); return None
if not input_image_path or not os.path.exists(input_image_path): logger.error(f"Runway Gen-4 needs input image. Path invalid: {input_image_path}"); return None
image_data_uri = self._image_to_data_uri(input_image_path)
if not image_data_uri: return None
runway_duration = 10 if target_duration_seconds >= 8 else 5
runway_ratio_str = self._map_resolution_to_runway_ratio(self.video_frame_size[0], self.video_frame_size[1])
base_name_runway, _ = os.path.splitext(scene_identifier_filename_base); output_video_filename = base_name_runway + f"_runway_gen4_d{runway_duration}s.mp4" # Corrected base name usage
output_video_filepath = os.path.join(self.output_dir, output_video_filename)
logger.info(f"Runway Gen-4 task: motion='{text_prompt_for_motion[:100]}...', img='{os.path.basename(input_image_path)}', dur={runway_duration}s, ratio='{runway_ratio_str}'")
try:
task_submission = self.runway_ml_client_instance.image_to_video.create(model='gen4_turbo', prompt_image=image_data_uri, prompt_text=text_prompt_for_motion, duration=runway_duration, ratio=runway_ratio_str)
task_id = task_submission.id; logger.info(f"Runway Gen-4 task ID: {task_id}. Polling...")
poll_interval=10; max_polls=36; start_poll_time = time.time()
while time.time() - start_poll_time < max_polls * poll_interval:
time.sleep(poll_interval); task_details = self.runway_ml_client_instance.tasks.retrieve(id=task_id)
logger.info(f"Runway task {task_id} status: {task_details.status}")
if task_details.status == 'SUCCEEDED':
output_url = getattr(getattr(task_details,'output',None),'url',None) or (getattr(task_details,'artifacts',None) and task_details.artifacts[0].url if task_details.artifacts and hasattr(task_details.artifacts[0],'url') else None) or (getattr(task_details,'artifacts',None) and task_details.artifacts[0].download_url if task_details.artifacts and hasattr(task_details.artifacts[0],'download_url') else None)
if not output_url: logger.error(f"Runway task {task_id} SUCCEEDED, but no output URL. Details: {vars(task_details) if hasattr(task_details,'__dict__') else task_details}"); return None
logger.info(f"Runway task {task_id} SUCCEEDED. Downloading: {output_url}")
video_response = requests.get(output_url, stream=True, timeout=300); video_response.raise_for_status()
with open(output_video_filepath,'wb') as f:
for chunk in video_response.iter_content(chunk_size=8192): f.write(chunk)
logger.info(f"Runway Gen-4 video saved: {output_video_filepath}"); return output_video_filepath
elif task_details.status in ['FAILED','ABORTED','ERROR']:
em = getattr(task_details,'error_message',None) or getattr(getattr(task_details,'output',None),'error',"Unknown Runway error.")
logger.error(f"Runway task {task_id} status: {task_details.status}. Error: {em}"); return None
logger.warning(f"Runway task {task_id} timed out."); return None
except AttributeError as ae: logger.error(f"RunwayML SDK AttrError: {ae}. SDK/methods changed?", exc_info=True); return None
except Exception as e: logger.error(f"Runway Gen-4 API error: {e}", exc_info=True); return None
def _create_placeholder_video_content(self, td, fn, dur=4, sz=None):
# (Keep as before)
if sz is None: sz = self.video_frame_size; fp = os.path.join(self.output_dir, fn); tc = None
try: tc = TextClip(td, fontsize=50, color='white', font=self.video_overlay_font, bg_color='black', size=sz, method='caption').set_duration(dur); tc.write_videofile(fp, fps=24, codec='libx264', preset='ultrafast', logger=None, threads=2); logger.info(f"Generic placeholder video: {fp}"); return fp
except Exception as e: logger.error(f"Generic placeholder video error {fp}: {e}", exc_info=True); return None
finally:
if tc and hasattr(tc, 'close'): tc.close()
def generate_scene_asset(self, image_generation_prompt_text, motion_prompt_text_for_video,
scene_data, scene_identifier_filename_base,
generate_as_video_clip=False, runway_target_duration=5):
# <<< THIS IS THE CORRECTED METHOD with fixed DALL-E loop >>>
base_name, _ = os.path.splitext(scene_identifier_filename_base)
asset_info = {'path': None, 'type': 'none', 'error': True,
'prompt_used': image_generation_prompt_text,
'error_message': 'Asset generation init failed'}
input_image_for_runway_path = None
base_image_filename = base_name + ("_base_for_video.png" if generate_as_video_clip else ".png")
base_image_filepath = os.path.join(self.output_dir, base_image_filename)
if self.USE_AI_IMAGE_GENERATION and self.openai_api_key:
max_retries = 2; attempt_count_dalle = 0
for attempt_num_dalle in range(max_retries):
attempt_count_dalle = attempt_num_dalle + 1
try: # DALL-E attempt try block
logger.info(f"Attempt {attempt_count_dalle} DALL-E (base img): {image_generation_prompt_text[:70]}...")
client_oai = openai.OpenAI(api_key=self.openai_api_key, timeout=90.0)
response_oai = client_oai.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")
img_url_oai = response_oai.data[0].url
revised_prompt_oai = getattr(response_oai.data[0],'revised_prompt',None)
if revised_prompt_oai: logger.info(f"DALL-E revised: {revised_prompt_oai[:70]}...")
img_response_get = requests.get(img_url_oai,timeout=120); img_response_get.raise_for_status()
pil_img_oai = Image.open(io.BytesIO(img_response_get.content))
if pil_img_oai.mode!='RGB': pil_img_oai=pil_img_oai.convert('RGB')
pil_img_oai.save(base_image_filepath); logger.info(f"DALL-E base img saved: {base_image_filepath}")
input_image_for_runway_path=base_image_filepath
asset_info={'path':base_image_filepath,'type':'image','error':False,'prompt_used':image_generation_prompt_text,'revised_prompt':revised_prompt_oai}
break # Success, exit loop
except openai.RateLimitError as e_rl: logger.warning(f"OpenAI RateLimit Att {attempt_count_dalle}:{e_rl}.Retry...");time.sleep(5*attempt_count_dalle);asset_info['error_message']=str(e_rl)
except openai.APIError as e_api_oai: logger.error(f"OpenAI APIError Att {attempt_count_dalle}:{e_api_oai}");asset_info['error_message']=str(e_api_oai);break
except requests.exceptions.RequestException as e_req_oai: logger.error(f"Requests Err DALL-E Att {attempt_count_dalle}:{e_req_oai}");asset_info['error_message']=str(e_req_oai);break
except Exception as e_gen_oai: logger.error(f"General DALL-E Err Att {attempt_count_dalle}:{e_gen_oai}",exc_info=True);asset_info['error_message']=str(e_gen_oai);break
if asset_info['error']: logger.warning(f"DALL-E failed after {attempt_count_dalle} attempts for base img.")
if asset_info['error'] and self.USE_PEXELS:
logger.info("Trying Pexels for base img.");pqt=scene_data.get('pexels_search_query_감독',f"{scene_data.get('emotional_beat','')} {scene_data.get('setting_description','')}");pp=self._search_pexels_image(pqt,base_image_filename);
if pp:input_image_for_runway_path=pp;asset_info={'path':pp,'type':'image','error':False,'prompt_used':f"Pexels:{pqt}"}
else:current_em=asset_info.get('error_message',"");asset_info['error_message']=(current_em+" Pexels failed for base.").strip()
if asset_info['error']:
logger.warning("Base img (DALL-E/Pexels) failed. Using placeholder.");ppt=asset_info.get('prompt_used',image_generation_prompt_text);php=self._create_placeholder_image_content(f"[Base Placeholder]{ppt[:70]}...",base_image_filename);
if php:input_image_for_runway_path=php;asset_info={'path':php,'type':'image','error':False,'prompt_used':ppt}
else:current_em=asset_info.get('error_message',"");asset_info['error_message']=(current_em+" Base placeholder failed.").strip()
if generate_as_video_clip:
if not input_image_for_runway_path:logger.error("RunwayML video: base img failed.");asset_info['error']=True;asset_info['error_message']=(asset_info.get('error_message',"")+" Base img miss, Runway abort.").strip();asset_info['type']='none';return asset_info
if self.USE_RUNWAYML:
video_path=self._generate_video_clip_with_runwayml(motion_prompt_text_for_video,input_image_for_runway_path,base_name,runway_target_duration)
if video_path and os.path.exists(video_path):asset_info={'path':video_path,'type':'video','error':False,'prompt_used':motion_prompt_text_for_video,'base_image_path':input_image_for_runway_path}
else:logger.warning(f"RunwayML video failed for {base_name}. Fallback to base img.");asset_info['error']=True;asset_info['error_message']=(asset_info.get('error_message',"Base img ok.")+" RunwayML video fail; use base img.").strip();asset_info['path']=input_image_for_runway_path;asset_info['type']='image';asset_info['prompt_used']=image_generation_prompt_text
else:logger.warning("RunwayML selected but disabled. Use base img.");asset_info['error']=True;asset_info['error_message']=(asset_info.get('error_message',"Base img ok.")+" RunwayML disabled; use base img.").strip();asset_info['path']=input_image_for_runway_path;asset_info['type']='image';asset_info['prompt_used']=image_generation_prompt_text
return asset_info
def generate_narration_audio(self, text_to_narrate, output_filename="narration_overall.mp3"):
# (Keep as before)
if not self.USE_ELEVENLABS or not self.elevenlabs_client or not text_to_narrate: logger.info("11L skip."); return None; afp=os.path.join(self.output_dir,output_filename)
try: logger.info(f"11L audio (Voice:{self.elevenlabs_voice_id}): {text_to_narrate[:70]}..."); asm=None
if hasattr(self.elevenlabs_client,'text_to_speech')and hasattr(self.elevenlabs_client.text_to_speech,'stream'):asm=self.elevenlabs_client.text_to_speech.stream;logger.info("Using 11L .text_to_speech.stream()")
elif hasattr(self.elevenlabs_client,'generate_stream'):asm=self.elevenlabs_client.generate_stream;logger.info("Using 11L .generate_stream()")
elif hasattr(self.elevenlabs_client,'generate'):logger.info("Using 11L .generate()");vp=Voice(voice_id=str(self.elevenlabs_voice_id),settings=self.elevenlabs_voice_settings)if Voice and self.elevenlabs_voice_settings else str(self.elevenlabs_voice_id);ab=self.elevenlabs_client.generate(text=text_to_narrate,voice=vp,model="eleven_multilingual_v2");
with open(afp,"wb")as f:f.write(ab);logger.info(f"11L audio (non-stream): {afp}");return afp
else:logger.error("No 11L audio method.");return None
if asm:vps={"voice_id":str(self.elevenlabs_voice_id)}
if self.elevenlabs_voice_settings:
if hasattr(self.elevenlabs_voice_settings,'model_dump'):vps["voice_settings"]=self.elevenlabs_voice_settings.model_dump()
elif hasattr(self.elevenlabs_voice_settings,'dict'):vps["voice_settings"]=self.elevenlabs_voice_settings.dict()
else:vps["voice_settings"]=self.elevenlabs_voice_settings
adi=asm(text=text_to_narrate,model_id="eleven_multilingual_v2",**vps)
with open(afp,"wb")as f:
for c_chunk in adi: # Renamed c to c_chunk
if c_chunk:f.write(c_chunk)
logger.info(f"11L audio (stream): {afp}");return afp
except Exception as e_11l:logger.error(f"11L audio error: {e_11l}",exc_info=True);return None # Renamed e to e_11l
def assemble_animatic_from_assets(self, asset_data_list, overall_narration_path=None, output_filename="final_video.mp4", fps=24):
# (Keep as in the version with robust image processing, C-contiguous array, debug saves, and pix_fmt)
if not asset_data_list: logger.warning("No assets for animatic."); return None
processed_clips = []; narration_clip_mvpy = None; final_composite_video_clip = None # Renamed variables
logger.info(f"Assembling from {len(asset_data_list)} assets. Frame: {self.video_frame_size}.")
for i, asset_info_dict in enumerate(asset_data_list): # Renamed asset_info to asset_info_dict
asset_p, asset_t, scene_d = asset_info_dict.get('path'), asset_info_dict.get('type'), asset_info_dict.get('duration', 4.5)
scene_n, key_act = asset_info_dict.get('scene_num', i + 1), asset_info_dict.get('key_action', '')
logger.info(f"S{scene_n}: Path='{asset_p}', Type='{asset_t}', Dur='{scene_d}'s")
if not (asset_p and os.path.exists(asset_p)): logger.warning(f"S{scene_n}: Not found '{asset_p}'. Skip."); continue
if scene_d <= 0: logger.warning(f"S{scene_n}: Invalid duration ({scene_d}s). Skip."); continue
current_scene_clip_mvpy = None # Renamed current_scene_mvpy_clip
try:
if asset_t == 'image':
# ... (Robust image processing logic from previous full version) ...
pil_img_opened = Image.open(asset_p); logger.debug(f"S{scene_n}: Loaded img. Mode:{pil_img_opened.mode}, Size:{pil_img_opened.size}")
img_rgba_converted = pil_img_opened.convert('RGBA') if pil_img_opened.mode != 'RGBA' else pil_img_opened.copy()
thumb_img = img_rgba_converted.copy(); res_filter = Image.Resampling.LANCZOS if hasattr(Image.Resampling,'LANCZOS') else Image.BILINEAR; thumb_img.thumbnail(self.video_frame_size,res_filter)
canvas_for_rgba = Image.new('RGBA',self.video_frame_size,(0,0,0,0)); x_offset,y_offset=(self.video_frame_size[0]-thumb_img.width)//2,(self.video_frame_size[1]-thumb_img.height)//2
canvas_for_rgba.paste(thumb_img,(x_offset,y_offset),thumb_img)
final_rgb_for_pil = Image.new("RGB",self.video_frame_size,(0,0,0)); final_rgb_for_pil.paste(canvas_for_rgba,mask=canvas_for_rgba.split()[3])
debug_path_pre_numpy = os.path.join(self.output_dir,f"debug_PRE_NUMPY_S{scene_n}.png"); final_rgb_for_pil.save(debug_path_pre_numpy); logger.info(f"DEBUG: Saved PRE_NUMPY_S{scene_n} to {debug_path_pre_numpy}")
numpy_frame = np.array(final_rgb_for_pil,dtype=np.uint8);
if not numpy_frame.flags['C_CONTIGUOUS']: numpy_frame=np.ascontiguousarray(numpy_frame,dtype=np.uint8)
logger.debug(f"S{scene_n}: NumPy for MoviePy. Shape:{numpy_frame.shape}, DType:{numpy_frame.dtype}, C-Contig:{numpy_frame.flags['C_CONTIGUOUS']}")
if numpy_frame.size==0 or numpy_frame.ndim!=3 or numpy_frame.shape[2]!=3: logger.error(f"S{scene_n}: Invalid NumPy. Skip."); continue
image_clip_base = ImageClip(numpy_frame,transparent=False).set_duration(scene_d)
moviepy_debug_frame_save_path=os.path.join(self.output_dir,f"debug_MOVIEPY_FRAME_S{scene_n}.png"); image_clip_base.save_frame(moviepy_debug_frame_save_path,t=0.1); logger.info(f"DEBUG: Saved MOVIEPY_FRAME_S{scene_n} to {moviepy_debug_frame_save_path}")
image_clip_with_fx = image_clip_base
try: end_scale_kb=random.uniform(1.03,1.08); image_clip_with_fx=image_clip_base.fx(vfx.resize,lambda time_t:1+(end_scale_kb-1)*(time_t/scene_d) if scene_d>0 else 1).set_position('center')
except Exception as e_kb: logger.error(f"S{scene_n} Ken Burns error: {e_kb}",exc_info=False)
current_scene_mvpy_clip = image_clip_with_fx
elif asset_t == 'video':
# ... (Video processing logic from previous full version) ...
source_video_file_clip=None
try:
source_video_file_clip=VideoFileClip(asset_p,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=source_video_file_clip
if source_video_file_clip.duration!=scene_d:
if source_video_file_clip.duration>scene_d:temp_video_clip_obj=source_video_file_clip.subclip(0,scene_d)
else:
if scene_d/source_video_file_clip.duration > 1.5 and source_video_file_clip.duration>0.1:temp_video_clip_obj=source_video_file_clip.loop(duration=scene_d)
else:temp_video_clip_obj=source_video_file_clip.set_duration(source_video_file_clip.duration);logger.info(f"S{scene_n} Video clip ({source_video_file_clip.duration:.2f}s) shorter than target ({scene_d:.2f}s).")
current_scene_mvpy_clip=temp_video_clip_obj.set_duration(scene_d)
if current_scene_mvpy_clip.size!=list(self.video_frame_size):current_scene_mvpy_clip=current_scene_mvpy_clip.resize(self.video_frame_size)
except Exception as e_vidload:logger.error(f"S{scene_n} Video load error '{asset_p}':{e_vidload}",exc_info=True);continue
finally:
if source_video_file_clip and source_video_file_clip is not current_scene_mvpy_clip and hasattr(source_video_file_clip,'close'):source_video_file_clip.close()
else: logger.warning(f"S{scene_n} Unknown asset type '{asset_t}'. Skip."); continue
if current_scene_mvpy_clip and key_act: # Text Overlay
try:
text_overlay_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
text_overlay_s_time=0.25
if text_overlay_dur > 0:
text_clip_obj=TextClip(f"Scene {scene_n}\n{key_act}",fontsize=self.VIDEO_OVERLAY_FONT_SIZE,color=self.VIDEO_OVERLAY_FONT_COLOR,font=self.video_overlay_font,bg_color='rgba(10,10,20,0.7)',method='caption',align='West',size=(self.video_frame_size[0]*0.9,None),kerning=-1,stroke_color='black',stroke_width=1.5).set_duration(text_overlay_dur).set_start(text_overlay_s_time).set_position(('center',0.92),relative=True)
current_scene_mvpy_clip=CompositeVideoClip([current_scene_mvpy_clip,text_clip_obj],size=self.video_frame_size,use_bgclip=True)
else: logger.warning(f"S{scene_n}: Text overlay duration zero. Skip text.")
except Exception as e_txtclip:logger.error(f"S{scene_n} TextClip error:{e_txtclip}. No text.",exc_info=True)
if current_scene_mvpy_clip:processed_clips.append(current_scene_mvpy_clip);logger.info(f"S{scene_n} Processed. Dur:{current_scene_mvpy_clip.duration:.2f}s.")
except Exception as e_asset_loop:logger.error(f"MAJOR Error S{scene_n} ({asset_p}):{e_asset_loop}",exc_info=True)
finally:
if current_scene_mvpy_clip and hasattr(current_scene_mvpy_clip,'close'):
try: current_scene_mvpy_clip.close()
except: pass
if not processed_clips:logger.warning("No clips processed. Abort.");return None
transition_val=0.75
try:
logger.info(f"Concatenating {len(processed_clips)} clips.");
if len(processed_clips)>1:final_composite_video_clip=concatenate_videoclips(processed_clips,padding=-transition_val if transition_val>0 else 0,method="compose")
elif processed_clips:final_composite_video_clip=processed_clips[0]
if not final_composite_video_clip:logger.error("Concatenation failed.");return None
logger.info(f"Concatenated dur:{final_composite_video_clip.duration:.2f}s")
if transition_val>0 and final_composite_video_clip.duration>0:
if final_composite_video_clip.duration>transition_val*2:final_composite_video_clip=final_composite_video_clip.fx(vfx.fadein,transition_val).fx(vfx.fadeout,transition_val)
else:final_composite_video_clip=final_composite_video_clip.fx(vfx.fadein,min(transition_val,final_composite_video_clip.duration/2.0))
if overall_narration_path and os.path.exists(overall_narration_path) and final_composite_video_clip.duration>0:
try:narration_clip_mvpy=AudioFileClip(overall_narration_path);final_composite_video_clip=final_composite_video_clip.set_audio(narration_clip_mvpy);logger.info("Narration added.")
except Exception as e_narr:logger.error(f"Narration add error:{e_narr}",exc_info=True)
elif final_composite_video_clip.duration<=0:logger.warning("Video no duration. No audio.")
if final_composite_video_clip and final_composite_video_clip.duration>0:
output_vid_path=os.path.join(self.output_dir,output_filename);logger.info(f"Writing video:{output_vid_path} (Dur:{final_composite_video_clip.duration:.2f}s)")
final_composite_video_clip.write_videofile(output_vid_path,fps=fps,codec='libx264',preset='medium',audio_codec='aac',temp_audiofile=os.path.join(self.output_dir,f'temp-audio-{os.urandom(4).hex()}.m4a'),remove_temp=True,threads=os.cpu_count()or 2,logger='bar',bitrate="5000k",ffmpeg_params=["-pix_fmt", "yuv420p"])
logger.info(f"Video created:{output_vid_path}");return output_vid_path
else:logger.error("Final clip invalid. No write.");return None
except Exception as e_vid_write:logger.error(f"Video write error:{e_vid_write}",exc_info=True);return None
finally:
logger.debug("Closing all MoviePy clips in `assemble_animatic_from_assets` finally block.")
all_clips_to_close_list = processed_clips + ([narration_clip_mvpy] if narration_clip_mvpy else []) + ([final_composite_video_clip] if final_composite_video_clip else [])
for clip_to_close_item in all_clips_to_close_list:
if clip_to_close_item and hasattr(clip_to_close_item, 'close'):
try: clip_to_close_item.close()
except Exception as e_final_close: logger.warning(f"Ignoring error while closing a clip: {type(clip_to_close_item).__name__} - {e_final_close}")